The importance of behavioral data to identify online fake reviews for tourism businesses: a systematic review

Introduction

In the last four decades, the continuously growing sector of tourism has been supported by the development of information and communication technologies (ICT) (Papathanassis & Buhalis, 2007; Buhalis & Law, 2008). In the 21st century, the digital revolution in social sciences and tourism should be taken into account, as it is one of the important factors that make the tourism industry competitive (Moutinho, Ballantyne & Rate, 2011; Saura & Bennett, 2019).

Nowadays, consumers use different social platforms, such as social networking sites (SNS), consumer review sites, blogs, and social communities in order to communicate and share their purchase experiences and behavior regarding for products and brands with other consumers (Cheung & Thadani, 2012; Chew, Metheney & Teague, 2017; Hubert et al., 2017).

The continuously developing technologies and the wide spread use of the Internet in several industries have empowered the evolution from traditional word-of-mouth to electronic word-of-mouth (eWOM) (Gottschalk & Mafael, 2017; Manes & Tchetchik, 2018). eWOM is embodied in online reviews that customers write for other customers. The content of online reviews depends on the experience that these specific customers have with purchased products or services (Munar & Jacobsen, 2014; Saura, Reyes-Menendez & Alvarez-Alonso, 2018a). This fact has an important consequence for businesses, as there is a power shift from companies to consumers (Hennig-Thurau, Walsh & Walsh, 2003; Reyes-Menendez et al., 2018). The abovementioned power shift is particularly important in certain industries, such as the tourism industry where customers pay close attention to the opinion of previous travelers (Papathanassis & Knolle, 2011). For that reason, online reviews are a powerful communication tool for tourism businesses. Tourism businesses are companies such as hotels, ancillary services, transportation companies and restaurants (Riegner, 2007; Reyes-Menendez, Saura & Palos-Sánchez, 2018b).

With the growth of the Internet, the number of online reviews has increased as well, exerting a significant influence on customers’ purchase decision making (Bennett, Yábar & Saura, 2017). The growing relevance of this type of communication is particularly important on social platforms where it takes the form of online reviews. However, what happens when this information does not represent the objective reality? Are companies, rather than real consumers, writing these reviews?

In 2019, the Federal Trade Commission (FTC) denounced, for the first time, a company that advertised products to lose weight on Amazon for writing false reviews on this platform. In a press interview, the director of FTC, Andrew Smith, said that false reviews adversely affect both consumers and companies, as they represent a breach of market norms. For their part, Amazon representatives declared that they would take legal action against those fake reviews and invest significant economic and human resources to ensure that the reviews of the products presented on their platform are true and up-to-date (Shu, 2019).

Overall, consumers are becoming increasingly aware that many of the reviews on social network sites are fraudulent. To show this trend in numbers, following Önder & Ulrich Gunter (2016) we first made a search with Google Trends, a tool used previously (Önder, Irem & Ulrich Gunter, 2016) to identify past search trends in Google on various topics of interest. The results are shown in Figs. 1 and 2.

In Fig. 1, the searches on “fake reviews” made by users on Google are shown with a solid line, while searches on “fake news” are shown with a dotted line. As can be seen in Fig. 1, the number of searches on both topics has increased from 2004 to 2018. However, the increase is dramatically more pronounced for the searches related to “fake news”. The dynamics of the growth in the number of searches on “fake reviews” only is shown in Fig. 2. As can be seen in Fig. 2, the number of searches on “fake reviews” has steadily increased throughout the period 2004–2018, and this number continues to grow.

In the next step, we checked the importance of the topic of fake reviews for the scientific community. This was done with a search in Web of Science (WOS), a scientific database that indexes scientific articles.

Similarly to the results obtained with Google Trends, the WOS findings suggest that, throughout 2004–2018, there has been a considerable growth in the number of articles published on the issues of fake news and fake online reviews. Furthermore, again consistent with the results of using Google Trends, the scientific community has been more interested in the topic of fake news, as we found 248 papers that include the term “fake news” in the title, while there were only 48 papers with the term “fake reviews” in the title.

Figure 3 shows the dynamics in the number of citations to articles addressing fake online reviews throughout the period 2004–2018. As can be seen in Fig. 3, the first citations appeared in 2013. However, it was not until 2014 that scientific interest in fake online reviews skyrocketed, and it continues to grow today. Of note, according to the publication terms of the journals included in the JCR (Journal Citation Report) Index to which Web of Science (WOS) belongs, the publications that appeared in 2018 will begin to get cited in the next months or years. This explains why the publications from 2018 do not follow the growth trend as compared to the previous years.

The results shown in Figs. 1–3 underscore the importance of the topic of fake online reviews for consumers, companies, and the scientific community. Accordingly, in the present paper, our major goal is to identify directions of current research to address the problem of fake reviews on tourism platforms.

The remainder of this paper is structured as follows. After a brief literature review in ‘Literature Review’, we present the methodology used in the present study in ‘Methodology’. Results are reported in ‘Exploratory analysis of results’. The paper concludes with a discussion of implications of our findings (‘Implications’) and general conclusions (‘Conclusions’).

Literature Review

Over the last years, many studies have investigated the impact of online reviews on consumer purchase behavior and decision making (Chevalier & Mayzlin, 2003; Riegner, 2007; Gretzel & Yoo, 2008). A strong influence of online reviews has also been highlighted in numerous industrial statistic reports (e.g., Reyes-Menendez et al., 2018).

Electronic word-of-mouth has become an important concept for tourism businesses (Papathanassis & Knolle, 2011). According to Litvin, Goldsmith & Pan (2008) and Pai et al. (2013), eWOM is the most important source of information that drives consumer purchase behavior in the hospitality and tourism services sectors. In the last several decades, advances in information and communication technologies (ICTs) have transformed both travelers’ behavior and the tourism industry (Buhalis & Law, 2008). Nowadays, the number of travelers who access the Internet to book hotel rooms via third-party intermediaries is continuously increasing (Luo, Chen & Zheng, 2016). Furthermore, several studies demonstrated that about two-thirds of customers prefer to read online consumer reviews about a hotel, rather rely on the hotel’s own descriptions. Such online reviews are visited by hundreds of millions of potential hotel visitors every year (Reyes-Menendez, Saura & Martinez-Navalon, 2019)

Therefore, in order to obtain a better understanding of the continuously increasing impact of eWOM on different social platforms and its effect on the decision making and behavior of hotel consumers, reviews on online travel sites and social networking sites should be taken into account (Saura, Rodriguez Herráez & Reyes-Menendez, 2019). Yet, a recently emerging issue with online reviews is that some online reviews are fake. Although most online platforms have their own false review detection algorithms (Cheng, Tseng & Chung, 2017), these algorithms are sometimes limited in scope and filter only 16% of published fake reviews (Luca & Zervas, 2016). Therefore, there is a clear need to improve the existing algorithms and elaborate new approaches. Many studies have sought to do just that (e.g., Elmurngi & Gherbi, 2017; Munzel, 2016; Zhang et al., 2016). To this end, various methodologies have been used, some of which will be discussed in the remainder of this paper.

Methodology

Following Kazakov & Predvoditeleva (2015), in the present study, we aimed to provide an overview of previous research on the state of art of online fake reviews in tourism social networking sites. We focused on the analysis of users’ ability to detect real or fake reviews. To this end, we critically examined the available literature on tourism fake reviews and behavioral approaches to analyze and identify them for tourism businesses. The systematic literature review focused on the following two main topics: (i) fake reviews; (ii) tourism. Following Sherman et al. (2005) and Banerjee, Chua & Kim (2015), we used a randomized controlled process to select the main topics and consequent search terms “fake reviews” and “tourism”.

The scientific databases of Scopus, PubMed, PsyINFO, ScienceDirect, and Web of Science were used to collect relevant studies on the issue at stake. Of note, when performing the search by “Title” in the scientific database Web of Science, only one article met the aforementioned search requirement, that both “fake reviews AND tourism” were contained in the title. Therefore, following Saura, Palos-Sánchez & Suárez (2017), we included the articles that were initially obtained as a result of the search, prioritizing those that dealt with reviews, even if they were not specifically focused on tourism platforms. We reasoned that the insights reported in these studies could be extended to address the problem of false reviews on tourism platforms. The search yielded a total of 124 articles; after different filters were applied (see Fig. 4), a total of 17 studies were selected for further analysis. The Boolean operator AND was applied to optimize the results. All articles were analyzed by reading the titles and abstracts and selecting the ones which met the inclusion criteria. Next, we analyzed the 17 selected papers. The data were collected in June 2018 using AMSTAR (2017), a tool initially designed to assess the quality of articles based on their abstracts (Shea et al., 2009). In this way, we ensured that only high-quality studies were included in the dataset. In the process of article selection, we also followed the recommendations formulated by Van den Bosch & Ode Sang (2017). These recommendations include keyword search in several databases, predefined inclusion criteria, and data extraction based on selected keywords.

To this end, following Saura, Palos-Sánchez & Suárez (2017), we used the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2009 Flow Diagram. This method, introduced by Moher et al. (2009), provides guidelines to develop Systematic Reviews and Meta-Analyses that include conceptual and practical advances in the science of systematic reviews. One of the phases of the PRISMA Flow Diagram is discarding the articles that have inadequate or inconclusive terms. The terms considered as inadequate or inconclusive are those that a priori may correspond to the keywords; however, when reading the article in depth, it is observed that they are not within the scope of the investigation. These terms can be misleading, as in the case of reviews that may be either tourist reviews or peer reviews.

Our aim was to achieve the highest possible amount of evidence in the results based on high-quality studies. Some of the variables used in AMSTAR to evaluate the quality of the systematic review were (i) the relationship of the research question to the criteria included in the study; (ii) the extraction of data from at least two independent researchers; (iii) the quality of the literature review, (iii) identification and definition of concepts; and (iv) the quality of the conclusions stated in the study.

Exploratory Analysis of Results

The Systematic Literature Review (RSL) was proposed by Saura, Palos-Sánchez & Suárez (2017) and Bassett (2015) as a development tool to carry out an exploratory analysis of previously reported results. Such a literature review is used to evaluate researchers’ interest in a specific topic (Saura, Reyes-Menendez & Palos-Sanchez, 2018b). A literature review is an exploratory methodology and consists of collecting and re-analyzing available findings. A literature review usually includes both primary and secondary sources.

Luo & Zhong (2015) and Comerio & Strozzi (2018) conducted literature reviews and applied exploratory analysis specifically for tourism businesses, while Huete-Alcocer (2017) focused on the transformation of the traditional word-of-mouth into electronic word-of-mouth, and the behavioral implications of this transition for the tourism industry.

A summary of the studies selected for further analysis in the present review is shown in Table 1. Table 1 presents the authors of relevant studies, as well as the main contents of those studies. Based on this information, we categorized the reviewed studies in several groups. For instance, the studies that can be applied to all sectors were categorized as “All Industries”. A study classified into the category of “E-Commerce” was included in the analysis since, for example, TripAdvisor is considered an E-commerce platform in the tourism sector. Studies on restaurants were categorized into the “Restaurant—Hospitality Industry” group. Finally, articles on hotels and other tourist services were grouped into the “Tourism” category.

The studies summarized in Table 1 demonstrate the growing interest in the concept of fake reviews and social networking sites, particularly in the hospitality and tourism industries. Some of these studies (e.g., Banerjee & Chua, 2014; Banerjee et al., 2015; Cardoso, Silva & Almeida, 2018; Chang et al., 2015; Deng & Chen, 2014; Hunt, 2015; Lappas, Sabnis & Valkanas, 2016a; Lappas, Sabnis & Valkanas, 2016b; Li, Feng & Zhang, 2016; Munzel, 2016) focus on the Tourism industry category, while others fall into the hospitality industry category (Chen, Guo & Deng, 2014; Li et al., 2014; Li et al., 2018; Luca & Zervas, 2016). Some works (Lin et al., 2014; Zhang et al., 2016; Ramalingam & Chinnaiah, 2018) were included as part of the analysis because their results can be implemented in every industry that allows consumers to write reviews, including the tourism industry. Elmurngi & Gherbi (2018) analyzed false reviews in E-commerce, considering that TripAdvisor is the most important e-commerce platform in the hospitality industry; therefore, this study might be of interest to the present study (Reyes-Menendez, Saura & Alvarez-Alonso, 2018a).

The interest in the concept of fake reviews is underpinned by two factors. First, as demonstrated in several studies, the currently available algorithms of false review detection remain largely ineffective. For instance, Luca & Zervas’ (2016) results demonstrate that only 16% of the false reviews are filtered on the Yelp platform—particularly, those that have more extreme content, either positive or negative; this suggests that the remaining 84% of reviews are not filtered and may be false. Second, as demonstrated in several studies, false reviews negatively impact companies’ visibility. For instance, Lappas, Sabnis & Valkanas (2016a) and Lappas, Sabnis & Valkanas (2016b) report that, with only 50 false reviews in some markets, competitors can be overtaken in terms of visibility. In this sense, Hunt (2015) focuses on the example of the UK Advertising Standards Authority that found against the tourism social platform TripAdvisor. These figures explain the concern of establishments in the tourism sector about the phenomenon of false reviews.

In what follows, we review the methodologies used in previous research on false reviews in the tourism industry. Particular attention is paid to the unit of analysis focused on in previous studies and the behavioral approach to the analysis and identification of online fake reviews in tourism.

Implications

Conclusions

This exploratory study has defined the scope and identified recent avenues of research on fake online reviews in the tourism industry. Interestingly, even though consumers might be aware that some comments are fictitious, they still rely on them to make decisions (Manes & Tchetchik, 2018).

However, although many studies have investigated the impact of eWOM in the tourism industry (Mauri & Minazzi, 2013; Vermeulen & Seegers, 2009; Xie et al., 2011), further research on the impact of new approaches is necessary to detect fake online reviews for tourism businesses.

As suggested by our literature review, to counteract this trend, tourism companies must constantly improve their methods of detecting false reviews. These methods are mainly based on algorithms (Banerjee & Chua, 2014; Banerjee et al., 2015), and the improvements are especially based on behavioral approaches (Li, Feng & Zhang, 2016; Zhang et al., 2016). In addition, in order not to lose their visitors’ trust (Ladhari & Michaud, 2015), platforms that allow users to write reviews about their experiences with tourism companies take legal action against false reviews (Hunt, 2015). When a company spots fake reviews, it has the right to take action. TripAdvisor has fought fraudulent reviews in thousands of lawsuits.

The systematic review of the literature undertaken in the present study allows us to make the following three conclusions.

First, our results demonstrate that there is an ever-growing interest, among scientific and consumer communities alike, in the credibility of online reviews in the tourism sector with an interdisciplinary approach. Not only have Computer Science and Information Systems (e.g., Chen, Guo & Deng, 2014; Chang et al., 2015; Li et al., 2014; Zhang et al., 2016) published research about fake reviews but also Management, Business and Operation Research and Management Science (e.g., (Lappas, Sabnis & Valkanas, 2016a; Lappas, Sabnis & Valkanas, 2016b; Luca & Zervas, 2016); (Munzel, 2016). Consumer concern about fake reviews is particularly pronounced (see Figs. 1 and 2).

Second, the results of both scientometric analysis and our literature review suggest that the methods of Computer Science are most frequently used in terms of the development and improvement of the methods to detect false reviews. Most publications included in the literature review (14) fall into the category of algorithms (eight) and sentiment analysis (six), while only a few of them (three) use other methods.

Third, our systematic review of the literature highlights the importance of further development of new methods for identifying false reviews based on the following criteria: (1) those based on emotional approaches; (2) those based on semantic analysis; and (3) those based on new technologies.

Some of these methods are based on a behavioral and emotional analysis of the text of reviews in the tourism sector (Banerjee & Chua, 2014; Li et al., 2018; Luca & Zervas, 2016; Munzel, 2016; Zhang et al., 2016). To this end, sentiment analysis and textual analysis were used (Chen, Guo & Deng, 2014; Deng & Chen, 2014; Elmurngi & Gherbi, 2017; Lappas, Sabnis & Valkanas, 2016a; Lappas, Sabnis & Valkanas, 2016b; Lin et al., 2014; Zhang et al., 2016).

Other methods are focused on the semantic analysis of the reviews (Li, Feng & Zhang, 2016; Luca & Zervas, 2016) and provide guidelines to follow in order to detect fraudulent reviews. Some of these clues are based on extreme comments and ratings, as well as on comments lacking detail or reviews where first-person pronouns (e.g., I, me) are widely used to simulate sincerity.

Finally, some methods to improve the detection of fake reviews are mostly supported by new technologies (Li et al., 2014) and present novel solutions for fake review detection. One of these technologies is the Blockchain, which requires proof of payment in order to publish a review.

The limitations of this study relate to the number of studies reviewed and the databases consulted. Although the authors consulted the main scientific databases—Scopus, PubMed, PsyINFO, ScienceDirect and Web of Science—there are more databases available for consultation.

While our review of the literature has highlighted several important issues related to fake reviews in the tourism sector, further research that would perform in-depth analysis of specific aspects presented in this paper is needed. Among such possibilities is using quantitative techniques to measure the impact of fake reviews on social networking sites. Another promising area of future research is studying the behavioral aspects of users who write online reviews for tourism businesses.

Supplemental Information