MASARYKOVA
U N I V E R Z I T A
E K O N O M I C K O - S P R Á V N Í F A K U L T A
The Real Problem of
Fake Online Reviews
Diplomová práce
M A T Ě J P R A C H A Ř
Vedoucí práce: doc. Ing. Dušan Mladenovic, Ph.D
Katedra podnikové ekonomiky a managementu
Program Podniková ekonomika a management
Brno 2025
IUI UN I
ECON
T H E REAL PROBLEM OF FAKE O N L I N E REVIEWS
Bibliografický záznam
Autor:
Název práce:
Studijní program:
Vedoucí práce:
Rok:
Počet stran:
Klíčová slova:
MATEJ PRACHAŘ
Ekonomicko-správní fakulta
Masarykova univerzita
Katedra podnikové ekonomiky a managementu
The Real Problem of Fake Online Reviews
Podniková ekonomika a management
doc. Ing. Dušan Mladenovič, Ph.D.
2025
101
eWOM, Online Reviews, Fake Reviews, Review
Authenticity, Booking Intention, Hospitality
2
T H E REAL PROBLEM O F FAKE O N L I N E REVIEWS
Bibliographic record
Author: MATĚJ PRACHAŘ
Faculty of Economics and Administration
Masaryk University
Department of Business Management
Title of Thesis: The Real Problem of Fake Online Reviews
Degree Programme: Business Economy and Management
Supervisor: doc. Ing. Dušan Mladenovič, Ph.D.
Year: 2025
Number of Pages: 101
Keywords: eWOM, Online Reviews, Fake Reviews, Review
Authenticity, Booking Intention, Hospitality
3
T H E REAL PROBLEM OF FAKE O N L I N E REVIEWS
HUN I
E C 0 N
UnSHHTtDVl UMEUEHZI T»
Í Kunám c K n - S P S U T N I I A K U L T I
LI F • v ••• 4 1 * . 602 CC B R H d
l í : D D B 1 Q E S 4
l i t : C I » D Í 1 B ! ! 4
Z A D Á N I
D I P L O M O V É P R Á C E
Akademicky rok 2025'2026
Student: Bů. Malej Prachař
Program Podniková ekonomika a management
N á i c v pracc: Skutečný problém falešných online recenzi
Název p r i a n g l i c k y ; Real Problem of Fake Online Reviews
Cil p r i u , postup a pou
l i l i metody:
Cíl práce:
Cílem letů práce je analyzoval rostoucí problém íaleänyth (podvodných) recenzi.
Zejména se zamatu jo na zkoumáni Sdhůpnosli Spotfebiielú rozpoznat laleärté Online recenze,
a pf esvedčovaci' silu falešných online rocenzf pli vyvolávání nakupn ih* zámeru.
Postup práce a pouzilé melody:
Z metodologicke ho hlediska a v závislosti na tľli výzkumu muža rento výzku m vyui ival jak
kvaniiiativni (napr. kvazioipérimanty, průzkumy, espsrimenly aid.), lak kvalitativní (napi.
rozhovory) metody sboru dal a Opirat £0 0 primárni 1 Sekundárni dala.
Výzku-ria auerda Ly rn-Ĺ-la za'ir'icvat. ". Pich ĽCI rclcvariTi u a-:l_u ÍO'.'U'IC ľlculcry.
2. SpacHikaci metodologickeho prístupu; 3. Analýzu výsledku; 4. Diskusi zjisienľ a záverečné
poznámky.
Rozsah gralických
práci:
Podle pokynů vedoucího práce
Rozsah prače bez pří­
loh:
60 - BO stran
Literatura Kim. J. M.. Park. K. K. C , a Mariáni, M. M. (2023). Do online review readers react rfflerently
when exposed to credible versus lake online reviews?. Journal of Business Research,
154. 113377.
Román, B., Ftiquelme, 1. P., Ä lacobucef. D. (2023). Faké or credible? Antecederrs and
consequences 01 perceived credibility in esaogeraiad online reviews. Journal ol Business
Research. 15Í, 11348*.
Song, y.. wane,, L., Zhang, z., s Hikkeŕova, L. (2023). Do take reviews promote to nsu •
merspurthase intention?. Journal of Business Research, 164.113871.
WalllWr. M.. Jakobi, T.. Watson, S J . . & Stevens. G . (2023). A Ěy/Ětematie lileraluíů review
about the consumers' side ol fake review dateeiion-wtiicri cues do consumers use to
determine tho veracity ol online user reviews?. Computers in Human Behavior Reports.
10, 10027B.
Vedoucí práce: doe. Ino,. Dušan Mladsnovir, Ph.D.
Pracoviště vedoucího
prácí:
Ekonomicko-Ěprávnr lakulta
Katedra podnikové ekonomiky a managementu
4
T H E REAL PROBLEM OF FAKE O N L I N E REVIEWS
Anotace
Předmětem diplomové práce „Skutečný problém falešných online recenzí"
je zkoumání vlivu vybraných signálů v online recenzích na vnímanou
autenticitu a ochotu zákazníků rezervovat ubytování v hotelovém
sektoru. V teoretické části práce jsou vymezeny pojmy word-of-mouth a
eWOM, popsány rozdíly mezi autentickými a falešnými recenzemi a přístupy
k jejich odhalování. Praktická část je založena na experimentálním
dotazníkovém šetření, ve kterém jsou analyzovány čtyři typy signálů
ovlivňujících autenticitu recenzí. V závěru práce jsou shrnuly hlavní poznatky
výzkumu a diskutovány jejich teoretické i manažerské implikace.
6
T H E REAL PROBLEM O F FAKE O N L I N E REVIEWS
Abstract
The subject of the thesis "The Real Problem of Fake Online Reviews" is to
examine the influence of selected cues in online reviews on perceived
authenticity and customers' willingness to book accommodation in the
hotel sector. The theoretical part of the thesis defines the terms word-ofmouth
and eWOM, describes the differences between authentic and fake
reviews, and approaches to detecting them. The practical part is based
on an experimental questionnaire survey, which analyzes four types of
cues influencing the authenticity of reviews. The conclusion of the thesis
summarizes the main findings of the research and discusses their theoretical
and managerial implications.
7
T H E REAL PROBLEM OF FAKE O N L I N E REVIEWS
Declaration of generative Al use
During the preparation of this thesis, the author used ChatGPT (version
5.1) to draft the experimental review versions. ChatGPT and Perplexity
were also used to improve the clarity, grammar, and language of the thesis,
to support the identification of relevant literature, and to assist with
the organization and processing of questionnaire data. All AI assisted
outputs were reviewed, edited, and verified by the author, who takes full
responsibility for the final content of the thesis.
8
T H E REAL PROBLEM O F FAKE O N L I N E REVIEWS
Declaration
Prohlašuji, že jsem diplomovou práci na téma The Real Problem of Fake
Online Reviews vypracoval samostatně pod vedením doc. Ing. Dušana
Mladenoviče, Ph.D. a uvedl v ní všechny použité literární a jiné odborné
zdroje v souladu s právními předpisy, vnitřními předpisy Masarykovy
univerzity a vnitřními akty řízení Masarykovy univerzity a Ekonomicko-správní
fakulty MU.
Brno December 15, 2025
MATĚJ PRACHAŘ
9
T H E REAL PROBLEM O F FAKE O N L I N E REVIEWS
Acknowledgements
I would like to express my sincere thanks to my thesis supervisor doc.
Ing. Dušan Mladenovič, Ph.D., for his guidance, valuable feedback, and expert
advice, which significantly contributed to the completion of this thesis.
I would also like to thank my family and friends for their continuous
support and encouragement throughout the writing of this thesis and
during my entire course of study.
5 Masarykova univerzita 11
T A B L E OF CONTENTS
Table of Contents
List of Figures 15
List of Tables 16
Glossary 17
List of Appendices 18
1 Introduction 19
2 Theoretical background 22
2.1 WOMandeWOM 22
2.2 Fake and authentic online reviews 23
2.3 Understanding the motivation for fake review creation 31
2.4 Uncovering Fake Online Reviews 34
3 Methodology 41
3.1 Research design 41
3.2 Data collection 46
3.3 Data analysis 47
4 Results 49
4.1 Demographics 49
4.2 Scale reliability 51
4.3 Manipulation checks 51
4.4 Group comparisons: Perceived authenticity 53
4.5 Group comparisons: Booking intention 58
4.6 Relationship between perceived authenticity and booking
intention 63
5 Discussion 68
6 Theoretical contributions and managerial implications 72
7 Limitations and future research 75
13
T A B L E O F CONTENTS
8 Conclusion 77
Bibliography 79
Appendix A Questionnaire copy 91
Appendix B Review versions 96
Appendix C Generative AI prompts 98
14
LIST OF FIGURES
List of Figures
Figure 1: Example of possible opinionated opportunism 25
Figure 2: Example of review showing signs of non-immediacy, excessive
emotiveness and lack of specificity 27
Figure 3: Example of a reviewer using an incoherent username 30
Figure 4: Example of spam accounts with no profile picture 31
Figure 5: Example of an overly negative fake online review 34
Figure 6: Example of a credible reviewer profile 36
Figure 7: Fake reviewer posing as actor Chris Hemsworth 38
Figure 8: Conceptual framework 45
Figure 9: Mean perceived authenticity across conditions 54
Figure 10: Games-Howell post-hoc comparisons for perceived
authenticity 58
Figure 11: Mean booking intention across conditions 60
Figure 12: Games-Howell post-hoc comparisons for booking intention
63
Figure 13: Pearson correlation between perceived authenticity and
booking intention 64
15
LIST OF T A B L E S
List of Tables
Table 1: Sub-items used in generative review creation 43
Table 2: Measurement of constructs 44
Table 3: Sample demographics 50
Table 4: Manipulation check results for each condition 53
Table 5: Descriptive statistics for perceived authenticity 54
Table 6: Welch ANOVA results 56
Table 7: Descriptive statistics for booking intention 59
Table 8: One-way ANOVA results 61
Table 9: Effects of experimental conditions on perceived authenticity 65
Table 10: Mediation model predicting booking intention 66
Table 11: Bootstrapped indirect effects via perceived authenticity 67
16
GLOSSARY
Glossary
WOM - Word-of-mouth
eWOM - Electronic Word-of-Mouth
AI - Artificial Intelligence
ML - Machine Learning
MC - Manipulation Check
ANOVA - Analysis of Variance
BO - Baseline condition
SI - Specificity condition
VI - Validation condition
01 - Overlap condition
11 - Implausibility condition
17
LIST O F APPENDICES
List of Appendices
Text Appendices
Appendix A Questionnaire copy 91
Appendix B Review versions 96
Appendix C Generative AI prompts 98
18
INTRODUCTION
1 Introduction
In today's day and age of digital transformation, online reviews have
become an integral part of consumer decision-making across multiple
commercial sectors. Nowadays, potential customers rely ever so heavily
on these reviews as credible feedback and support for future decisions
in their purchasing behaviour, with plenty of research indicating that a
very significant percentage of consumers opt to first read online reviews
before making a purchase (Le Loc Tuan Ly etal., 2022), establishing them
as a cheap, yet powerful marketing tool (Hu et al., 2012), essentially
forcing businesses to maintain a positive online presence to prevent loss
of sales and sustain their competitive edge.
The prevailing importance of keeping a positive "review score" is
further underlined by various researchers. If we were to quantify this
impact, according to Zhang et al. (2016), up to 80% of consumers
reconsider their initial intention to purchase upon encountering
negative reviews, whereas 87% confirm their decision when exposed to
positive consumer feedback. Moreover, review-driven orientation and
choice-making have elevated online reviews to the position of the second
most reliable source of information, surpassed only by suggestions made
by family and friends (Salehan & Kim, 2016).
It can be said with great confidence that online reviews, whether
positive or negative, wield a significant influence on businesses and
customers alike, and are a force to be reckoned with in the marketplace
(Fernandes et al., 2022). However, not all sectors are influenced the
same, with some showing a greater deal of review dependence compared
to others. Porrit (2017) identifies the hospitality sector as the most
susceptible to online reviews, noting that more consumers seek
evaluations of hotels and restaurants than any other type of business. Of
those actively searching for reviews, 33% of people admit they would not
visit a restaurant with a less than 4-star rating. Bulchand-Gidumal et al.
(2023) elucidate this phenomenon by adding that consumers are often
more likely to seek reviews for services, due to their intangible nature,
making them more prone to uncertainly compared to mass-produced
goods.
Unfortunately, the growth and expansion of online reviews have
been accompanied by a corresponding increase in the unsolicited
phenomenon of fake online reviews, affecting customers, businesses, and
19
INTRODUCTION
the overall efficiency of the market (Wu et al., 2020), as various
companies, feeling the need to maintain a positive online image, have
resorted to unethical practices, such as manipulating and posting fake
online reviews to deceive customers, despite potential legal backlash
(Plotkina et al., 2020). The motivation behind this manipulation is
evident, as it is well-established that there is a strong correlation
between hotels actively managing their online reputation, managing
positive online ratings, and improving performance indicators, such as
booking rates and revenue per room (Gabbard, 2023).
However, artificial inflation of such ratings through review
manipulation may lead to customer dissatisfaction, feelings of betrayal,
and the aforementioned ineffectiveness and loss of trust in online
customer feedback. To mitigate the adverse impacts of fake reviews,
proactive measures could be implemented to educate consumers on the
issue, such as informing them about the prevalence of fake reviews, along
with equipping them with various methods of detection, enabling them
to make more educated choices and alter their purchasing behavior
accordingly (Huang et al., 2023).
In conclusion, the importance of online reviews and their profound
influence on consumers is undeniable. Due to the ever-growing realm of
e-commerce, online reviews exert a greater impact on more customers
than ever, shaping their decisions and purchasing behavior. Regrettably,
the rising prevalence of fake reviews poses an unprecedented challenge,
capable of disturbing market effectiveness and instilling distrust in
customers, jeopardizing the credibility of online reviews as a whole.
The aim of this master's thesis is to analyze the specifics of the
problem of fake online reviews within the setting of the Czech Republic.
Given the influence of online reviews on the hospitality sector, where
consumers are most likely to read reviews before completing a purchase
(Porrit, 2017), this thesis will explore the problem of fake reviews in this
domain. The thesis consists of a theoretical part, providing a
comprehensive review of existing literature on the topic, and a practical
part, exploring consumers' perceptions of online review authenticity,
through the analysis of both primary and secondary data.
The aim of the thesis will be fulfilled through answering the
following research questions:
20
INTRODUCTION
RQl: How do various review cues affect perceived authenticity?
RQ2: How does perceived authenticity influence booking intention?
RQ3: What are the different types of fake online reviews?
RQ4: What are the key aspects of differentiating between fake and
authentic online reviews?
21
THEORETICAL BACKGROUND
2 Theoretical background
2.1 WOM and eWOM
For this thesis discussing online reviews, the concept of eWOM, or electronic
word of mouth will be widely used and discussed. However, to
begin with, it is important to outline the basics of its "predecessor", the
classical concept of WOM, or word of mouth. As stated by Dellarocas
(2003), WOM is regarded as one of the oldest forms of information sharing.
It has been defined in an array of ways over time. According to an
early definition by Katz & Lazarsfeld (1966), WOM is the sharing of marketing-related
information among customers, which is crucial in determining
how they behave and how they feel about goods and services. The
concept of WOM was further explored by Arndt (1967), who characterized
it as a person-to-person communication channel in which a communicator
exchanges information with a recipient, who interprets it as noncommercial,
especially when discussing brands, goods, or services. It is
due to this absence of a commercial incentive, that people perceive this
interpersonal mode of communication as more reliable. It is frequently
agreed upon that WOM is one of the most important drivers of customer
behaviour, particularly in circumstances concerning intangible goods.
This is especially apparent in sectors where previous customer experience
is essential to decision-making such as hospitality and tourism
(Daugherty & Hoffman, 2014). Furthermore, studies regularly demonstrate
that consumers view word-of-mouth (WOM) as more reliable and
credible than traditional media sources (Cheung & Thadani, 2012), underscoring
WOM's dependability as a communication tool in influencing
consumer decisions. However, WOM generally takes place in face-to-face
or personal conversations in which participants are constrained by privacy
and proximity, making it more difficult to share experiences and reviews
among a larger scale of consumers (King et al., 2014).
Naturally, with the expansion of technology, namely the internet, a
new kind of word-of-mouth, called eWOM has arisen, collecting and
spreading the opinions of more customers than ever before. In ways that
conventional, localized WOM could not accomplish, this digital progression
has allowed customers all over the world to access and express their
thoughts and opinions on certain products and services. One of the most
recent definitions of eWOM, labelled as eWOM 2.0, highlighting the
22
THEORETICAL BACKGROUND
multi-directional highly dynamic communication between all interested
parties is as follows: "eWOM 2.0 communication refers to digital interactions
occurring between consumers and various stakeholders (e.g., including
brands, experts, and fellow consumers), which are initiated by consumer-generated
online content concerning a product, service, or brand."
(Liu et al., 2024). Upon review of relevant literature, it is clear there are
many distinctions between WOM and eWOM, extending beyond the different
environments in which they are transmitted. According to HueteAlcocer
(2017), information shared through eWOM is more readily accessible
and can spread rapidly and at any time. However, such messages
often tend to be perceived as less credible, due to their anonymous nature.
The author also notes a difference in privacy, noting that these messages
can be accessed by anyone, anytime and from anywhere. Ly and
Huyen (2023) add, that it is this difference in accessibility and speed of
diffusion that can lead to a higher impact on purchase intention as opposed
to traditional WOM. However, in cases where eWOM lacks credibility,
due to factors such as fake, biased, or unreliable reviews, customer
purchase intentions tend to decline, highlighting the vitality of credibility
and its impact on purchase decisions within the eWOM context. Recent
conceptual work further suggests that WOM is moving beyond traditional
text-based eWOM and toward Al-mediated "synthetic WOM,"
where generative systems curate or generate recommendation messages
(Mladenovic et al., 2024) and immersive "metaWOM" in
Metaverse-like environments (Mladenovic et al., 2023), both developments
underlining that questions of WOM authenticity and credibility
will likely persist, and perhaps intensify, as review communication
moves into richer virtual or Al generated settings.
2.2 Fake and authentic online reviews
As already mentioned above, online reviews have evolved into an important
driver of consumer behavior, especially within the hospitality industry.
When it comes to choosing where to go, eat, or stay, customers
mostly rely on these evaluations for information (Xiao et al., 2021). It is
often the case, that consumers consider online reviews, whether positive
or negative, to be more influential than advertisements because of their
perceived honesty and relatability (Soares et al., 2022). Due to their influence,
the authenticity of online reviews remains a crucial factor to be
23
THEORETICAL BACKGROUND
considered, as inauthentic feedback can quickly erode consumer trust
and undermine the overall effectiveness of these reviews, emphasizing
the importance of differentiating between fake and authentic reviews.
(Balakrishnan et al., 2024).
Fake online reviews are often characterized as deceptive evaluations,
deliberately created to mislead customers with the intent of either
damaging or enhancing a product's reputation, effectively tampering
with consumer's purchase decisions (Wang et al., 2023). As per a study
conducted by Istanbulluoglu & Harris (2023), fake reviews can be classified
into the following categories:
• Equity equalizing - Deliberately faked positive reviews
meant to "balance" or offset negative ones, which is
sometimes viewed as unjust. Reviewers defend this as an
attempt to restore equity and save a business they believe is
being treated unfairly.
• Friendly flattery - Reviews, that are exaggerated but based
on actual experiences. The reviewer uses hyperbole to
compliment or flatter, usually to benefit friends or
acquaintances connected to the business.
• Opinionated opportunism - Exaggerated or negatively
slanted reviews created to serve the reviewer's personal
interests. Reviewers may use the unfavourable review as
leverage to demand payment, discounts, or even just
retribution for what they perceive as an injustice.
• Malicious profiteering - Entirely fabricated and intended
to hurt a company for the author's own benefit. The goal is
often to press the business into addressing the false
allegations in order to receive payment or other benefits.
24
THEORETICAL BACKGROUND
Figure 1: Example of possible opinionated opportunism
Bill D Pope Jr Pediatric Dentistry
5124 N 10th St. McAllen. TX
2.2 91 reviews ©
Local Guide 18 reviews
If -k -k * * a month ago
This place reminds of a dental office that's from the 50s. Super outdated.
lb Like
Response from the owner a month ago
Hi Isabella. We'd like to apologize for your experience We make an effort to delight all our patients and
we re sorry we didnt meet your expectations Please give us a call at (956) 630-6026 se^we can make
this right - Regards. Bill 0 Pope Jr Pediatric Dentistry ^
Source: DePompa and Molina, 2022
It is no wonder, that whether positive or negative, deceptive online
reviews are a force to be reckoned with. As research suggests, nearly 74
% of online consumers read reviews before making a purchase
(Daneshyari, 2023). Zhang et al. (2016) further note, that around 87 %
of potential customers tend to reaffirm their decision to purchase a product
after reading positive reviews, while nearly 80 % of shoppers tend to
reconsider shopping for a product or service, upon reading negative
ones. Furthermore, while estimates tend to vary, research suggests that
around 20-30 % of reviews posted online may be fake (Sanjay et al.,
2024). Based on this data, it is fairly clear that the issue of fake reviews
is quite widespread, and that the likelihood of a consumer stumbling
upon them and being influenced by them is significant, making it ever so
important to be able to successfully distinguish an authentic review from
a fake one.
How can we then differentiate between authentic and fake reviews?
Despite this process being far from straightforward, research suggests
that there are distinguishable patterns or "cues", that can help distinguish
them, helping businesses and customers navigate the online review
environment more effectively. Research conducted by Li et al.
(2020) suggests, that the main cues that we should look for when contemplating
the authenticity of a review are cues related to the structure
25
THEORETICAL BACKGROUND
of the review itself, or linguistic cues and cues that focus more on the author
of the review. The former can be then further subdivided into general
linguistic cues, observing aspects such as usage of certain words,
punctuation, and review structure, and psychological cues, relating to the
deeper mental and emotional processes expressed in the review text, reflecting
patterns associated with deceptive behaviors.
2.2.1 General linguistic cues
As mentioned before, general linguistic cues tend to focus on the review
itself. Abri et al. (2020) have constructed a pattern of analysis of transcribed
textualfeatures to identify fake reviews. While analysing text, the
following features are often investigated:
• Quantity - Number of words, sentences, specific phrases.
Fake reviews often contain filler words or redundant infor-
mation.
• Complexity - The structure of sentences and clauses, where
fake reviews might show overly complex or simple
structures.
• Non-immediacy - Distancing language such as passive voice
or vague terms, to make less direct statements about the
product.
• Expressiveness/emotiveness - Level of emotionality,
normally computed by counting adjectives and adverbs
against verbs and nouns.
• Diversity - Indicates the range of unique words. Fake
reviews display less diversity, which may indicate
automated inauthentic reviews.
• Informality - The degree to which casual language is used,
including the ratio of typos to number of words.
• Specificity - Contains information on time, place, and
sensory characteristics. In order to seem more genuine, fake
reviews may be too descriptive or they may lack specificity
altogether.
26
THEORETICAL BACKGROUND
Figure 2: Example of review showing signs of non-immediacy, excessive
emotiveness and lack of specificity
it it it it it Awesome
By ProductsGalore Customer on 20 January 2017
Verified Purchase
This product is unique and revolutionary.
Everyone should buy this.
• Comment
Was this review helpful to you? |Yes| | No | Report abuse
Source: wikiHow Staff, 2019
A study conducted by Wang et al. (2021) suggests, that fake reviews
do indeed often exhibit more exaggerated emotional content, as opposed
to authentic ones, with the authors often using more emotional language
with the intention of misleading readers. This claim is further supported
by Chua & Chen (2022), who also report on these claims of emotional
exaggeration, where reviewers tend to frequently use strong language,
to convey emotions such as surprise or anger. Authors of fake reviews
may use various forms of exaggeration to appear more credible, using
"exaggerators" like exclamation points and firm words, such as "always"
or "never" (Banerjee & Chua, 2017). However, in a later study, Banerjee
(2022) states, that the belief that fake reviews are always more exaggerated
than authentic ones may not always be correct, especially in the hospitality
setting where context plays a large role as well. Factors such as
the type of hotel, whether luxury or budget, greatly influence the level of
exaggeration in both fake and authentic reviews. Wang & Kuan (2022)
further state, that while fake reviews often mirror authentic ones on the
message level, corresponding with overall credible, general themes, they
often fail on the formulation level, where variations in readability, pronoun
usage, and sentence structure expose their dishonest origin. Fake
reviews also frequently fail to include particular experiences that would
provide the review legitimacy, and they are vague and excessively broad
(Carr & Piercy, 2021). Additionally, false reviews often leave out
27
THEORETICAL BACKGROUND
information regarding the product's quality, price, and subsequent customer
service (Petrescu etal., 2018).
2.2.2 Psychological cues
The presence of psychological cues in fake online reviews had been
widely explored, particularly through the lens of interpersonal deception
theory, categorizing said cues into affective cues, cognitive cues, social
cues, and perceptual cues (Li et al., 2020; Wang & Kuan, 2022).
Affective or also emotional cues can be referred to as affective expressions
within communication (De Paulo etal., 2003). According to the
interpersonal deception theory, deceitful individuals typically exhibit
more intense emotional expressions, including anxiety and guilt, than
honest individuals (Porter etal., 2012; Vrij etal., 2019). These emotions,
also labelled as "leakage" or "non-strategic" however, tend to occur primarily
in face-to-face communication, as opposed to thought-out online
review creation, where reviewers often employ more strategic emotions,
like purposefully using more emotive language to attract potential clients
or using negative emotions to harm rivals' reputations while emphasizing
positive emotions to draw attention to a company's advantages
(Wang et al., 2022).
Cognitive cues refer to indicators in communication, that reveal the
level of cognitive processing while a person is writing or speaking and
are linked to markers of mental effort, such as response latency, certainty,
and coherence (De Paulo et al., 2003). While, based on interpersonal
deception theory, dishonest individuals do express a higher
amount of mental effort as opposed to truth-tellers (Burgoon & Buller,
2015), studies conducted by Li etal. (2020) and Wang etal. (2022) argue,
that in the online environment, the relationship between cognitive cues
and review authenticity is not significant and consequently not fundamental
to distinguishing between fake and authentic reviews.
Social cues signify language cues, that include reference to family,
friends, gender, and social activities (Pennebaker et al., 2015). Li et al.
(2020) state, that numerous studies have discovered, that individuals
who are lying frequently utilize third-person pronouns as opposed to
first-person ones to make their message less personal and make it easier
to deceive. This has been noted in both in-person and online conversa-
tions.
28
THEORETICAL BACKGROUND
Lastly, perceptual cues include statements about a person's sensory
experiences, such as seeing, feeling tasting and smelling. These cues are
frequently used to communicate various sensory impressions of hospitality
services, such as the tastes, scents, and presentation of food in restaurants
(Pennebaker et al., 2015). As per Li etal. (2020), perceptual expressions
are quite frequently used in online reviews in the hospitality
sector, as it is more challenging to describe the aforementioned impressions
online than in face-to-face communication. Concerning fake reviews,
Banerjee and Chua (2014) have found these reviews to contain
more visual and touch-related cues as opposed to their authentic counterparts.
Hauch et al. (2015) further add that perceptual cues are an essential
tool in identifying fake reviews, as these reviews are often written
from imagination, rather than genuine experience of the reviewed businesses,
arguing that adding perceptual cues convincingly would be problematic
and that such cues may be more easily recognized as not authentic
and fabricated.
2.2.3 A person leaving the review
As stated previously by Li et al. (2020), to differentiate between authentic
and fake reviews, not only can we examine the language and the structure
of the review itself, but it is also advisable to inspect the credibility
of the author posting the review. This may include factors such as the
locality of the author to the reviewed establishment, where local reviewers
were found to be more likely to post fake reviews, as they often have
more in-depth information about the restaurant, allowing them to more
easily create convincing fabricated content.
Another way to assess the credibility of a reviewer is through their
username. In the case that a profile is using a nonsensical username, often
containing arbitrary groupings of numbers and letters, it may mean
that this profile is one or one of a group of profiles created to post fake
reviews. When one review under such a username is detected, it is usually
enough to raise suspicion. This suspicion is intensified when there
are more reviews present, each with a nonsensical username, within the
same set of reviews. This pattern often suggests that the reviews may
have been generated by bots or fake accounts rather than genuine customers
(Harrison-Walker & Jiang, 2023).
29
THEORETICAL BACKGROUND
Figure 3: Example of a reviewer using an incoherent username
These is the gratest headphones ever!!! Super high-quality sound.
You can connect without worry to your Bluetooth (including Bose
SimpleSync technology). The headphones deliver up to 20 hours
of wireless play, you should definitely buy one
m b 7 3 9 1
Age
Gender
Member since Apr 2021
2 reviews
5.0 average review score
0 people found this helpful
Source: Bassig, n.d.
Further cues assessing the authenticity of the reviewer may include
whether or not the profile is accompanied by a profile picture. As Fong
(2022) states in his study, it is often the case that profiles responsible for
writing fake reviews tend to not have a profile picture attached to them,
and in the case that they do, they are likely to either display the face of
others or not display a face at all. The author argues that these reviewers
are aware of their misdemeanour and deliberately decide not to provide
a profile picture to protect their real identity. While assessing the authenticity
of a review, it is also important to consider the motivating factors
that led the reviewer to create it (Harrison-Walker & Jiang 2023).
30
THEORETICAL BACKGROUND
Figure 4: Example of spam accounts with no profile picture
Joseph Garcia
1 review
if if if if if 2 hours ago
Customer service is clear and kind. They do care users.
Amos Ferrell
1 review
if if if if if 2 hours ago
Customer service is clear and kind. They do care users.
| f c Like
Source: Blackwood, 2020
To summarize, assessing the credibility of a review may require
looking beyond its content and focusing on the reviewer. Factors, such as
the author's locality, username, or the presence of details such as profile
pictures, can offer important clues about a review's authenticity.
2.3 Understanding the motivation for fake review
creation
As mentioned above, individuals may be motivated by a large
scope of diverse incentives when writing fake reviews, with each of them
relating to different underlying motives. To get a grasp of the motivation
behind fake review creation, it is important to examine these motives
themselves. As stated by Zaman et al. (2023) and others, in line with selfdetermination
theory, this motivation can be classified into intrinsic and
extrinsic categories (Malik et al., 2024; Victor et al., 2024).
• Intrinsic motivation - Refers to such behavior, that is
motivated by each individual's own internal sense of curiosity
or satisfaction, uninfluenced by external rewards.
31
THEORETICAL BACKGROUND
In the context of fake online reviews, intrinsic motivation
may stem from personal emotion towards a brand or a certain
establishment, such as strong like or dislike (Zaman et
al., 2023). Furthermore, research conducted by Blank (2015)
discusses how performative elements and humour in online
spaces can also serve as intrinsic motivators for fake reviews.
• Extrinsic motivation - Refers to actions or behaviour influenced
by external factors, such as motivations to receive financial
incentives, complimentary items, and various other
benefits. Studies indicate that monetary rewards tend to
lower moral emotions such as guilt in individuals, making
them more prone to engaging in dishonest behaviors (Malik
et al., 2024).
As stated, the motivation to engage in fake review posting can
emerge from various sources, stemming either from inside or outside
influence and may persuade individuals to post either positive or negative
reviews.
2.3.1 Motivation for posting positive fake reviews
The motivation for posting positive fake reviews often stems from
various extrinsic motives, such as the ones listed above. In today's competitive
environment, companies may resort to endorsing positive fake
review posts to gain an advantage.
This phenomenon is more prevalent among companies, offering
lower-quality products, as opposed to those selling higher-quality ones
(Wu & Qiu, 2023). One of the strategies these companies might employ
is known as "sock puppet", describing a situation, where people within
the company itself are responsible for creating fake accounts and posting
positive reviews (Schoolov, 2020). Oftentimes, in saturated environments
such as the hospitality sector, businesses are often likely to resort
to using positive fake reviews, as they feel the need to maintain high ratings
in order to keep their competitive edge. As Zhang et al. (2022) have
found, restaurants are often more likely to solicit positive fake reviews
when their competitors boast a high number of positive reviews, potentially
causing a cycle of deceiving customers and artificially boosting perceived
value. The process of businesses offering individuals some form
of incentive for posting positive evaluations of their brand or product is
32
THEORETICAL BACKGROUND
quite widespread, as confirmed by He et al. (2021). They have found, that
on Facebook alone, many groups, where businesses recruit fake reviewers
exist. On average, each of these groups has around 16,000 members,
with each group producing almost 600 fake review postings per day. In
these postings, sellers often encourage people to buy their product and
leave a five-star review, in return for a full refund and sometimes even
an additional payment on top. Victor et al. (2024) state, that when incentivized,
individuals with egoistic moral frameworks are more likely to
post fake positive reviews, perceiving this process as a low-risk method
to gain rewards without facing significant backlash. It is often due to the
positive nature of these reviews, that individuals see this practice as
purely a way to support businesses, enabling them to more easily justify
their actions and escape feelings of guilt (Malik et al., 2024).
Reviewers can also write fake positive reviews without the promise
of any form of compensation, purely out of their own intrinsic beliefs. For
example, customers experiencing a phenomenon known as "brand love"
have been known to deliberately manipulate positive reviews, to feel
useful and helpful to their favourite brand (Thakur et al., 2018). It has
also been observed that customers may leave fake positive reviews for
businesses they believe are operating more ethically and morally, or for
those with which they have any form of social ties, such as family or
friends (Zaman etal., 2023).
2.3.2 Motivation for posting negative fake reviews
As with positive reviews, extrinsic motivation can also be an influence
in the publishing of negative fake reviews, with some of the main
factors of gain being financial benefits and social status (Zaman et al.,
2023). Studies often show, that particularly in the accommodation industry,
customers do tend to exaggerate or even completely fabricate negative
experiences, to seek some form of compensation, whether in the
form of financial reimbursement or any form of complimentary offerings,
such as free meals, stays, and other benefits (Gossling et al., 2018). Further
research by Moon etal. (2021) identifies additional motives, such as
direct financial compensation from companies aiming to harm their competition,
as well as motivations in customers with egoistical orientations,
who write negative reviews in an attempt to enhance their social status
by showcasing expertise and opinion superiority, despite the review not
being necessarily valid.
33
THEORETICAL BACKGROUND
Intrinsic motives are also present amid negative reviews, often
stemming from feelings of anger or betrayal among enraged customers
seeking revenge (Thakur et al., 2018). These individuals may leave negative
reviews to damage a business's reputation and ultimately attempt
to make the business fail, with the fake negative review serving as a way
to release their anger (Wu et al., 2020).
Figure 5: Example of an overly negative fake online review
We visited your restaurant on 29 February. My husband ordered a halloumi
burger and I had the chicken pie. The staff were rude, my husband's food
was cold, and to top it all off it was way overpriced
Source: Bromley, 2022
2.4 Uncovering Fake Online Reviews
As can be stated based on the literature review conducted so far, the
problem of fake online reviews has been growing ever so significantly,
planting doubt in consumers' minds, diminishing their trust in online reviews,
and subsequently unfairly damaging purchase intentions and
businesses as a whole. It is for these reasons, that it is crucial to be able
to confidently and reliably distinguish between authentic and fake online
reviews. In recent years, research has focused on machine-based detection
techniques, relying on algorithms and large data sets to confidently
label a review as genuine or fake. However, due to the consumer's inadequate
awareness of this topic and the limited availability of machinebased
support tools, it is also important for customers to be able to identify
fake reviews and for researchers to understand the processes behind
such recognition (Walther et al., 2023). The following sections provide
an analysis of the differences between human and machine-based review
detection.
34
THEORETICAL BACKGROUND
2.4.1 Human detection
Unlike machine-based detection, relying on automated processes,
human detection relies on a sense of judgment, knowledge, and past experience
to successfully recognize cues, suggesting that a review is authentic
or fake. Whereas automated detection models have been able to
achieve accuracy rates of around 90 % on a hotel dataset, and nearly 94
% on a restaurant dataset (Ren et al., 2023), humans accuracy remains
notably lower, ranging from about 60-80 % (Shukla et al., 2019), which
is more akin to random guessing rather than informed, deliberate decisions
(Plotkina et al., 2020). There are multiple suspected reasons for
this, Wang et al. (2015) mention the consumer's often low determination
and effort put into gathering relevant information, which would help
them better identify authenticity, while Siddiqi et al. (2020) name other
factors, such as the restrictive nature of our cognitive abilities, namely
the limits of our attention span, short-term memory, and capacity to digest
a certain amount of information. Drawing inspiration from recent
studies by Harrison-Walker & Jiang (2023) and Bucher (2024), the following
cues were selected to be examined as factors influencing the perceived
authenticity of reviews in human detection.
Cues that a review is authentic
Based on a search of the literature mentioned above (HarrisonWalker
& Jiang, 2023), the following cues were selected as factors influencing
readers towards perceiving a review as authentic: Reviewer Validation,
Purchase Validation, Review Detail, and Negative Reviews.
Reviewer Validation refers to cues observed by customers when
evaluating the credibility of a review via its author. Factors influencing
the credibility of a reviewer include variables such as the number of reviews
written along with the ratio of helpful votes received (Sharma &
Aggarwal, 2019). Along with the amount of customer feedback labelling
a reviewer as helpful, other ranking methods, such as a leaderboard
among other reviewers (e.g., top 100 reviewers) also play a role. Highranking
reviewers are more likely to invoke feelings of credibility (Baek,
Ahn & Choi, 2012). Another factor to be considered in reviewer validation
is reviewer identification, as consumers need to contemplate the
identity of the reviewer, before deciding on a review's authenticity (Chatterjee
etal., 2023). Cues such as the inclusion of the reviewer's true name
35
THEORETICAL BACKGROUND
or photograph generally heighten the reviewer's credibility (Liu & Park,
2015). Ideally, reviewers, who use their real names, have a profile picture,
and hold a high reviewer rating appear more honest and tend to
invoke trust in customers (Lee & Choeh, 2016).
An example of a credible reviewer can be seen in Figure 6, including factors
such as a high number of reviews written, or the presence of a real
name and profile picture.
Figure 6: Example of a credible reviewer profile
Q D Q D D 4 / 7 / 2 0 1 6
I took a Mixed Class at this studio yesterday as part of
Classpass. This was my first Barre class and it was a solid
workout which definitely pushed me to my limits during the
class without inducing fatigue.
The instructor - Sarah was very encouraging and very
Source: McCabe, 2019
Purchase Validation pertains to signs, indicating that the author of
a review has genuinely purchased the reviewed product or service and
has first-hand experience with it, rendering him eligible to authentically
review it. A clear indicator is the inclusion of photos or videos of the
product in the review. Li et al. (2020) reveal that since the majority of
fake reviews do not contain such evidence, a review that includes it tends
to gain significant credibility. Chatterjee etal. (2023) further support this
claim, suggesting that based on empirical research, customers should
ideally prefer reviews that contain photos or videos, over those that do
not. Further signs may include the presence of a "Verified Buyer" symbol,
essentially confirming that the reviewer has bought the product (Lappas,
Sabnis & Valkanas, 2016). Another approach to this is used by websites,
that do not use such symbols but only allow users who have purchased
on the site to leave a review (Donaker, Kim and Luca, 2019).
Review Detail describes the amount of information about a product
or service contained within the review. This can include factors such as
the number of distinct details included, or the degree of specificity to
which each detail is described (Harrison-Walker & Jiang, 2023). Zhang,
Wang& Wu (2021) suggest that the higher the level of detail in an online
review, the higher its perceived authenticity, as consumers are more
I Shivani P.
Chicago, IL
i 319 friends
O 177 reviews
tD 200 photos
Elite 19
36
THEORETICAL BACKGROUND
likely to trust reviews with detailed descriptions of products and experiences.
This is in line with subsequent research by Kim, Park and Mariani
(2023), who have found that authentic reviews often contain more thorough
details and more precise evaluation of the reviewed product or ser-
vice.
Negative Reviews - As it is frequently the case, that people attribute
negative information with more credibility than positive information,
consumers often tend to place more importance on it and usually allow
it to have more impact on their decisions. For this reason, it is not uncommon
for customers to attribute more authenticity to negative reviews
than to positive ones (Harrison-Walker & Jiang 2023). According
to Jha & Shah (2021), the presence of negative reviews can increase the
perceived authenticity of a product's overall review set, as customers feel
it is necessary to have some sort of opposition to positive feedback. This
aligns with the findings of Le etal. (2022), who argue that consumers are
often sceptical in the case that all reviews are positive, viewing them as
potentially manipulated.
Cues that a review is fake
As for the indicators predicting review fakeness from the perspective of
the consumers, the following cues were selected: Reviewer Implausibility,
Review Imbalance, Review Overlap, Review Grammar.
Reviewer Implausibility, similar to Reviewer Validation, is one of
the cues examining not the content of the review, but rather the authenticity
of the reviewer. Unfortunately, there is no shortage of situations,
where fake reviews are generated by bots - automated accounts holding
no credibility or value to the consumer (Khalid, 2019). It is also common
to find not only bots littering the online marketspace but also human users
creating several fake profiles to spam fake reviews. These fake accounts
often feature nonsensical usernames, comprising unsystematic
combinations of numbers and letters (Harrison-Walker & Jiang 2023). It
is for this reason, that many websites have nowadays adopted a series of
measures to confirm a reviewer's authenticity, such as performing verification
checks or providing information such as reviewer ratings, past
reviews, and others (Khalid, 2019). As discussed earlier in the thesis, it
is also quite common for fake reviewers to either not add a profile picture
37
THEORETICAL BACKGROUND
to their account at all or to assume another person's identity, as illustrated
by Figure 7.
Figure 7: Fake reviewer posing as actor Chris Hemsworth
Review Imbalance refers to the overall balance of the review's message,
assessing whether or not the review is entirely one-sided, such as
purely positive or negative, or whether it contains a more balanced mix
of information. Research has shown that a balanced perspective in a review,
covering a mix of both strengths and weaknesses of a product or
service, often helps mitigate perceptions of reviewer bias, making the review
appear more honest and credible. Conversely, reviews that are either
excessively positive or negative tend to raise suspicions about their
authenticity, thus lowering consumer's trust (Yan & Hua, 2021). Suspicions
also when an imbalance occurs not only within a single review but
also across the whole set of reviews. As already discussed earlier in relation
to the Negative Reviews cue, a lack of balance within the whole review
set, such as the complete absence of negative reviews tends to raise
suspicion among customers, particularly those with prior experience encountering
deceptive reviews (Kollmer et al., 2022).
Review Overlap or in other words, the presence of similar linguistical
patterns across multiple reviews is often a notable cue in discovering
fake reviews. According to the literature, one of the most common ways
consumers detect fake online reviews is by discovering multiple reviews
38
THEORETICAL BACKGROUND
using similar or identical language, phrasing, or even entire sentences
(Schoolov, 2020). Not only the structure of the review but also the time
of its posting is an important factor to consider. Whenever there is a
larger number of reviews posted at the same or nearly the same time,
they immediately appear suspicious, as this may indicate that these reviews
had to be submitted within a defined period, potentially suggesting
they were solicited (Dragan, 2016). Recent research approaches focused
on machine-learning-based fake review detection often rely on identifying
these overlapping language patterns and repetitive phrasing among
reviews to distinguish those written by bots or paid reviewers (Mohawesh
et al., 2024). However, it is essential to distinguish between different
interpretations of review overlap. While the cue discussed in this
subsection refers specifically to factors such as identical or very similar
sentences, words, and phrases, some sources define overlapping reviews
rather as those sharing similar content or themes. In this sense, overlapping
reviews may suggest that many consumers have had similar experiences
with the product or service, indicating a form of consensus that
can enhance the review's perceived credibility (Pooja & Upadhyaya,
2022).
Review Grammar describes, as the name suggests, the extent to
which a review is grammatically correct. Research shows, that grammatical
inaccuracy and spelling mistakes were the third most common cue
consumers used to label a review as being fake (Schoolov, 2020). Wu et
al. (2020) support this finding noting that customers often perceive reviews
that are grammatically well-written as credible and authentic,
whereas those with poor grammar are more likely to be labelled as fake.
2.4.2 Machine based detection
Machine-based detection of fake online reviews relies on the use of artificial
intelligence (Al) and the concept of machine learning (ML). Ling
(2023) describes ML as a field of study, in which computers can learn and
improve independently without requiring explicit human programming.
This means that instead of being hardcoded with specific instructions,
machines can analyse data and improve their performance over time
based on that data. Crawford et al. (2015) describe ML techniques as divided
into supervised, semi-supervised, and unsupervised.
39
THEORETICAL BACKGROUND
Supervised learning techniques are amongst the most commonly
used in machine learning, using a dataset with predetermined labels,
where the right answer is already known. Using this method, the algorithm
is then trained on the patterns instilled in this fixed set of data, allowing
it to anticipate outcomes for previously unknown data.
Semi-supervised learning combines techniques used both in supervised
and unsupervised learning. For training, it uses a combination of
labelled and unlabelled data, usually a small amount of labelled data and
a much larger collection of unlabelled data. This method works particularly
well in situations containing an extensive amount of unlabelled
data, such as detecting review spam. As semi-supervised learning often
employs fewer labelled instances than supervised learning while still
providing a greater structure than unsupervised techniques, it can prove
to be a great enhancer of accuracy and effectivity within the learning pro-
cess.
Unsupervised learning techniques operate solely on unlabelled data.
The primary objective of these algorithms is to uncover concealed patterns
and connections within the data without the use of predefined outcomes
or labels. In doing so, unsupervised learning often involves clustering
data based on similarities and subsequently discovering new patterns
and correlations.
Recent advancements in technologies such as machine learning and
artificial intelligence, have led to a significant improvement in the ability
to detect fake online reviews. For example, Salokhe (2024) demonstrated
how with the use of a supervised learning algorithm named "Random
Forest", the accuracy of distinguishing between authentic and fake
reviews has reached an impressive 91.72%, while Al-Saad (2024) notes,
that models such as "BERT" or "RoBERTa" have been able to detect even
the subtlest of differences in language patterns between authentic and
fake reviews, achieving a detection accuracy of 97.1%.
40
METHODOLOGY
3 Methodology
As stated previously, the thesis is to be divided into two parts. The objective
of the theoretical part is to thoroughly examine and synthesize the
existing literature and current viewpoints on the topic of fake online reviews,
highlighting the various types of fake reviews prevalent in today's
online environment, together with exploring the different methods of
differentiating between fake reviews and authentic ones. The practical
section then aims to investigate the strength of various factors, impacting
consumers' perceptions of review authenticity, as well as the influence
of authenticity on hotel booking intention.
The theoretical part of the thesis will be completed through the analysis
of secondary data gathered from previous works on the topic. The study
conducted in the practical section of the thesis will rely on quantitative
data collected by conducting primary research via a questionnaire,
aimed at consumers of hospitality services, such as hotels and restaurants,
within the Czech Republic. The questionnaire items were adapted
from previous studies conducted by Jean Harrison-Walker (2023) and
Bucher (2024), both focusing on the specific cues consumers use to assess
whether a review is authentic or fake, while the scales were adapted
from the work of (Roman et al., 2023). To answer the research questions,
the results of this primary research will then be processed using SPSS
Statistics along with other relevant tools.
3.1 Research design
As mentioned above, the study builds upon the research conducted by
Harrison-Walker (2023) and Bucher (2024), in which, several cues were
established as relevant to consumer's perception of review authenticity.
The cues in question being Implausibility and Overlap, signalling a review
is fake, paired with Specificity and Validation, signalling a review
is authentic.
While these previous works had primarily focused on whether individual
cues influence the perception of authenticity at all - that is,
whether they act as signals of authenticity or as signals of possible falsification,
what remained unexplored was the strength of individual cues
and their comparison among each other. This is where the main contribution
of this thesis lies. The aim is not only to verify whether the
41
METHODOLOGY
individual cues impact perceived authenticity, but also to determine how
strong their effects are in comparison with each other. The study therefore
directly manipulates these four cues, quantitatively compares their
relative effects, and simultaneously tests how these differences subsequently
influence the willingness to book a hotel.
Building on this reasoning, a between-subjects experimental study
was conducted, in which each respondent was randomly assigned to one
of five versions of an online hotel review. All review texts were generated
using Open AI's ChatGPT. The used prompts and full review texts can be
seen in Appendices B and C. To ensure consistency, the Baseline review
was created first, describing a standard positive hotel stay in approximately
80-100 words. The four manipulated reviews were then derived
from this baseline version and were constructed to be as comparable as
possible within the constraints of the experimental design, matching the
Baseline in length, the same 5-star rating and a similar overall structure,
with the main difference being only the presence of the tested cues.
Therefore, five review versions were created:
• BO - Baseline (neutral review without a cue)
• SI - Specificity
• V I - Validation
• 01 - Overlap
• II - Implausibility
To operationalize each cue, a single sub-item was selected from a pool of
possible formulations. Each chosen sub-item was identified as the most
representative of its corresponding cue, while adhering to a minimalconfound
principle, ensuring that the review versions differed only in the
intended cues and not in other unintended aspects. Although the initial
pool of available sub-items for each cue was relatively broad, several options
were incompatible with the constraints of the experimental design.
Some sub-items would have introduced additional variation in review
length, visual elements, and other factors, which could have compromised
the comparability of the stimuli across conditions. Therefore, only
the sub-items that aligned with both the theoretical definition of each cue
and the practical requirement of keeping the reviews comparable were
retained. The specific sub-items used to create the four manipulated review
versions can be seen in Table 1.
42
METHODOLOGY
Table 1: Sub-items used in generative review creation
Cue Selected sub-item
Specificity SI: Reviews that are very specific in detail.
Validation VI: Reviews where the reviewer is identified by his/her real name,
along with a form of "Verified buyer" badge.
01: Multiple reviews in a set of reviews that sound or appear
similar to each other.
Overlap
Implausibility II: Reviews that use excessive cliches (such as, "This is the
last x you will ever have to buy!")
Source: Own elaboration, adaptedfrom Harrison-Walker (2023) and Bucher (2024)
To ensure adequate reliability and validity, the survey items were
adapted from established measurement scales in prior studies.
The dependent construct "perceived authenticity" was measured
using three items adapted from Román etal. (2023), quantifying to which
extent the displayed review was seen as genuine and trustworthy. The
second key construct "booking intention" was also adapted from Román
et al. (2023), measuring the respondents' subsequent intention to book
the hotel described in the review. In addition, a set of manipulation
checks was developed for this study, in order to assess whether or not
the participants noticed the specific cue (Specificity, Validation, Overlap,
Implausibility) distinguishing their review version from the Baseline
version. These items reflected the specific signals participants were expected
to pick up on when retrospectively evaluating the review. All
items were measured on a 7-point Likert scale, as literature suggests that
7-point scales offer a good balance between response sensitivity and
ease of use, and may better capture respondents' evaluations than 5point
scales (Finstad, 2010). The specific items used to measure the constructs,
along with the manipulation checks, are displayed in Table 2 be-
low.
43
METHODOLOGY
Table 2: Measurement of constructs
Construct
Item
code
Measurement item
Perceived au-
authl
I consider this online review to be...
thenticity
authl
1 = False, 7 = Authentic
auth2
I consider this online review to be...
auth2
1 = Untrustworthy, 7 = Trustworthy
I consider this online review to be...
auth3 1 = Deceptive, 7 = Honest
(Adaptedfrom Roman et ah, 2023)
Booking
intention
BI1
After reading this online review, I would consider
staying at this hotel in the future.
Booking
intention
1 = I completely disagree, 7 = I completely agree
After reading this online review, it is likely that I
BI2 would make a reservation here.
1 = I completely disagree, 7 = I completely agree
After reading this online review, I would give this
BI3
hotel a chance.
BI3 1 = I completely disagree, 7 = I completely agree
(Adaptedfrom Roman et ah, 2023)
Manipulation
checks
MC_spec
The review contained specific details (e.g., times,
names, specific elements).
Manipulation
checks
1 = I completely disagree, 7 = I completely agree
I noticed elements of review verification (e.g., autMC_val
hor name and/or "Verified Stay" label).
1 = I completely disagree, 7 = I completely agree
The text seemed formulaic and contained repetiMC_over-
tive information (repeated or nearly identical
lap phrasing).
1 = I completely disagree, 7 = I completely agree
44
METHODOLOGY
Construct , Measurement item
The text contained exaggerated/cliched statements
(e.g., "the best ever", "absolutely perfect",
MC_impl etc.).
1 = I completely disagree, 7 = I completely agree
(Own elaboration)
Source: Own elaboration
A simple conceptual framework of the study is presented below in Figure
8. The main independent variable is the experimental review condition,
represented by the five review versions (BO, VI, SI, II, 01). These conditions
are assumed to influence how authentic the review appears to respondents,
addressing RQ1. Perceived authenticity is then modelled as
the key mechanism, shaping participants' intentions to book the hotel,
corresponding to RQ2. Therefore, the framework assumes that the review
conditions affect booking intention indirectly through perceived
authenticity, while a possible direct effect of the review condition on
booking intention is included, which is tested for in the subsequent mediation
analyses.
Figure 8: Conceptual framework
BO
VI
Sil
l
01
Condition
RQl Perceived
authenticity
'*•--... (Mediation)
RQ2 Booking
intention
Source: Own elaboration
45
METHODOLOGY
3.2 Data collection
To achieve the objectives of the study, a questionnaire was used in order
to gather quantitative data from respondents. The questionnaire consisted
of three sections.
The first section included a simple question, purely asking whether
or not the participant had at least occasionally read online reviews of hotels.
If not, the questionnaire was ended straight away. This section
served to ensure that the final sample consisted of users with at least
minimal experience of online reviews.
The second section had randomly assigned the partaker to one of
the five review variations, using a redirecting randomization proxy, leading
to one of five versions in Google Forms. The respondents were then
asked to evaluate the review based on their personal impression and
subjective assessment. Immediately after seeing the review, the participants
were then shown the items measuring perceived authenticity,
booking intention, and the manipulation checks as described in the previous
subsection and summarised in Table 2. A single attention-check
item was also included in this section to filter out inattentive or careless
responses.
The third section addressed demographic details of the respondents,
including their age, gender, and education, along with a question on
how often they usually read online reviews before booking a hotel.
The questionnaire was translated to Czech, as the research is based
in the context of the Czech Republic, and was distributed to potential respondents
through online platforms such as social media, survey platforms
and other digital channels, as well as physical locations, with the
use of QR codes. The data collection took place over a three-week period
from November 3r d
, 2025, to November 24t h
, 2025.
3.2.1 Pilot survey
Before commencing the main part of data collection, the questionnaire
underwent pilot testing in order to test its comprehensibility, randomization
functionality and overall user-friendliness. The pilot version was
distributed to 15 respondents, who were asked to complete the questionnaire
in the same way as regular participants in the main study and
provide feedback. Based on their comments, a number of adjustments
were made to increase the clarity of the instructions, eliminate or at least
46
METHODOLOGY
lower the possibility of confounds and ensure that the measured effects
were evaluated purely on the basis of review cue manipulations and not
other content elements. The adjustments made after the pilot included:
• Specifying the estimated completion time, which was approximated
to around 5 minutes, allowing future respondents to be
given more accurate information at the beginning.
• Addition of a progress bar, increasing participant comfort and potentially
encouraging full completion of the questionnaire.
• Elimination of possible confounds - the star ratings of the reviews
were unified, the mention of price per night was removed from
the details mentioned in the Specificity review, or the selection of
a less vividness-inducing sub-item in the Validation review (no
photographic elements).
• Change in wording of manipulation check questions, the oftenmisunderstood
term "overlap" was replaced with the clearer and
more intuitive term "repeated information", which respondents
understood better.
• Correction of typographical errors and confusion that arose during
the generation of text using AI.
3.3 Data analysis
For all statistical analyses, IBM SPSS Statistics (version 29) was used,
supplemented by the PROCESS macro for mediation analysis (Model 4).
The analysis proceeded stepwise as follows:
1) A respondent profile analysis was first conducted to provide
a basic overview of the sample in terms of age, gender, education,
travel frequency and frequency of reading online hotel
reviews before booking.
2) Descriptive statistics (means, standard deviations, minimum
and maximum values) were computed for the main study
variables, namely perceived authenticity, booking intention,
47
METHODOLOGY
and the manipulation checks items, to provide an overview
of their distribution.
3) The internal consistency of the two multi-item variables was
examined using Cronbach's alpha. Item-total statistics were
reviewed to confirm all items contributed positively to scale
reliability and no items required removal.
4) Prior to running the main group comparisons, the assumptions
for ANOVA were assessed. Shapiro-Wilk tests and Q-Q
plots were used to examine normality within each group,
while Levene's test was used to assess the homogeneity of
variances. After conducting the omnibus group comparisons,
Games-Howell post hoc tests were used to identify specific
differences between the specific groups.
5) To quantify the association between perceived authenticity
and booking intention, a bivariate Pearson correlation coefficient
was computed, measuring the strength and direction
of the association between the two constructs.
6) Finally, a series of mediation models was estimated using the
PROCESS macro, in order to evaluate whether perceived authenticity
acted as a mediator between the review cues and
booking intention. In each model, the experimental condition
was entered as the predictor, perceived authenticity as the
mediator, and booking intention as the outcome. Indirect effects
were evaluated with 5,000 bootstrap resamples and
95% bias-corrected confidence intervals.
48
RESULTS
4 Results
4.1 Demographics
After data collection, the initial research sample contained a total of 328
respondents. From this number, 21 who had not passed the initial
screening question were excluded. These participants had stated that
they do not read online hotel reviews and therefore did not meet the condition
for continuing with the questionnaire. Furthermore, another 6 respondents
were subsequently omitted as a result of not passing the attention
check. Therefore, the final analysed sample consisted of 301 respondents,
whose demographic profile is displayed in Table 3.
In terms of gender, the sample was relatively balanced, although
women slightly outnumbered men (55.1%, n = 166), while men accounted
for 44.9 % of respondents (n = 135). In terms of age distribution,
the largest part of the sample fell into the 18-24 age group (46.2%), followed
by the 25-34 age group (35.8%). The 35-44 age group accounted
for 13% of participants, while older age groups were significantly less
represented, which is to be expected in an online survey.
Regarding education, respondents with a university education
(Bachelor's Master's and higher) predominated, representing 56.2% of
the sample. This was followed by the group with secondary education
(35.2%), while other levels of education were less strongly represented.
The travel behavior of respondents shows a relatively active pattern.
Almost half of the participants (45.5%) reported using accommodation
services 3-5 times in the past year. This was followed by 30.2%
who stayed 1-2 times, and 12% who reported 6-10 stays over the year.
A smaller proportion of respondents (5.6%) used accommodation more
than 11 times, while only 6.6% indicated that they had not stayed in a
hotel at all during the past twelve months.
Lastly, the section examining review-reading frequency prior to
booking revealed that the vast majority of respondents (69.1%) almost
always (in more than 75% of stays) read online reviews before choosing
a hotel, followed by 19.9% of participants selecting they read reviews often
(51-75%). Additionally, 7.3% had reported they read these review
sometimes (25-50%), while a mere 3.7% stated that they do so rarely
(less than 25% stays).
49
RESULTS
Table 3: Sample demographics
Variable Value Frequency %
Gender Male 135 44.9
Female 166 55.1
Age 18-24 139 46.2
25-34 108 35.8
35-44 39 13
45-54 11 3.7
55+ 4 1.3
Education Primary 6 2
High School 106 35.2
Higher vocational 20 6.6
Bachelor's degree 95 31.6
Master's degree 68 22.6
Higher 6 2
Hotel stays in
past year Not once 20 6.6
1-2 times 91 30.2
3-5 times 137 45.5
6-10 times 36 12
11 and more 17 5.7
50
RESULTS
Variable Value Frequency %
How often do you read Rarely (<25 % stays)
reviews before booking?
11 3.7
Sometimes (25-50 %) 22 7.3
Often (51-75 %) 60 19.9
Almost always (>75 %) 208 69.1
Source: Own elaboration
4.2 Scale reliability
As mentioned before, two multi-item scales were employed in the questionnaire:
perceived authenticity (three items) and booking intention
(three items), each rated on a 7-point Likert scale. Before conducting further
analyses, it is crucial to examine the internal consistency of these
scales using Cronbach's alpha. The perceived authenticity scale showed
excellent reliability (a = .959), indicating very strong consistency among
the measured items. Similarly, the booking intention scale also demonstrated
excellent internal consistency (a = .923), meaning both scales
have met the conditions for sufficient reliability.
For both constructs, scale scores were computed as the arithmetic
mean of the three items, with higher values indicating higher perceived
authenticity and higher booking intention.
4.3 Manipulation checks
Before commencing the analysis of individual review cues, manipulation
checks were used to confirm that the intended cues were salient to respondents
at the group level. Each version of the questionnaire included
all four manipulation check items (Specificity, Validation, Overlap,
51
RESULTS
Implausibility), rated on a 7-point Likert scale. Although each respondent
was randomly assigned to only one review version, all four manipulation
check items were presented in order to reduce demand characteristics
and avoid revealing the focus of the manipulation. For each condition,
only the item corresponding to the cue present in the displayed review
served as the focal manipulation check and was subsequently analysed.
The remaining three items served purely as masking items and
were not interpreted as manipulation checks for that specific condition.
For descriptive reporting values of 5 or higher on the focal manipulation
check item were considered an illustrative threshold of successful recognition
of the manipulation, but the primary evaluation of manipulation
effectiveness relied on group-level comparisons of the continuous manipulation
check scores. As the manipulation checks were conducted after
exposure to the experimental reviews, respondents were not excluded
based on these items and the >5 rates were regarded as descriptive
data. Although this differs slightly from the more common practice
of dropping failed manipulation-check cases, recent research has warned
that such post-treatment exclusions can introduce selection bias and differential
attrition across conditions, potentially distorting the estimated
effects (Aronow et al., 2019; Varaine, 2023).
The results suggest that all four experimental manipulations were
fairly effective. According to Games-Howell post-hoc tests, each focal
condition scored significantly higher on its own manipulation check than
the other four conditions, while one-way Welch ANOVAs revealed significant
between-group differences on the pertinent MC items across all
four cues (all p < .001, with medium to large effect sizes). In other words,
while the non-focal conditions often did not significantly differ from one
another on these items, respondents in SI reported the highest level of
specific detail on MC_spec, those in VI reported the strongest signs of
reviewer verification on MC_val, 01 respondents strongly noticed repeated
information on MC_overlap, and II respondents perceived the
greatest degree of exaggerated or cliched wording on MC_impl. Pass-rate
descriptives followed the same pattern. In every manipulated condition,
more than 80% of respondents correctly identified the intended cue,
with the highest success rate achieved in the Validation condition (94%
MC-pass). Conversely, the lowest, but still high values were recorded for
Specificity (83% MC-Pass). The manipulation-check means, standard deviations,
and MC-pass percentages are summarised in Table 4.
52
RESULTS
Table 4: Manipulation check results for each condition
Condition
Relevant
item
MC M
(condition)
M
(baseline)
SD
MC-pass
rate (% >
5)
Specificity MC_spec 5.45 3.22 1.40 83
Validation MC_val 6.25 1.90 1.21 94
Overlap MC_overlap 5.86 3.97 1.29 91
Implausibility MCJmpl 5.64 3.05 1.60 88
Source: Own elaboration
A-A Group comparisons: Perceived authenticity
The following section examines how the perception of authenticity had
differed across the five experimental review conditions. First, descriptive
statistics for each condition are reported, followed by tests of assumptions.
Subsequently, the results of the omnibus group comparison and
the corresponding post-hoc tests are presented.
4.4.1 Descriptive statistics
The descriptive statistics for perceived authenticity across the five experimental
conditions are presented in Table 5. The highest mean authenticity
was observed in the Validation condition (VI; M = 5.19, SD =
1.39), closely followed by the Baseline (BO; M = 4.86, SD = 1.75) and Specificity
conditions (SI; M = 4.60, SD = 1.79). Authenticity scores then decline
in the Overlap condition (01; M = 3.70, SD = 1.60) and especially the
Implausibility condition (II; M = 2.46, SD = 1.25). Already, these descriptive
results suggest significant differences in perceived authenticity
across the conditions, with Validation increasing and Implausibility
clearly decreasing mean perceived authenticity.
53
RESULTS
Table 5: Descriptive statistics for perceived authenticity
Condition N Mean (M) SD
Baseline (BO] 58 4.86 1.75
Implausibility (11] 59 2.46 1.25
Overlap (01) 57 3.70 1.60
Specificity (SI) 64 4.60 1.79
Validation (VI) 63 5.19 1.39
Total 301 4.18 1.84
Source: Own elaboration
For a better visualisation of the mean distribution, Figure 9 presents
a mean plot of perceived authenticity by condition.
Figure 9: Mean perceived authenticity across conditions
B0 II 01 SI VI
condition_num
Source: Own elaboration in SPSS
54
RESULTS
4.4.2 Test of assumptions
As the primary goal of the research was to compare perceived authenticity
across the five experimental conditions, a one-way between-subjects
ANOVA test was deemed appropriate. Before conducting the analysis, the
standard ANOVA assumptions must be assessed, namely sample independence,
the normality of residuals within each group, and the homogeneity
of variances across groups (Field, 2018).
The first assumption, independence, states that in order to ensure
valid results, the observations must be uncorrelated. This assumption
was ensured by the study's between-subjects design, which required
each respondent to complete the questionnaire only once and assigned
them at random to a single review condition, preventing any measurement
from being repeated or shared across conditions. The independence
assumption of ANOVA is typically not problematic in designs where
each case represents a distinct respondent (Rabusic, 2019).
The second assumption, normality of residuals was examined using
Shapiro-Wilk tests and Q-Q plots for each condition. The Shapiro-Wilk
test was significant for four of the five groups (BO, II, SI, VI, p < .01),
indicating clear deviations from normality, while the Overlap condition
did not significantly deviate from normality (p = .122). Although the
Shapiro-Wilk tests were significant for most conditions, inspection of the
Q-Q plots indicated only slight deviations from normality, usually common
for Likert-type data. These deviations are not expected to threaten
the robustness of ANOVA, especially using samples of our size (N ~ 60
per group)(Ghasemi & Zahediasl, 2012).
The final assumption, homogeneity of variances across groups was
assessed using Levene's test. The test based on the mean was significant,
F(4, 296) = 3.88, p = .004, showing that the assumption was not met. As
this was the case, the group comparison for perceived authenticity was
conducted using the more robust Welch's ANOVA, resolving the issue of
unequal variances (Gravetter & Wallnau, 2017).
4.4.3 Welch ANOVA results
Following up on the assumption test results, a Welch one-way ANOVA
was used to test for the overall differences in perceived authenticity
across the five manipulation conditions. The subsequent analysis revealed
a statistically significant effect of condition on perceived authenticity,
F (4,146.83) = 39.38, p < 0.001, indicating that mean authenticity
55
RESULTS
ratings differed across the conditions. The effect size, where omega
squared was used as a bias-corrected alternative appropriate for Welch
ANOVA, was large, suggesting that which review version participants
saw accounted for a notable proportion of the variability in perceived
authenticity (oo2
= .28; 95% CI [.18, .35]).
Table 6: Welch ANOVA results
Test dfl df2 F p w2
Welch ANOVA 4 146.83 39.38 < .001 .28
Source: Own elaboration
4.4.4 Post-hoc tests
As the overall Welch ANOVA was significant, subsequent Games-Howell
post-hoc tests were conducted to identify which conditions differed in
perceived authenticity. The detailed comparisons are displayed below in
Figure 10, which presents the mean differences and adjusted p-values for
all condition pairs.
The findings show a distinct difference between two "problematic"
conditions and a set of reviews with comparatively high levels of authenticity.
In the Implausibility condition (II), all pairwise differences
reached statistical significance at p < .001, suggesting that reviews in II
were deemed significantly less authentic than reviews in any of the other
four conditions. This implies that overly exaggerated and cliched language
can be particularly detrimental to credibility, resulting in authenticity
ratings that are significantly lower than those of a neutral baseline
review.
The Overlap condition (01) showed a similar, albeit a slightly less
pronounced pattern. Compared with Baseline, 01 was also associated
with significantly lower authenticity scores and differed from the Specificity
and Validation conditions, indicating that repetitive or formulaic
wording reduces perceived credibility. The differences between 01 and
II, however, were smaller and less consistent, suggesting that these two
manipulations form a "low-authenticity tier," with Overlap in a moderately
negative position and Implausibility ranking the lowest.
In contrast, despite minor numerical variations in their means, conditions
Baseline (BO), Specificity (SI), and Validation (VI) did not significantly
differ from each other in the post-hoc tests (all p >.24). As a result,
56
RESULTS
these three criteria seem to create a cluster of reviews that are perceived
as similarly genuine. Within this higher-authenticity cluster, Validation
showed the highest mean score, followed closely by Specificity and the
Baseline condition. However, because none of these differences were statistically
significant, any numerical ordering should be interpreted with
caution.
In conclusion, it is possible to say the results of the Games-Howell
tests depict a fairly straightforward pattern. The condition II consistently
produced the lowest authenticity ratings, while the condition 01
represents a moderately lower level of perceived authenticity. Higher authenticity
scores can be seen among conditions BO, SI and VI, where condition
VI was ranked the highest in terms of mean authenticity. However,
it is important to state, that the differences among these groups (BO,
SI, VI) were not found to be statistically significant, and therefore it is
more appropriate to understand them as a cluster with similarly high
perceived authenticity, rather than as a precisely ranked order.
57
RESULTS
Figure 10: Games-Howell post-hoc comparisons for perceived authenticity
Multiple Comparisons
Dependent Variable: auth_mean
Games-Howell
(1) condition_num 0) condition_num
Mean
Difference (1J)
Std. Error Sig.
BO 11 2.40444* .28094 <.001
01 1.16031* .31217 .003
SI .25790 .32058 .929
VI -.32841 .28867 .786
11 BO -2.40444* .28094 <.001
01 -1.24413* .26691 <.001
SI -2.14654* .27670 <.001
VI -2.73285* .23900 <.001
01 BO -1.16031* .31217 .003
11 1.24413* .26691 <.001
SI -.90241* .30836 .033
VI -1.48872* .27504 <.001
SI BO -.25790 .32058 .929
11 2.14654* .27670 <.001
01 .90241* .30836 .033
VI -.58631 .28454 .244
VI BO .32841 .28867 .786
11 2.73285* .23900 <.001
01 1.48872* .27504 <.001
SI .58631 .28454 .244
*. The mean difference is significant at the 0.05 level.
Source: Own elaboration in SPSS
4.5 Group comparisons: Booking intention
Similarly to perceived authenticity, one-way analyses were also conducted
to examine the effect of the review conditions on booking intention.
The variable was once again measured on a 7-point Likert scale,
where higher scores marked higher intention of booking, whereas lower
scores indicated a lower intention to stay in the reviewed hotel. The following
subsections first report on the descriptive statistics for booking
intention in relation to each condition. Subsequently, the mandatory
58
RESULTS
tests of assumptions are carried out, followed by the relevant omnibus
group comparison, and the corresponding post-hoc tests.
4.5.1 Descriptive statistics
Table 7 presents the descriptive statistics for booking intention across
the experimental conditions. Mean booking intention was lowest in the
Implausibility condition (II; M = 3.58, SD = 1.51), and still relatively low
compared to the other conditions, in the Overlap condition (01; M = 4.26,
SD = 1.44). Similar to perceived authenticity, higher means can be observed
for the Baseline (BO; M = 5.20, SD = 1.47), Specificity (SI; M = 5.05,
SD =1.44) and Validation (VI; M = 5.33, SD = 1.56) conditions. Overall,
these results suggest that so called "negative" cues, such as Implausibility
and Overlap are associated with a lower intention to book, while the remaining
conditions show higher intentions of booking. These findings
align with those regarding perceived authenticity, where conditions Implausibility
and Overlap too produced the lowest mean rankings, suggesting
that cues which undermine a review's perceived authenticity also
seem to depress the reader's intention to book, whereas cues associated
with higher authenticity are accompanied by higher booking intentions.
Table 7: Descriptive statistics for booking intention
Condition N Mean (M) SD
Baseline (BO) 58 5.20 1.47
Implausibility (11) 59 3.58 1.51
Overlap (01) 57 4.26 1.44
Specificity (SI) 64 5.05 1.44
Validation (VI) 63 5.33 1.56
Total 301 4.70 1.84
Source: Own elaboration
For a cleaner visual comparison of the means, Figure 11 presents a
mean plot of booking intention by condition.
59
RESULTS
Figure 11: Mean booking intention across conditions
Means Plots
3.50
BO II 01 SI VI
condition num
Source: Own elaboration in SPSS
4.5.2 Test of assumptions
As in the analysis for perceived authenticity, the assumptions for a oneway
ANOVA on booking intention, namely independence, normality and
homogeneity of variances were examined.
Independence of observations was once again ensured by the between-subjects
survey design, where each participant was randomly assigned
and contributed only one set of booking intention assessments.
As for normality, Shapiro-Wilk tests were conducted, indicating statistically
significant deviations from a perfect normal distribution in
groups BO, SI, VI and II, all p < .01, with only the 01 condition not significantly
departing from normality (p = .109). Nevertheless, as with the
previous analysis, a visual inspection of histograms and Q-Q plots across
all conditions suggests an approximately normal distribution, with no extreme
outliers, indicating only moderate departures from normality, typical
for Likert-type data of this sample size. Therefore, as was the case in
the analysis of perceived authenticity, we can consider a one-way ANOVA
sufficiently robust to such deviations supposing other assumptions are
met, and thus the analysis was deemed appropriate to proceed.
Homogeneity of variances was assessed using Levene's test, which
was non-significant, F(4, 296) = 0.57, p = .684, meaning that the variance
60
RESULTS
of booking intention did not vary significantly across conditions. As the
independence of observations, approximately normal distributions, and
homogeneity of variances checks were successful, the use of a standard
one-way ANOVA to compare booking intention scores across groups was
deemed appropriate in this case.
4.5.3 One-way ANOVA results
A one-way ANOVA revealed a statistically significant effect of the review
condition on booking intention, F (4, 296) = 16.47, p < .001. The effect
size fell into the small to medium range, citing oo2
= .17 (95% CI [.09, .24]),
showing that approximately 17% of the variance in booking intention
was accounted for by the experimental manipulation of review cues. The
effect was somewhat smaller, though broadly comparable to the effect
exerted on perceived authenticity (oo2
= .28), suggesting the cues have
demonstrated a stronger effect on authenticity than booking intention,
while however still maintaining a meaningful impact on both outcomes.
Given this significant omnibus effect, post-hoc comparisons were conducted
to more precisely identify how conditions differed from one another
in booking intention.
Table 8: One-way ANOVA results
Test dfl df2 F p w2
One-way ANOVA 4 296 16.47 < .001 .17
Source: Own elaboration
4.5.4 Post-hoc tests
Following the significance of the one-way ANOVA, Games-Howell tests
were once again conducted, examining which conditions differed significantly
in booking intention. The full results are displayed in Figure 12,
showing mean differences and the adjusted p-values for all pairs of con-
ditions.
Participants exposed to the Implausibility condition (II) reported
the lowest willingness to book, showing consistency with the mean plot.
Their booking intention scores were significantly lower than those in the
Baseline, Specificity, and Validation conditions (all p < .001), suggesting
61
RESULTS
that excessively emotional and cliched language not only undermines authenticity
perceptions, but also significantly dampens behavioral inten-
tions.
The Overlap condition (01) also reduced booking intention relative
to several other versions. Although the mean remained slightly above the
midpoint of the scale, booking intention in 01 was still significantly lower
than in the Baseline, SI, and VI conditions (p = .007-.024). Rather than
indicating a complete rejection of the hotel, this pattern reflects a moderate
reluctance to book.
Regarding conditions II and 01, these negative cues did not significantly
differ from one another in booking intention when compared directly
(p = .104). This suggests that, from a behavioral perspective, both
types of suspicious cues exert a broadly similar dampening effect on intentions,
even though II appears somewhat more extreme atthe descriptive
level.
The three higher-intention conditions, Baseline (BO), Specificity
(SI), and Validation (VI), again formed a relatively homogeneous cluster
with elevated willingness to book and no significant differences among
them (allp>.75).
Taken together, the results mirror the earlier post-hoc findings for
perceived authenticity, where conditions II and 01 also yielded lower
authenticity ratings, while the remaining conditions formed a high scoring
cluster. However, in the authenticity analysis, II and 01 differed significantly
from each other, while for booking intention they converged.
This means that participants distinguished between the two negative
conditions in terms of perceived authenticity but treated both with a similar
level of problematicity while considering whether to book the hypothetical
hotel.
62
RESULTS
Figure 12: Games-Howell post-hoc comparisons for booking intention
Multiple Comparisons
Dependent Variable: Bl_mean
Games-Howell
(1) condition_num (J) condition_num
Mean
Difference (1J)
Std. Error Sig.
BO 11 1.61348* .27469 <.001
01 .93809* .27050 .007
SI .14332 .26355 .983
VI -.13793 .24765 .981
11 BO -1.61348* .27469 <.001
01 -.67539 .27304 .104
SI -1.47016* .26615 <.001
VI -1.75141* .25042 <.001
01 BO -.93809* .27050 .007
11 .67539 .27304 .104
SI -.79477* .26183 .024
VI -1.07602* .24582 <.001
SI BO -.14332 .26355 .983
11 1.47016* .26615 <.001
01 .79477* .26183 .024
VI -.28125 .23814 .762
VI BO .13793 .24765 .981
11 1.75141* .25042 <.001
01 1.07602* .24582 <.001
SI .28125 .23814 .762
*. The mean difference is significant at the 0.05 level.
Source: Own elaboration in SPSS
4.6 Relationship between perceived authenticity and
booking intention
In the previous sections, we examined how the various experimental
conditions influenced perceived authenticity and booking intention separately.
The present section then investigates the relation between these
two constructs and whether perceived authenticity helps to explain
63
RESULTS
differences in booking intention across the conditions. Specifically, the
analyses first examine the strength of the bivariate association between
perceived authenticity and booking intention, then followed by an evaluation
of whether perceived authenticity mediates the effect of review
condition on booking intention.
4.6.1 Correlation analysis
In order to quantify the basic association between the two main constructs,
a Pearson correlation was computed between perceived authenticity
and booking intention scores. As can be seen in Figure 13 below,
the analysis showed a strong positive relationship, r = .82, p < .001, N =
301, indicating that higher perceived authenticity of the reviews was also
associated with higher intentions of booking the hypothetical hotel. This
also shows that changes in perceived authenticity alone account for
about 67% of variance in booking intention (r2
~ .67). This substantial
correlation suggests that perceptions of authenticity are closely related
to participants' booking intentions and supports examining authenticity
as a potential mediator in subsequent analyses.
Figure 13: Pearson correlation between perceived authenticity and booking
intention
Correlations
authjrean Bljnean
authjmean Pearson Correlation 1 .824
Sig. (2-tailed) <.001
N 301 301
Bljnean Pearson Correlation .824 1
Sig. (2-tailed) <-001
N 301 301
**. Correlation is significant at the 0.01 level (2-
tailed).
Source: Own elaboration in SPSS
4.6.2 Mediation analysis
In order to further understand how the experimental manipulation conditions
affect consumers, an additional set of analyses tested whether the
64
RESULTS
effects of the various conditions on booking intention are transmitted
through perceived authenticity.
A mediation analysis using PROCESS Model 4 with multicategorical
coding of the independent variable was carried out to investigate these
effects. Four dummy variables (X1-X4) were used to represent the five
conditions, with the Baseline condition serving as the reference category.
5000 bootstrap samples were used to estimate indirect effects, with perceived
authenticity (authmean) designated as the mediator and booking
intention (BImean) as the outcome variable.
Table 9 displays the regression modelling of perceived authenticity
as a function of the four condition dummies. Only the two negative cues
Implausibility (XI) and Overlap (X2) significantly decreased perceived
authenticity in comparison to the Baseline, which is consistent with the
previous ANOVA results. The biggest drop was caused by Implausibility
(B = -2.404, p < .001), while Overlap produced a moderate but significant
decline (B = -1.160, p < .001). Conversely, there was no significant difference
between the Baseline and Specificity (X3) and Validation (X4) (both
p > .25). These patterns copy previous findings that while the negative
cues caused noticeable and significant declines in authenticity perceptions,
the authenticity-enhancing cues stabilized rather than raised them.
Table 9: Effects of experimental conditions on perceived authenticity
Predictor Coefficient (B) SE P 9 5 % CI
XI Implausibility -2.404 .290 <.001 -2.976 to -1.833
X2 Overlap -1.160 .293 <.001 -1.737 to -.584
X3 Specificity -.258 .285 .366 -.818 to .302
X4 Validation .328 .286 .251 -.234 to .891
Constant 4.862 .206 <.001 4.456 to 5.268
Source: Own elaboration
Across all models, perceived authenticity was found to be a powerful
and reliable predictor of booking intention. Authenticity had a significant
positive impact (B = 0.706, p < .001), as demonstrated in Table 10,
65
RESULTS
indicating that participants were significantly more inclined to reserve
the hotel when the review seemed more genuine, supporting the notion
that consumers' assessment of online reviews is largely influenced
by authenticity. None of the four condition dummies demonstrated significant
direct effects on booking intention when authenticity was included
into the model (all p > .48). This suggests that once perceived authenticity
was taken into account, the cues had no independent effect on
willingness to book. The lack of direct effects indicates that the mediator
was the main channel through which the manipulations had an impact.
Table 10: Mediation model predicting booking intention
Predictor Coefficient (B) SE P 9 5 % CI
XI Implausibility .083 .182 .648 -.274 to -.441
X2 Overlap -.119 .169 .482 -.453 to .214
X3 Specificity .039 .161 .811 -.278 to .355
X4 Validation -.094 .162 .562 -.412 to .224
Authenticity (M) .706 .033 <.001 .641 to .770
Source: Own elaboration
The bootstrapped indirect effects are summarized in Table 11. Both
conditions Implausibility and Overlap showed significant negative indirect
effects on booking intention, with confidence intervals clearly excluding
zero. This indicates that these manipulations lowered willingness
to book specifically because they reduced perceived authenticity.
Conversely, the indirect effects for Validation and Specificity were not
significant, which is in line with their minimal influence on authenticity
past baseline levels. In other words, the positive cues did not significantly
change authenticity enough to meaningfully influence booking intention.
As a whole, the pattern demonstrates that the cues influenced behavior
only through their impact on perceived authenticity, and while negative
cues carried behavioral consequences, positive cues were stabilizing but
not elevating.
66
RESULTS
Table 11: Bootstrapped indirect effects via perceived authenticity
Condition Indirect effect BootSE 95% CI
XI Implausibility -1.697 .209 -2.113 to-1.282
X2 Overlap -.819 .221 -1.242 to -.372
X3 Specificity -.182 .225 -.619 to .254
X4 Validation .232 .206 -.166 to .649
Source: Own elaboration
Overall, the mediation analysis suggests that booking intention was
indeed impacted by perceived authenticity within the studied sample.
The experimental cues affected booking intention only indirectly,
through their influence on authenticity, with the strongest effects observed
for the two negative cues. These results demonstrate the disproportionate
influence of credibility-reducing cues relative to credibilityenhancing
ones, while further emphasizing the significance of authenticity
in consumers' evaluations of online reviews.
67
DISCUSSION
5 Discussion
The following section aims to summarize the main empirical results of
the study, while also situating them in the context of existing literature
on online review authenticity and hotel booking behaviour. It first considers
how strongly the four manipulated cues (II, 01, SI, VI) influenced
perceived authenticity and whether these effects allow for the cues to be
meaningfully ordered from the least credible, to the most persuasive authenticity
signal, thereby addressing RQ1.
The discussion then turns to RQ2, by examining the extent to which
perceptions of authenticity, irrespective of their origin, shape consumer's
intentions of booking a hotel. To achieve this, the chapter integrates
evidence from the ANOVA results, the correlation between perceived
authenticity and booking intention, and the mediation models, to
outline how review cues and authenticity perceptions jointly influence
consumers' booking decisions. The answers to both RQ1 and RQ2 are developed
throughout this discussion and summarized explicitly at the end
of the chapter.
Addressing RQ1: How do various review cues affect perceived authenticity?,
the results indicate that the four manipulated cue conditions
can be meaningfully distinguished based on how they each shape authenticity
perceptions.
Across the results, the Implausibility condition (II) stood out as the
single most damaging pattern, producing the strongest decrease in perceived
authenticity. These findings align with the presented literature,
stating that exaggerated, overly broad or unrealistic statements are
among the typical linguistic patterns found in deceptive online reviews
(Carr & Piercy, 2021). Such formulations were characterized in the theoretical
context, as going against readers' expectations of authentic experience
sharing therefore serving to raise suspicion and to encourage
consumers to perceive the information as deliberately fabricated, rather
than experientially grounded. Additional support for these findings is
provided by Yan and Hua (2021), who demonstrate that excessively positive
or favourable reviews frequently cause suspicion and erode customer
trust, particularly when claims seem too good to be true. In this
way, the implausible phrases in II likely acted as instant "red flags", presumably
triggering strong suspicion, rather than persuasion, meaning
that these attempts to "oversell" a hotel risk backfiring by eroding trust
68
DISCUSSION
in both the review and the establishment, rather than enhancing its im-
age.
Review authenticity was also undermined by the Overlap condition
(01), albeit to a lesser degree than Implausibility. This pattern aligns
with the theoretical viewpoint that consumers expect genuine reviews to
represent unique, personally framed experiences, while in overlapping
reviews readers frequently interpret recurring or strikingly similar linguistical
patterns as indications of coordinated or non-genuine authorship
(Schoolov, 2020). This interpretation is consistent with recent computational
methods, where in order to differentiate authentic reviews
from those generated by bots or paid reviewers, machine-learning based
fake review detection techniques often rely on finding overlapping linguistic
patterns and repetitive phrasing (Mohawesh et al., 2024). Such
automated systems' use of overlap as a diagnostic indicator suggests the
possibility that human readers might apply a similar heuristic. Therefore,
respondents' perception of the formulaic or repeating phrasing as
proof of mechanical rather than genuine authorship likely explains the
moderate, but still significant decrease in authenticity for condition 01.
In contrast, the Specificity condition (SI) raised perceived authenticity
in comparison to both Implausibility and Overlap. This impact is
compatible with the previously discussed literature, citing that authentic
reviews typically contain fuller, more precise descriptions of time, place
and sensory features, serving as the opposite of misleading patterns,
such as broad, vague and inconcrete reviews (Carr & Piercy, 2021). This
interpretation is further supported by various sources, citing that authentic
online reviews typically contain thorough, specific descriptive
content and more accurate evaluation (Zhang, Wang & Wu, 2021; Harrison-Walker
and Jiang, 2023; Kim, Park and Mariani, 2023). As a result,
participants in the present study likely regarded the specific descriptions
in SI as genuine experiences that are more difficult to fabricate, boosting
the text's credibility by eliminating ambiguity and thus giving the review
a more authentic appearance.
The final condition Validation (VI), similarly raised perceived authenticity
in contrast to the negative cues. This result is consistent with
the theoretical claim that authenticity judgments are heavily influenced
by review validation factors (Chatterjee et al., 2023). As stated earlier,
fake reviews usually utilize anonymous or nonsensical usernames, often
made up of random letter and number combinations, raising suspicion in
customers (Harrison-Walker & Jiang, 2023). Furthermore, as noted by
69
DISCUSSION
Fong (2022), reviews posted by profiles lacking any form of identity validation,
such as utilizing clearly unrealistic photos and fictitious identities,
are frequently indicative of reviewers trying to conceal their true
identities, therefore indicating possible fabrication and undermining the
review's credibility. Conversely, and consistent with the results of the
present study, research indicates that reviews made by authors possessing
cues of real identity, such as profile photos, real names or high
ratings are seen as more trustworthy and honest (Liu & Park, 2015; Lee
& Choeh, 2016). For this reason, many platforms have adapted verification
techniques, such as indicators like "Verified Buyer" or "Verified Stay"
badges, further enhancing perceptions of legitimacy (Khalid, 2019; Lappas,
Sabnis & Valkanas, 2016). This suggests that the validation components
in VI, more specifically the presence of genuine name and a "Verified
Stay" badge, acted as strong indicators of accountability and trustworthiness.
As such, respondents presented with the VI condition likely
perceived the reviews as originating from an identifiable, trustworthy
author, strengthening perceptions of authenticity through identitybased
credibility, contrasting with Specificity, which enhanced authenticity
primarily through factual detail.
Taken together, the results provide an ordering of how the studied
review cues affected authenticity perceptions in the sample. Overall, the
cues form a coherent gradient from highly inauthentic (II) through moderately
inauthentic (01), followed by a pair of authenticity enhancing
conditions (SI and VI). This pattern offers a theoretically grounded answer
to RQ1, indicating the relative strength by which different cues may
influence consumers' evaluations of the authenticity of online reviews.
As for RQ2: How does perceived authenticity influence booking
intention?, the findings showed a significant positive association (r =
.82) between perceived authenticity and booking intention. Although
these correlations do not suggest causality, the trend is consistent with
studies indicating that customers' reactions to online reviews are
strongly shaped by their perceived credibility (Filieri et al., 2015). As
credible and trustworthy reviews help reduce uncertainty and boost decision
confidence, it may explain why perceived authenticity emerged as
such a strong predictor of booking related intentions in the present
study. This interpretation is particularly relevant in the hospitality context,
where consumers tend to rely on online reviews more heavily than
in many other domains. Reviews tend to carry a significant impact on
travellers' behavioral intentions, as they are often seen as a more
70
DISCUSSION
relevant source of information than traditional advertising (Soares et al.,
2022; Xiao etal., 2021). Because of this increased dependence, decisions
pertaining to hotels may be more heavily influenced by credibility signals
embedded in reviews. Given that customers rely more heavily on peer
generated information, cues that strengthen or weaken credibility can
have a significant impact on their willingness to book.
This reasoning is further supported by the group comparisons. Similar
to their detrimental effect on authenticity, conditions Implausibility
(II) and Overlap (01) produced significantly lower booking intentions
than the Baseline condition. Respondents seemed to dismiss a review's
informational value after it triggered suspicions, whether through implausible
claims or repetitive phrasing subsequently decreasing their
willingness to make a reservation. This is in line with previous research,
showing that cues signalling manipulation or low credibility increase
mistrust and reduce customers' reliance on the information (Filieri et al.,
2015; Wang et al., 2023). By contrast, conditions Specificity (SI) and Validation
(VI) did not raise booking intention over baseline levels, suggesting
that credibility enhancing elements may stabilize rather than
raise evaluations once a neutral level of trust is achieved.
The mechanism was further expanded on by the mediation analysis.
The cues' direct effects on booking intention were not substantial, while
their indirect effects through authenticity were significant for conditions
Implausibility and Overlap. This implies that negative cues affected
booking intention by first lowering authenticity perceptions, subsequently
translating into a decreased willingness to book. Such patterns
further reflect evidence that reviews that are seen as misleading or superficial
reduce their persuasive impact, whereas those that appear
trustworthy or grounded in experience tend to produce more favourable
behavioral responses (Kim, Park & Mariani, 2023).
Overall, these findings point to a theoretically sound response to
RQ2, suggesting that perceived authenticity acts as an important psychological
filter connecting review cues to booking intention. In this study,
maintaining authenticity was associated with rather stable booking intentions,
while decreased authenticity perceptions were consistently
linked to lower willingness to book. Thus, authenticity appears to play a
central role in eWOM processing shaping the extent to which review information
becomes persuasive, is treated as neutral, or is largely disre-
garded.
71
THEORETICAL CONTRIBUTIONS AND MANAGERIAL IMPLICATIONS
6 Theoretical contributions and managerial
implications
Based on the empirical results, the following section summarises the
main theoretical contributions and practical recommendations derived
from the study.
The first theoretical implication is that as opposed to establishing a
binary distinction between "authentic" and "fake", the review cues were
ranked on a credibility gradient. Several authenticity and fakery cues
have been identified in previous research, namely Harrison-Walker and
Jiang (2023) and Bucher (2024), however their relative value has not
been compared. By empirically ranking these cues, the current study expands
on this concept by demonstrating that Implausibility carries the
most detrimental impact on perceived authenticity, followed by Overlap,
while Specificity and Validation serve as signals moderately increasing
authenticity, but not exceeding baseline levels. These cue effects are in
line with Bucher's results, which indicated that specificity and validation
were indicators of increased authenticity, while implausibility and overlap
were the main indicators of low credibility, while further expanding
on these results by providing a more nuanced understanding of how consumers
interpret contextual cues in online reviews by ranking these cues
based on their impact.
Furthermore, the results support that the primary mechanism tying
review cues to booking intention is perceived authenticity. Authenticity
showed a very strong correlation with booking intention, and mediation
analyses revealed that cue effects on in booking intention were transmitted
almost entirely through perceived authenticity. These results provide
experimental proof of authenticity's vital role in eWOM processing
and complement previous studies emphasizing credibility as a major factor
influencing consumer decision making in the hospitality industry
(Filieri et al., 2015).
Lastly, the study suggests an asymmetry in how consumers react to
cues that increase or decrease credibility. While Specificity and Validation
purely assisted in keeping both variables at baseline levels, without
raising them further, Implausibility and Overlap significantly decreased
both perceived authenticity and booking intention in comparison to the
Baseline condition. This pattern indicates that while positive cues do not
further increase credibility beyond a neutral point, negative cues may
72
THEORETICAL CONTRIBUTIONS AND MANAGERIAL IMPLICATIONS
have disproportionately greater negative effects. This is consistent with
research demonstrating that reviews that appear overly positive or deceptive
are frequently viewed as suspicious and have the potential to disproportionately
impact credibility assessments and future booking decisions
(Filieri, 2016; Kollmer et al., 2022).
The study findings also carry several practical implications for hospitality
providers and online review platforms.
Firstly, as results suggest the Validation condition helped maintain
perceived authenticity at baseline levels, shielding booking intention
from the declines seen with negative cues, experience and identity-based
verification signals should be actively strengthened by hospitality providers
and online review platforms. This could entail enhancing or introducing
forms of "Verified Stay" badges on platforms that do not currently
have them, encouraging visitors to write reviews using their real names
or providing extra features such as the ability to attach photos from the
stay. Hotels may facilitate this by simply suggesting guests to enclose
these details during check-out, or in the case of on-site review systems,
by designing the review interface to encourage the use of identifiable information
or uploading a photo as part of the feedback process.
Furthermore, accommodation providers should try to encourage
customers to create detailed and experience-rich reviews. As seen in the
results, Specificity reduced uncertainly and helped in maintaining perceived
authenticity and preventing declines in booking intention. Hotels
may encourage this by subtly directing guests towards specific information,
such as asking "What specifically did you enjoy during your
stay?", or by providing guiding subcategories like breakfast quality,
check-in experience, cleanliness, noise levels and so on.
Finally, the findings imply that it is strategically counterproductive
to create or solicit positive fake reviews. Although competitive pressure
in the hospitality sector may steer businesses towards artificially inflating
their rating Zhang et al. (2022), these fabricated reviews often contain
implausible statements or repetitive, template-like phrasing. As is
demonstrated in this study, such cues substantially reduce perceived authenticity
and, through that mechanism, lower booking intention. These
patterns frequently emerge when reviews are written by paid third parties
or automated bots, or when overly promotional language is used in
an attempt to improve ratings. When combined with further evidence
that inauthentic feedback erodes trust and diminishes the effectiveness
of online reviews (Balakrishnan et al., 2024), hotels are to be advised that
73
THEORETICAL CONTRIBUTIONS AND MANAGERIAL IMPLICATIONS
these attempts to manipulate review profiles may ultimately harm rather
than strengthen market performance.
74
LIMITATIONS AND FUTURE RESEARCH
7 Limitations and future research
Naturally, the present study holds several limitations, which are outlined
in this chapter, together with possible directions for future research.
The first limitation of the study concerns the use of Al-generated,
fictitious review stimuli. While this method allowed for a more precise
isolation of the examined conditions, it removed the richness and variability
of naturally occurring online reviews, which often contain multiple
overlapping cues, emotional tone or further contextual detail. In order to
improve ecological validity, future research could examine actual review
datasets or use hybrid designs that combine controlled manipulations
with real review content.
Furthermore, the participants generally evaluated a single review
(barring the Overlap condition, where two similar reviews were necessary
to create the repetition cue) before being asked to assess their willingness
to book. Although this approach was methodologically necessary,
it does not reflect real decision-making environments, where on average,
customers usually process up to nine reviews for each hotel they
are considering staying in (Harden-England, 2024). Therefore, future research
could investigate how consumers prioritize conflicting cues
within larger sets of reviews or test more realistic multi-review settings.
The study was also limited by the constraints of the experimental
design, which sought to minimise confounding factors by making the reviews
as similar as possible, for example by unifying length and avoiding
overly vivid elements. As a result, each condition was only represented
by one specific sub-item (e.g. one form of implausibility, one form of validation).
A follow-up study could investigate several operationalizations
of each condition, using different sub-items, or even their various combinations,
as is often the case in realistic review environments.
Additionally, the decision not to exclude manipulation check (MC)
failures could be seen as a limitation, as eliminating these respondents
would have reduced the usable sample to N = 274. Nevertheless, additional
analyses utilizing the MC-pass subsample produced the same general
pattern of results, indicating that this decision had little impact on
the study's findings.
Finally, the data were collected exclusively in the context of the
Czech Republic, and the sample skewed young, with 82% of respondents
being between the ages of 18 and 34. Since perceptions of credibility,
75
LIMITATIONS AND FUTURE RESEARCH
communication norms, and sensitivity to linguistic cues may differ
across regions and age groups, this restricts the findings' cultural and demographic
generalizability. Older travellers, for instance, may base their
decisions on altered heuristics or place more weight on different cues
than younger, digitally savvy users, who made up the majority of the
studied sample. In order to assess whether the cue hierarchy and behavioral
effects seen in this study hold true for larger populations, future research
may benefit from employing cross-country or multilingual designs
and more age-diverse samples.
76
CONCLUSION
8 Conclusion
The aim of the thesis was to examine how various review cues influence
consumers' perceptions of authenticity, and consequently, their intentions
of booking a hotel, in the context of the broader issue of fake online
reviews. Following up on existing research on eWOM, review credibility
and fake review detection, the study focused on cues that prior literature
has identified as significant markers of either genuine experiences (reviewer
or stay validation; detailed, specific review content) or as indicators
of possible fabrication (implausible content; overlapping, templatelike
reviews). Although these cues were established in previous research,
they have rarely been compared directly or ranked in terms of their relative
strength of influence. The present experiment therefore aimed to
contribute to that gap by testing these cues alongside one another and
exploring whether they can be meaningfully organised from the most inauthentic
to the most authentic review patterns.
In the theoretical part of the thesis, the foundations were established
by defining WOM and eWOM and describing the main distinctions
between traditional and digital word-of-mouth. Subsequently, differences
between fake and authentic reviews were summarized along with
the cues typically associated with each. Ultimately, the motivations behind
the creation of fake reviews, along with machine learning and human-based
methods for identifying fake reviews were explored.
In the practical part, an online experiment was carried out with the
aim of investigating how four review conditions (Specificity, Validation,
Overlap, and Implausibility) affect perceived authenticity and booking
intention. Survey respondents were randomly assigned to one of five
conditions, each of which exposed them to a carefully manipulated review.
The participants were then asked to evaluate the review's perceived
authenticity and their willingness to make hotel reservations after
reading it. The results of the subsequent analyses suggested a hierarchy
of cue effects, where Specificity and Validation assisted in maintaining
authenticity at baseline levels, while Implausibility and Overlap produced
the largest reductions in authenticity and booking intention.
Additionally, subsequent correlation and mediation analyses revealed
that the main mechanism connecting cue exposure to booking intention
was perceived authenticity.
77
CONCLUSION
Overall, this thesis contributes by providing a concise theoretical
summary of fake online reviews in the hospitality industry and by
providing empirical evidence of how particular review cues influence
perceived authenticity and booking intention. The findings clarify the
mechanisms through which review content influences consumer decision-making
and provide insight useful for both research and practice in
building more trustworthy review environments
78
BIBLIOGRAPHY
Bibliography
Abri, F., Gutierrez, L. F., Namin, A S., Jones, K. S., & Sears, D. R. W. (2020).
Linguistic Features for Detecting Fake Reviews. In 2020 19th IEEE InterNational
Conference on Machine Learning and Applications
(ICMLA). https://doi.org/10.1109/icmla51294.2020.00063
Alsaad, N. M. M. B. (2024). Transformer-Based Language Deep Learning
detection of fake reviews on online products. Deleted Journal, 20(3),
2368-2378. https://doi.org/10.52783/jes.4083
Arndt, J. (1967). Role of Product-Related Conversations in the Diffusion
of a New Product. Journal of Marketing Research, 4, 291-295.
https://doi.org/10.1177/00222437670040030
Aronow, P. M., Baron, J., & Pinson, L. (2019). A Note on Dropping Experimental
Subjects who Fail a Manipulation Check. Political Analysis, 27(4),
572-589. https://doi.Org/10.1017/pan.2019.5
Baek, H., Ahn, J., & Choi, Y. (2012). Helpfulness of Online Consumer Reviews:
Readers' objectives and review cues. International Journal of Electronic
Commerce, 17[2], 99-126. https://doi.org/10.2753/jecl086-
4415170204
Balakrishnan, N., Chandru, C, Karimulla Sheriff, K., & Priyadharshini, T.
(2024). Decoding Authenticity in Online Review using Machine Learning.
2024 International Conference on Cognitive Robotics and Intelligent
Systems (ICC ROBINS), 34, 421-424. https://doi.org/10.1109/icc-rob-
ins60238.2024.10533938
Banerjee, S. (2022). Exaggeration in fake vs. authentic online reviews
for luxury and budget hotels. International Journal of Information Management,
62,102416. https://doi.Org/https://doi.org/10.1016/j.ijinfo-
mgt.2021.102416
Banerjee, S., & Chua, A. Y. K. (2014). A linguistic framework to distinguish
between genuine and deceptive online reviews. 501-506. Paper presented
at International MultiConference of Engineers and Computer
Scientists, Hong Kong.
79
BIBLIOGRAPHY
Banerjee, S., & Chua, A. (2017). Theorizing the textual differences between
authentic and fictitious reviews: Validation across positive, negative,
and moderate polarities. Internet Research, 27(2), 321-337.
https://doi.org/10.1108/IntR-ll-2015-0309
Banerjee, S., & Chua, A. Y. K. (2021). Calling out fake online reviews
through robust epistemic belief. Information & Management, 58(3),
103445. https://doi.Org/https://doi.org/10.1016/j.im.2021.103445
Bassig M. (n.d.). Tools for Fake Review Detection: How to Spot Fake Reviews.
ReviewTrackers. Retrieved January 13, 2025, from
https://www.reviewtrackers.com/blog/fake-reviews/
Blackwood, M. (2020). Hundreds of Fake Reviews / Fake 5-Star Review
Network. Google Business Profile Help Community. Retrieved January
13, 2025, from https://support.google.com/busi-
ness/thread/25606944?hl=en&msgid=26834961
Blank, T. J. (2015). Faux your entertainment: Amazon.com product reviews
as a locus of digital performance. Journal of American
Folklore, 128(509), 286-297. https://doi.org/10.5406/jamer-
folk.128.509.0286
Bucher, P. (2024). Skutečný problém online recenzí. Diplomová práce.
Brno: Masarykova univerzita, Ekonomicko-správní fakulta. Available at:
https://is.muni.cz/auth/th/mo07y/
Bulchand-Gidumal, ]., & Melián-González, S. (2023). Fighting fake
reviews with blockchain-enabled consumer-generated reviews. Current
Issues in Tourism, 27(5), 739-753.
https://doi.org/10.1080/13683500.2023.2173054
Burgoon, J. K., & Buller, D. B. (2015). Interpersonal deception theory.
The International Encyclopedia of Interpersonal Communication, 1-
6. https://doi.org/10.1002/9781118540190.wbeicl70
Carr, C. T., & Piercy, C. W. (2021). Linguistic cues to misrepresentation
in online employer reviews. Communication Reports, 34(3), 179-191.
https://doi.org/10.1080/08934215.2021.1973053
80
BIBLIOGRAPHY
Chatterjee, S., Chaudhuri, R., Kumar, A., Wang, C. L., & Gupta, S. (2023).
Impacts of consumer cognitive process to ascertain online fake review:
A cognitive dissonance theory approach. Journal of Business Research,
154, 113370. https://doi.Org/10.1016/j.jbusres.2022.113370
Cheung, C. M., & Thadani, D. R. (2012). The impact of electronic wordof-mouth
communication: A literature analysis and integrative
model. Decision Support Systems, 54(1), 461-470.
https://doi.Org/10.1016/j.dss.2012.06.008
Chua, A. Y. K, & Chen, X. (2022). Online "helpful" Lies: An Empirical
Study of Helpfulness in Fake and Authentic Online Reviews. In Lecture
notes in computer science (pp. 91-99).
https://doi.org/10.1007/978-3-030-96957-8_10
Crawford, M., Khoshgoftaar, T. M., Prusa, J. D., Richter, A. N., & Najada, H.
A. (2015). Survey of review spam detection using machine learning
techniques. Journal of Big Data, 2(1).
https://doi.org/10.1186/s40537-015-0029-9
Daugherty, T., & Hoffman, E. (2013). eWOM and the importance of capturing
consumer attention within social media. Journal of Marketing
Communications, 20(1-2), 82-102.
https://doi.org/10.1080/13527266.2013.797764
Dellarocas, C. (2003). The digitization of word of mouth: Promise and
challenges of online feedback mechanisms. Management Science,
49(10), 1407-1024. https://doi.org/10.1287/mnsc.49.10.1407.17308
DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K,
& Cooper, H. (2003). Cues to deception. Psychological Bulletin, i29(l),
74-118. https://doi.Org/10.1037/0033-2909.129.l.74
DePompa, R., & Molina, D. (2022, October 17). Five Star Fakes: Small
business owners feel big impact from fake online reviews. KJCT8. Retrieved
January 13, 2025, from
https://www.kjct8.com/2022/10/17/five-star-fakes-small-business-
owners-feel-big-impact-fake-online-reviews/
Donaker, G., Hyunjin, K, Luca, M. (2019) "Designing Better Online Review
Systems." Harvard Business Review 97, no. 6 (November-December
2019): 122-129.
81
BIBLIOGRAPHY
Dragan, L. (2016). Let's Talk About Amazon Reviews: How We Spot the
Fakes. Wirecutter. Retrieved October 17, 2024, from https://www.ny-
times.com/wirecutter/blog/lets-talk-about-amazon-reviews/
Fernandes, S., Panda, R., Venkatesh, V. G., Swar, B. N., & Shi, Y. (2022).
Measuring the impact of online reviews on consumer purchase
decisions - A scale development study. Journal of Retailing and Consumer
Services, 68,103066.
https://doi.Org/10.1016/j.jretconser.2022.103066
Field, A. (2018). Discovering statistics using IBM SPSS Statistics. SAGE
Publications Limited.
Filieri, R, Alguezaui, S., & McLeay, F. (2015). Why do travelers trust
TripAdvisor? Antecedents of trust towards consumer-generated media
and its influence on recommendation adoption and word of mouth.
Tourism Management, 51,174-185. https://doi.org/10.1016/j.tour-
man.2015.05.007
Finstad, K. (2010). Response interpolation and scale sensitivity: evidence
against 5-point scales. Journal of Usability Studies Archive, 5(3),
104-110. https://dl.acm.org/ft_gate-
way.cfm?id=2835437&ftid=1632947&dwn=l
Fong, L. H. N., Ye, B. H., Leung, D., & Leung, X. Y. (2022). Unmasking the
imposter: Do fake hotel reviewers show their faces in profile pictures?
Annals of Tourism Research, 93,
103321. https://doi.Org/10.1016/j.annals.2021.103321
Gabbard, D. (2023). The Impact of Online Reviews on Hotel
Performance. Journal of Modern Hospitality, 2, 26-36.
https://doi.org/10.47941/jmh.1558
Ghasemi, A., & Zahediasl, S. (2012). Normality Tests for Statistical Analysis:
A Guide for Non-Statisticians. International Journal of Endocrinology
and Metabolism, 10(2), 486-489.
https://doi.org/10.5812/ijem.3505
82
BIBLIOGRAPHY
Gossling, S., Hall, C. M., & Andersson, A. (2018). The manager's dilemma:
a conceptualization of online review manipulation strategies. Current Issues
in Tourism, 21(5), 484-503.
https://doi.org/10.1080/13683500.2015.1127337
Gravetter, F. J., & Wallnau, L. B. (2016). Statistics for the Behavioral Sciences.
Cengage Learning.
Harden-England, K. (2024, June 17). Accor Research reveals "Almost
every hotel or resort customer reads guest reviews." Travolution. Retrieved
December 5, 2025, from https://www.travolu-
tion.com/news/accor-research-reveals-almost-every-hotel-or-resort-
customer-reads-guest-reviews/
Harrison-Walker, L. J., & Jiang Y. (2023). Suspicion of online product reviews
as fake: Cues and consequences. Journal of Business Research, 160,
113780. https://doi.Org/10.1016/j.jbusres.2023.113780
Hauch, V., Blandon-Gitlin, I., Masip, J., & Sporer, S. L. (2015). Are computers
effective lie detectors? A Meta-Analysis of Linguistic Cues to Deception.
Personality and Social Psychology Review, 19(4), 307-342.
https://doi.org/10.1177/1088868314556539
He, S., Hollenbeck, B., Overgoor, G., Proserpio, D., & Tosyali, A. (2021).
Detecting fake-review buyers using network structure: Direct evidence
from Amazon. Proceedings of the National Academy ofSciences,
119(47). https://doi.org/10.1073/pnas.2211932119
Huang, Y., Miguel Villas-Boas, J., Zhao, M., thank Nano Barahona, W.,
Chen, X., Dean, M., Gaertig C, Handel, B., He, S., Hollenbeck, B., Iyer, G.,
Jeziorski, P., Kawai, K., Kolstad, J., Li, P., Moshary, S., Natan, 0., & Tadelis,
S. (2024). Unmasking the Deception: The Interplay between Fake Reviews,
Ratings Discrepancy, and Consumer Demand.
http://dx.doi.org/10.2139/ssrn.4621736
Hu, N., Bose, I., Koh, N. S., & Liu, L. (2012). Manipulation of online
reviews: An analysis of ratings, readability, and sentiments. Decision
Support Systems, 52(3), 674-684.
https://doi.Org/10.1016/j.dss.2011.ll.002
Huete-Alcocer, N. (2017). A Literature review of Word of Mouth and
Electronic Word of Mouth: Implications for Consumer Behavior. Frontiers
in Psychology, 8. https://doi.org/10.3389/fpsyg.2017.01256
83
BIBLIOGRAPHY
Istanbulluoglu, D., & Harris, L. C. (2023). Forms of falsified online reviews:
the good, the bad, and the downright ugly. European Journal of
Marketing, https://doi.org/10.1108/ejm-12-2022-0904
Jha, A. K., & Shah, S. (2021). Disconfirmation effect on online review
credibility: An experimental analysis. Decision Support Systems, 145,
113519. https://doi.Org/10.1016/j.dss.2021.113519
Katz, E., & Lazarsfeld, P. F. (1966). Personal influence: the part played
by people in the flow of mass communications. The American Catholic
Sociological Review, 17(4), 357. https://doi.org/10.2307/3708755
Khalid, A. (2019). What to know about Fake reviews when shopping
online. Quartz. Retrieved October 14, 2024,
from https://qz.com/1758846/how-to-detect-fake-reviews-while-
shopping-online
Kim, J. M., Park, K. K., & Mariani, M. M. (2023). Do online review readers
react differently when exposed to credible versus fake online reviews?
Journal of Business Research, 154,113377.
https://doi.Org/10.1016/j.jbusres.2022.113377
King, R. A., Racherla, P., & Bush, V. D. (2014). What we know and don't
know about Online Word-of-Mouth: A review and synthesis of the literature.
Journal of Interactive Marketing 28(3), 167-183.
https://doi.Org/10.1016/j.intmar.2014.02.001
Kollmer, T., Eckhardt, A., & Reibenspiess, V. (2022). Explaining consumer
suspicion: insights of a vignette study on online product reviews.
Electronic Markets, 32(3), 1221-1238.
https://doi.org/10.1007/sl2525-022-00549-9
Lappas, T., Sabnis, G., & Valkanas, G. (2016). The impact of fake reviews
on Online Visibility: A vulnerability assessment of the hotel industry. Information
Systems Research, 27(4), 940-961.
https://doi.org/10.1287/isre.2016.0674
Lee, S., & Choeh, J. Y. (2016). The determinants of helpfulness of online
reviews. Behaviour and Information Technology, 35(10), 853-
863. https://doi.org/10.1080/0144929x.2016.1173099
84
BIBLIOGRAPHY
Le, L. T., Ly, P. T. M., Nguyen, N. T., & Tran, L. T. T. (2022). Online reviews
as a pacifying decision-making assistant. Journal of Retailing and
Consumer Services, 64,102805.
https://doi.Org/https://doi.org/10.1016/j.jretconser.2021.102805
Li, L., Lee, K. Y., Lee, M., & Yang, S. (2020). Unveiling the cloak of deviance:
Linguistic cues for psychological processes in fake online reviews.
International Journal of Hospitality Management, 87,
102468. https://doi.Org/10.1016/j.ijhm.2020.102468
Ling Q. (2023). Machine learning algorithms review. Applied and Computational
Engineering, 4(1), 91-98. https://doi.org/10.54254/2755-
2721/4/20230355
Liu, H., Jayawardhena, C, Shukla, P., Osburg V., & Yoganathan, V.
(2024). Electronic word of mouth 2.0 (eWOM 2.0) - The evolution of
eWOM research in the new age. Journal of Business Research, 176,
114587. https://doi.Org/10.1016/j.jbusres.2024.114587
Liu, Z., & Park, S. (2015). What makes a useful online review? Implication
for travel product websites. Tourism Management, 47,140-
151. https://doi.Org/10.1016/j.tourman.2014.09.020
Ly, N. T. H., & Huyen, P. T. (2023). EWOM and purchase intention: A
brief look at the last decade by Bibliometrics. International Journal of
Research in Marketing Management and Sales, 5(2), 40-47.
https://doi.org/10.33545/26633329.2023.v5.i2a.131
Malik, A. Z., R, K. S., Paswan, A., & Kanwal, F. (2024). Incentivized fake
reviews: When cognitive reappraisal paves the way for an immoral
journey. Journal of Consumer Behaviour.
https://doi.org/10.1002/cb.2315
Mladenovic, D., Beheshti, M., Kolar, T., Ismagilova, E., & Dwivedi, Y. K.
(2024). Synthetic W0M? The emergence of Generative Artificial Intelligence-Induced
Recommendations. Journal of Computer Information
Systems, 1-18. https://doi.org/10.1080/08874417.2024.2409252
Mladenovic, D., Ismagilova, E., Filieri, R., & Dwivedi, Y. K. (2023b).
MetaWOM - toward a sensory word-of-mouth (WOM) in the metaverse.
International Journal of Contemporary Hospitality Management, 36(6),
2144-2163. https://doi.org/10.1108/ijchm-04-2023-0474
85
BIBLIOGRAPHY
Mohawesh, R., Salameh, H. B., Jararweh, Y., Alkhalaileh, M., & Maqsood,
S. (2024). Fake review detection using transformer-based enhanced
LSTM and RoBERTa. International Journal of Cognitive Computing in Engineering,
5, 250-258. https://doi.Org/10.1016/j.ijcce.2024.06.001
Moon, S., Kim, M., & Iacobucci, D. (2021). Content analysis of fake consumer
reviews by survey-based text categorization. International Journal
of Research in Marketing, 38(2), 343-364.
https://doi.Org/10.1016/j.ijresmar.2020.08.001
Ögüt, H., & Onur Tas, B. K. (2011). The influence of internet customer
reviews on the online sales and prices in hotel industry. The Service
IndustriesJournal, 32(2), 197-214.
https://doi.org/10.1080/02642069.2010.529436
Pennebaker, J. W., Boyd, R. L., Jordan, K, & Blackburn, K. (2015). The
Development and Psychometric Properties of LIWC2015. The University
of Texas at Austin, https://repositories.lib.utexas.edu/bit-
stream/2152/31333/3/LIWC2015_LanguageManual.pdf
Petrescu, M., O'Leary, K, Goldring D., & Mrad, S. B. (2018). Incentivized
reviews: Promising the moon for a few stars. Journal of Retailing and
Consumer Services, 41, 288-295.
https://doi.Org/10.1016/j.jretconser.2017.04.005
Plotkina, D., Munzel, A., & Pallud, J. (2020). Illusions of t r u t h Experimental
insights into human and algorithmic detections of fake
online reviews. Journal of Business Research, 109, 511-523.
https://doi.Org/10.1016/j.jbusres.2019.ll.069.
Pooja, K, & Upadhyaya, P. (2022). What makes an online review credible?
A systematic review of the literature and future research directions.
Management Review Quarterly, https://doi.org/10.1007/sll301-
022-00312-6
Porritt, S. (2017). Do Online Reviews Impact Some Industries More
Than Others? Brightlocal. From: https://www.bright-
local.eom/blog/online-reviews-impact-industries-others/#
86
BIBLIOGRAPHY
Porter, S. & Brinke, L. T., (2012). Cry me a river: Identifying the behavioral
consequences of extremely high-stakes interpersonal deception.
Law And Human Behavior, 36(6), 469-477.
https://doi.org/10.1037/h0093929
Rabušic, L., Soukup, P., & Mareš, P. (2019). Statistická analýza sociálněvědních
dat (prostřednictvím SPSS).
Ren, Z., Zhang X., Zhang, S., & Yang C. (2023). Fake review detection via
heterogeneous graph attention network. In Lecture notes in computer
science (pp. 364-376).
https://doi.org/10.1007/978-3-031-44216-2_30
Román, S., Riquelme, I. P., & Iacobucci, D. (2022). Fake or credible? Antecedents
and consequences of perceived credibility in exaggerated
online reviews. Journal of Business Research, 156,113466.
https://doi.Org/10.1016/j.jbusres.2022.113466
Salehan, M., & Kim, D. J. (2016). Predicting the performance of online
consumer reviews: A sentiment mining approach to big data analytics.
Decision Support Systems, 81, 30-40.
https://doi.Org/10.1016/j.dss.2015.10.006.
Salokhe, A. (2024). Fake Reviews Detection using Supervised Machine
Learning Algorithms. INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH
IN ENGINEERING AND MANAGEMENT, 08(05), 1-5.
https://doi.org/10.55041/ijsrem34850
Sanjay, S. K, Danti, A., Reddy, T. S., & Nath, V. (2024). Deep learningbased
model for computing percentage of fake in user reviews using
topic modelling techniques. e-Prime - Advances in Electrical Engineering
Electronics and Energy, 8,
100506. https://doi.Org/10.1016/j.prime.2024.100506
Schoolov, K. (2020). Amazon is filled with fake reviews and it's getting
harder to spot them, [online] Retrieved from
https://www.cnbc.com/2020/09/06/amazon-reviewsthousands-are-
fake-heres-how-to-spot-them.html
Sharma, H., & Aggarwal, A. G. (2019). What factors determine reviewer
credibility? Kybernetes, 49(10), 2547-2567.
https://doi.org/10.1108/k-08-2019-0537
87
BIBLIOGRAPHY
Shukla, Aishwarya and Wang, Weiguang and Gao, Guodong (Gordon)
and Agarwal, Ritu, Catch Me If You Can (2019) — Detecting Fraudulent
Online Reviews of Doctors Using Deep Learning.
http://dx.doi.org/10.2139/ssrn.3320258
Siddiqi, U. I., Sun, J., & Akhtar, N. (2020). Ulterior motives in peer and
expert supplementary online reviews and consumers' perceived deception.
Asia Pacific Journal of Marketing and Logistics, 33(1), 73-98.
https://doi.org/10.1108/apjml-06-2019-0399
Soares, M. A., Dolci, D. B., & Lunardi, G. L. (2022). Interaction Effects between
Online Reviews and Product Characteristics on Consumer's
Trust. Revista De Administratedo Contempordnea, 26(suppl 1).
https://doi.org/10.1590/1982-7849rac2022210114.en
Thakur, R. (2018). Customer engagement and online reviews. Journal of
Retailing and Consumer Services, 41, 48-59.
https://doi.Org/10.1016/j.jretconser.2017.ll.002
Varaine, S. (2022). How dropping subjects who failed manipulation
checks can bias your results: an illustrative case. Journal of Experimental
Political Science, 10(2), 299-305.
https://doi.org/10.1017/xps.2022.28
Victor, V, James, N., & Dominic, E. (2024). Incentivised dishonesty:
Moral frameworks underlying fake online reviews. International Journal
of Consumer Studies, 48(2). https://doi.org/10.llll/ijcs.13037
Vrij, A., Hartwig, M., & Granhag, P. A. (2019). Reading Lies: Nonverbal
communication and deception. Annual Review of Psychology, 70(1),
295-317. https://doi.org/10.1146/annurev-psych-010418-103135
Walther, M., Jakobi, T., Watson, S. J., & Stevens, G. (2023). A systematic
literature review about the consumers' side of fake review detection Which
cues do consumers use to determine the veracity of online user
reviews? Computers in Human Behavior Reports, 10,100278.
https://doi.org/10.1016/jxhbr.2023.100278
88
BIBLIOGRAPHY
Wang, E. Y., Fong, L. H. N., & Law, R. (2021). Detecting fake hospitality
reviews through the interplay of emotional cues, cognitive cues and review
valence. International Journal of Contemporary Hospitality Management,
34(1), 184-200.
https://doi.org/10.1108/ijchm-04-2021-0473
Wang, Z., Hou, T., Li, Z., & Song, D. (2015). Spotting fake reviewers using
product review graph. Journal of Computational Information Systems,
11(16), 5759-5767.
Wang, B., & Kuan, K. K. (2022). Understanding the Message and Formulation
of Fake Online Reviews: A Language-production Model Perspective.
AIS Transactions on Human-Computer Interaction, 14(2), 207-229.
https://doi.org/10.17705/lthci.00167
Wang, Y., Zamudio, C, & Jewell, R. D. (2023). The more they know: Using
transparent online communication to combat fake online reviews. Business
Horizons, 66(6), 753-764.
https://doi.Org/10.1016/j.bushor.2023.03.004
wikiHow. (2019). How to Spot Fake Product Reviews Online. Retrieved
January 14, 2025, from https://www.wikihow.life/Spot-Fake-Product-
Reviews-Online#
Wu, Y., Ngai, E. W. T., Wu, P., & Wu, C. (2020). Fake online reviews:
Literature review, synthesis, and directions for future research. Decision
Support Systems, 132,113280.
https://doi.Org/10.1016/j.dss.2020.113280
Wu, R., & Qiu, C. (2023). When Karma strikes back: A model of seller
manipulation of consumer reviews in an online marketplace. Journal of
Business Research, 155,113316.
https://doi.Org/10.1016/j.jbusres.2022.113316
Xiao, Q., Li, S., Zhang X., Zhang F., Yue, Q., & Wan, S. (2021). Deconstructing
online hospitality review systems. Journal of Organizational
and End User Computing, 34(2), 1-17.
https://doi.org/10.4018/joeuc.292523
89
BIBLIOGRAPHY
Yan, L., & Hua, C. (2021). Which reviewers are honest and caring? The
effect of constructive and prosocial information on the perceived credibility
of online reviews. International Journal of Hospitality Management,
99,102990. https://doi.Org/10.1016/j.ijhm.2021.102990
Zaman, M., Vo-Thanh, T., Nguyen, C. T., Hasan, R., Akter, S., Mariani, M., &
Hikkerova, L. (2023). Motives for posting fake reviews: Evidence from a
cross-cultural comparison. Journal of Business Research, 154,
113359. https://doi.Org/10.1016/j.jbusres.2022.113359
Zhang D., Zhou, L., Kehoe, J. L., & Kilic, I. Y. (2016). What Online
Reviewer Behaviours Really Matter? Effects of Verbal and Nonverbal
Behaviors on Detection of Fake Online Reviews. Journal of Management
Information Systems, 33{2], 456-481.
https://doi.org/10.1080/07421222.2016.1205907
Zhang Z., Li, Y., Li, H., & Zhang, Z. (2022). Restaurants' motivations to
solicit fake reviews: A competition perspective. International Journal of
Hospitality Management, 107,103337.
https://doi.Org/10.1016/j.ijhm.2022.103337
Zhang Z., Wang, X., & Wu, R. (2021). Is the devil in the details? Construal-level
effects on perceived usefulness of online reviews for experience
services. Electronic Commerce Research and Applications, 46,
101033. https://doi.Org/10.1016/j.elerap.2021.101033
90
QUESTIONNAIRE COPY
Appendix A Questionnaire copy
Dobrý den,
jmenuji se Matěj Prachař a jsem student Ekonomicko správní fakulty
Masarykovy univerzity.
Tento krátký dotazník je součástí mé diplomové práce na téma falešných
online recenzí v oblasti ubytovacích služeb.
Cílem tohoto výzkumu je zjistit, jak různé prvky textu recenze ovlivňují
její vnímanou autenticitu a případně ochotu rezervovat ubytování.
Co je „falešná online recenze"?
Falešná či klamavá online recenze je taková, která je napsaná s úmyslem
ovlivnit hodnocení podniku - buď ho neoprávněně zlepšit, nebo
naopak poškodit jeho pověst.
Taková recenze může být napsána člověkem i automaticky (robotem),
často na objednávku nebo z osobních důvodů.
Účast je dobrovolná a anonymní. Nejsou shromažďovány citlivé osobní
údaje; odpovědi budou použily výhradně pro akademické účely.
Vyplnění zabere asi 5 minut. Budu moc rád, když dotazník rozešlete dál
- každá odpověď se počítá. Děkuji mnohokrát!
Kontakt:
494081@mail.muni.cz
matejprachar@gmail.com
Děkuji za Vaši účast!
Pokračováním a odesláním souhlasu s účastí potvrzujete, že jste starší 18
let a souhlasíte se zpracováním anonymních odpovědí pro účely této studie.
91
QUESTIONNAIRE COPY
Ol: Čtete alespoň někdy online recenze hotelů nebo restaurací?
• Ano
• Ne
V následujícím bloku uvidíte krátkou online recenzi/recenze
týkající se pobytu v hotelu. Prosím, přečtěte si text pozorně.
Poté zodpovíte několik krátkých otázek k vašemu dojmu. Odpovídejte
podle svého pocitu, neexistují správné či špatné od­
povědi.
Hotel působil celkově velmi příjemně. Pokoj byl čistý a postel pohodlná, na recepci nás
přivítal vstřícný personál. Snídaně měla dostatečný výběr a prostředí jídelny bylo klidné.
Lokalita umožňuje snadno se dostat kamkoli, a přes noc byl v hotelu klid. Pobyt
hodnotím pozitivně, vše proběhlo bez komplikací a cítil jsem se zde dobře. Za mě dobrá
volba pro kratší i delší návštěvu města. Rád bych se vrátil, pokud budu v této oblasti
znovu ubytování potřebovat.
Note - Displayed review differedfor every questionnaire version, the present one is BO
Na základě zobrazené recenze prosím ohodnoťte následující
tvrzení:
02: Považuji tuto online recenzi za... (1-7)
• Falešnou (1)
• Skutečnou (7)
03: Považuji tuto online recenzi za... (1-7)
• Nedůvěryhodnou (1)
• Důvěryhodnou (7)
04: Považuji tuto online recenzi za... (1-7)
• Klamavou (1)
• Poctivou (7)
Na základě zobrazené recenze prosím ohodnoťte následující
tvrzení:
92
QUESTIONNAIRE COPY
Po přečtení této online recenze...
05:... bych v budoucnu zvažoval(a) ubytování v tomto hotelu. (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
06: ...je pravděpodobné, že bych si zde udělal(a) rezervaci. (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
07: ...bych dal(a) tomuto hotelu šanci. (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
Do jaké míry souhlasíte s následujícími tvrzeními?
08: Recenze obsahovala konkrétní detaily (např. časy, názvy,
konkrétní prvky). (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
09: Všiml(a) jsem si prvků ověřenosti recenze (např. jméno autora
a/nebo štítek,Ověřenýpobyť). (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
010: Text působil šablonovitě a obsahoval opakující se informace
(opakované či téměř shodné formulace). (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
011: V textu se objevovala přehnaná/klišovitá tvrzení (např.
"nejlepší vůbec", /'naprosto dokonalé" apod.). (1-7)
• Zcela nesouhlasím (1)
• Zcela souhlasím (7)
93
QUESTIONNAIRE COPY
011: Kolik je 2+2?
• 2
• 4
• 6
• 8
12: Váš věk?
• 18-24
• 25-34
• 35-44
• 45-54
• 55-64
• 65+
13: Pohlaví
• Muž
• Žena
14: Nejvyšší dosažené vzdělání
• Základní
• Středoškolské
• Vyšší odborné
• Bakalářské
• Magisterské
• Vyšší
15: Kolikrát jste v posledních 12 měsících využil/a služeb ubytovacího
zařízení?
• Ani jednou
• l-2x
• 3-5x
• 6-lOx
• l l x a v í c e
• Nevím
16: Jak často si před výběrem hotelu prohlížíte recenze?
• Téměř vždy (u více než 75 % pobytů)
94
QUESTIONNAIRE COPY
• Často (u 51-75 % pobytu)
• Někdy (u 25-50 % pobytu)
• Zřídka (u méně než 25 % pobytu)
Děkuji za vyplnění dotazníku!
Vaše odpovědi byly zaznamenány anonymně a budou použily výhradně
pro účely mé diplomové práce.
Pokud máte jakýkoliv dotaz, napište mi prosím na:
• matejprachar@gmail.com
• 494081(5)mail.muni.cz
Budu vděčný, pokud dotazník přepošlete dál. Každá odpověď mi pomůže
zpřesnit výsledky. Děkuji!
Můžete dotazník "odeslat".
95
REVIEW VERSIONS
Appendix B Review versions
BO Baseline
Hotel působil celkově velmi příjemně. Pokoj byl čistý a postel pohodlná, na recepci nás
přivítal vstřícný personál. Snídaně měla dostatečný výběr a prostředí jídelny bylo klidné.
Lokalita umožňuje snadno se dostat kamkoli, a přes noc byl v hotelu klid. Pobyt
hodnotím pozitivně, vše proběhlo bez komplikací a cítil jsem se zde dobře. Za mě dobrá
volba pro kratší i delší návštěvu města. Rád bych se vrátil, pokud budu v této oblastí
znovu ubytování potřebovat.
S1 Specificity
Příjezd proběhl hladce: check-in jsem dokončila v 14:05, pokoj č. 305 (Double)
navazoval na tichou chodbu. V koupelně byl sprchový gel i fén, v pokoji rychlovarná
konvice a dvě USB zásuvky u postele. Snídaně od 7:00 s čerstvým pečivem a míchanými
vejci. Na hlavní náměstí 3 zastávky tramvají z rohu ulice; zastávka je cca 150 m od
vchodu. Postel pohodlná, pokoj udržovaný, noc klidná. Personál na recepcí ochotně
poradil s dopravou. Celkově spokojenost; ubytování odpovídalo tomu, co jsem hledala.
VI Validation
Tereza K. • Ověřený pobyt S
• • • • •
Pobyt byl příjemný a bez problémů. Pokoj působil čistě a postel byla pohodlná.
Snídaňová část nabídla několik běžných možností a prostředí bylo klidné. Personál na
recepci byl vstřícný a orientace v hotelu jednoduchá. Lokalita mi vyhovovala pro přesun
po městě a v noci byl v hotelu klid. Celkově mám z ubytování pozitivní dojem a hodnotím
ho jako dobrou volbu pro krátkou návštěvu. Pokud se do města vrátím, ráda zvážím
znovu toto místo.
9 6
REVIEW VERSIONS
01 Overlap
Skvělé místo na víkend
Hotel má skvělou lokalitu, čistý pokoj a příjemný personál. Snídaně byla super a všechno
proběhlo bez komplikací. V noci klid, postel pohodlná, recepce ochotná poradit, kam na
procházku. Pokoj působil udržovaně a prostředí celého ubytování bylo příjemné. Oceňuji
check-in a check-out bez zdržení. Celkově velmi příjemný pobyt, místo bych doporučil
známým a rád bych se vrátil při další návštěvě města. Za mě dobrá volba pro krátký i
delší pobyt.
Skvělé ubytování na víkend
Ubytování ve výborné lokalitě, pokoj čistý a personál moc příjemný. Snídaně super, bez
komplikací. Noc byla klidná, postel pohodlná a recepce milá - získal jsem tipy na
procházku. Celé prostory působí udržovaně a atmosféra je příjemná. Check-in i checkout
proběhl rychle, bez zdržování. Celkově příjemný pobyt, hotel bych doporučil dál a
rád bych se vrátil při další cestě. Za mě spolehlivá volba pro krátký i delší pobyt.
11 Implausibility
Tohle ubytování je naprostá špička - nejlepší zážitek mého života. Snídaně dokonalá,
každé sousto geniální; už nikdy nechci jíst jinde. Personál perfektní, doslova čte
myšlenky a vše vyřeší dřív, než požádáte. Postel absolutně nejpohodlnější, spánek jako
na obláčku, ráno jsem se probudil jako nový člověk. Pokoj bez jediné vady, zářivě čistý a
vyladěný do posledního detailu. Po téhle zkušenosti už nemá smysl zkoušet cokoliv
jiného - tady je stoprocentní jistota. Doporučuji úplně všem, bez výjimky.
9 7
GENERATIVE A I PROMPTS
Appendix C Generative AI prompts
For condition Baseline (BO):
Napiš neutrálně pozitivní recenzi na hotelový pobyt v češtině o délce
přibližně 80-100 slov. Recenze má působit jako běžné
pětihvězdičkové hodnocení bez specifických detailů, bez ověřovacích
prvků a bez výrazných emocí. Text by měl být realistický, obecný a
popisovat standardní aspekty pobytu, jako je čistota pokoje, pohodlí,
snídaně, personál a celkový dojem. Nepřidávej žádné konkrétní názvy
hotelů ani přehnané či podezřelé formulace. Recenze má sloužit jako
baseline verze bez jakýchkoli výrazných obsahových či stylistických
signálů.
For remaining conditions:
Na základě baseline recenze vytvoř čtyři nové verze pětihvězdičkové
hotelové recenze v češtině, přičemž každá bude obsahovat jeden
specifický stylistický prvek. Délka každé recenze by měla být
přibližně 80-100 slov. Zachovej podobný tón a strukturu jako u
baseline verze, ale uprav obsah tak, aby odpovídal příslušné
stylistické manipulaci.
Použij tyto čtyři varianty:
Specificity (S1) - Reviews that are very specific in detail.
Validation (V1) - Reviews where the reviewer is identified by his/her
real name, along with a form of "Verified buyer" badge.
Overlap (01) - Multiple reviews in a set of reviews that sound or
appear similar to each other.
Implausibility (11) - Reviews that use excessive cliches (such as, "This
is the last x you will ever have to buy!")
9 8
9 9