Why do we trust in online reviews? Integrative literature review and future research directions

Introduction

Online reviews of such diverse “objects” as cameras, hotels, physicians, and university lecturers have become a mass phenomenon. Many internet users search through reviews of peer consumers, peer patients, peer students, etc. before taking decisions so that online reviews have gained a considerable impact in many areas of everyday lives (e.g., Chevalier & Mayzlin, 2006; Maslowska et al., 2017). Basing decisions on online reviews requires the trust of opinion seekers in the evaluations of their peers. Accordingly, trust in and credibility of online reviews are of specific importance for explaining the effects of reviews. However, trusting in online reviewers and their reviews is a high-risk undertaking. For example, the opinion seeker usually lacks information on both the reviewer’s motives to provide the review and the reviewer’s qualification to evaluate the reviewed object. Furthermore, the providers of the reviewed objects benefit from positive evaluations of their offerings and thus have strong incentives to influence reviews in their favor (Lappas, 2012). Yet despite these obstacles, surveys indicate that internet users widely trust online reviews: For example, an US industry survey (BrightLocal, 2022) found that 49 percent of consumers trust online reviews as much as they trust personal recommendations from family and friends.

Studying trust has become a main concern in research on online reviews. As this review article will show, I identified 70 research articles published in peer-reviewed journals that examined trust in online review contexts. However, a focused overview of their findings is still missing. There exist various literature reviews (Bore et al., 2017; C. M. K. Cheung & Thandani, 2010; Ismagilova et al., 2020; King et al., 2014; Rani & Shivaprasad, 2018) and conceptual frameworks (Moran & Muzellec, 2017) that address research on electronic word-of-mouth (eWOM) communication. These works are, however, neither focused exclusively on online reviews nor on trust and credibility. To close this gap, this article sets out to provide an integrative literature review of empirical studies on trust in online review contexts. It contributes to the field by collecting existing findings, offering a framework for organizing these findings, and identifying shortcomings and research desiderata to indicate future research directions. 

Theorizing Trust in Online Review Contexts

In this section, I will put forward a suggestion on how to theorize trust in online review contexts. To do so, I will first introduce my theoretical understandings of online reviews and trust. Then, I will bring the two understandings together to develop a framework for organizing the empirical findings in the field.

Online Reviews

Online reviews are usually discussed as a specific format of eWOM communication. eWOM is defined as “consumer-generated, consumption-related communication that employs digital tools and is directed primarily to other consumers“ (Babić Rosario et al., 2020). eWOM thus serves as an umbrella concept that includes not only online reviews, but also other types of computer-mediated peer-consumer conversations. In contrast to other eWOM formats, online reviews are usually posted on specific online review platforms. These platforms can be both integrated within retailer homepages (e.g., Amazon, Bookings), fan communities (e.g., Metalstorm, The Metal Archives), and social networking sites (e.g., Facebook), as well as created independently (e.g., Yelp, HealthGrades). Online reviews have been defined as “peer-generated product evaluations posted on company or third-party websites” (Mudambi & Schuff, 2008, p. 186). For the current study, I heavily draw on this definition but specify that I consider all studies relevant that examine peer evaluations in the review section of a review platform. These reviews may be posted on an actual platform or generated specifically for a scientific study. It should be noted that this approach leads to an exclusion of online reviews published on individual blogs, e.g., by social media influencers. I decided to exclude these reviews because the conditions under which trust in online reviews emerges on review platforms differs markedly from the conditions under which it emerges on blogs.

Trust

Trust is a common social phenomenon that can be observed in many, if not all, areas of everyday life. Yet, when trying to pinpoint trust, the fuzziness of the concept becomes apparent. Many researchers have tackled trust from the perspective of their respective fields and presented a wide range of conceptualizations (see Gefen et al., 2003, for an overview). However, many scholars agree that in a trust relationship, a trustor acts on the grounds of the expectation that a trustee acts in a specific way, although the trustee could also act differently (Barber, 1983; Gefen et al., 2003; Giddens, 1990; Hardin, 1992; Luhmann, 1968, 1975/2018; Möllering, 2001). For example, a consumer (aka the trustor) who reads an online review of a camera might expect that the reviewer (aka the trustee) has collected sufficient information about and gained extensive experience of the camera before publishing a review. The tricky point here is that the trustor cannot be sure whether the trustee actually acts in the expected way. This uncertainty makes trust risky because the trustee could always act differently than expected, but the trustor will find out whether the trustee fulfilled their expectations only after having trusted them (Gefen et al., 2003; Luhmann, 1968, 1975/2018). For example, a consumer buys the particular camera just to find out that the reviewer did not discuss relevant dysfunctionalities and weaknesses. Alternatively, some authors highlight the vulnerability of the trustor as a key characteristic of trust (e.g,. McGeer & Pettit, 2017; Mishra, 1996; Rousseau et al., 1998) and thus emphasize the consequences of entering a relationship whose outcomes are uncertain because they depend on the trustee.

For this review, I chose to draw on a concept that was, at its core, developed by Luhmann (1968, 1975/2018). Luhmann shares the definition of trust relationships that I just introduced. Imperative for the Luhmannian understanding of trust then is the notion of “own selectivity” (Luhmann, 1990). The concept of own selectivity starts from the observation that, in most situations, a person faces more than one option how to act and therefore must make a decision. It highlights that only this person can make the decision. In a next analytical step, Luhmann introduces a second person to the situation to highlight the social consequences of own selectivity. In their actions, the second person depends on the first person and therefore has to find ways to cope with the first person’s own selectivity as the general autonomy to decide what to do. For Luhmann, trust is a mechanism that helps the second person (the trustor) to do so by acting on the grounds of the assumption that the first person (the trustee) will act as expected. Other such mechanisms are, for example, familiarity, contracts, and hope (Kohring, 2004). A consequence of this understanding of trust is that trust is always addressed to another social actor such as online reviewers and not to entities without own selectivity such as online reviews. This position also explains why, in trying to be analytically rigorous, I use the somewhat cumbersome formulation “trust in online review contexts” instead of “trust in online reviews.”

Since trusting is a risky business, trustors try to identify reasons to trust (Kohring, 2004). Such reasons serve as legitimation for entering a trust relationship because they reduce the perceived risk. For instance, in the camera example, the educational background of the reviewer or the conformity of a review with other reviews on the same camera may function as reasons to trust. Reasons to trust usually refer to particular dimensions of trust (Kohring, 2004). While trust is regarded as a multidimensional concept (Gefen, 2002; Lewis & Weigert, 1985; Romero & Mitchell, 2017), there exists less consensus about which dimensions constitute trust. Some authors argue in support of general principles that address the risk and vulnerability of the trustor such as benevolence (McKnight & Chervany, 2001; Romero & Mitchell, 2017), honesty (Fukuyama, 1995; Seckler et al., 2015), or reliability (Mishra, 1996; Rotter, 1971). In contrast, Luhmann (1968, 1975/2018) argues that the dimensions of trust depend on the social context of the trust relationship. That is to say that the dimensions of trust will be different in different social contexts. For example, the dimensions of trust in a writer of online reviews differs from those of trust in a judge or in a politician because online review author, judge, and politician fulfill different functions in society. In this review, I follow this second line of reasoning.

A considerable debate in trust research circles around the question whether trust should be considered a belief, an attitude, an intention, or a combination of these (for an overview, see Gefen et al., 2003). The Luhmannian line of understanding trust brings up yet another possibility: From its perspective, trust can be regarded as a social relation between two (or more) persons. Accordingly, it holds that trust emerges in and through the relationship rather than “residing” in the trustor. This perspective allows to adjourn the debate on belief, attitude, and intention without neglecting its relevance.

Luhmann’s concept of trust has proven to be productive for the study of trust in mediated communication (e.g., Graf, 2018; Hoffjann, 2013; Kohring, 2004). Like any other concept, however, Luhmann’s concept opens specific perspectives while, at the same time, suffering from its blind spots. This is why I want to emphasize that there exist other trust concepts (e.g., Barber, 1983; Giddens, 1990; Lewis & Weigert, 1985; McGeer & Pettit, 2017; Möllering, 2001) that are worth exploring and that will open different perspectives on trust in online review contexts. Beyond the question of which particular concept to adopt, I suggest that engaging with more comprehensive concepts of trust is fruitful because it sharpens the analytical capabilities of research on trust in online review contexts.

Trust in Online Review Contexts

As Duffy (2017) pointed out, there are different forms of trust in online review contexts. To identify relevant trust relationships, I employed a model of the ideal online review process (Figure 1). The model identifies the various actors involved in the review process: An option provider (e.g., camera manufacturer or physician) offers a specific option (e.g., camera that can be bought or health treatment that can be received). A reviewer, who, ideally, has experience with using this option, composes a review to share this experience. The review is (usually) published on a platform, provided by a platform provider (e.g., Amazon or Healthgrades). A community of platform users assesses the quality of the review (e.g., via comments, flagging, helpfulness votes). An opinion seeker who is interested in the option consults the review as part of their decision making on whether to select the option. In addition to reading the review, the opinion seeker might collect further information provided by an external actor, i.e., an actor who is neither associated with the platform nor the option. Note also that the model only captures the most basic operation of the online review process, which is one opinion seeker consulting one review. In most cases, the opinion seeker will consult more than one review and compare the reviews with each other.

Figure 1. Model of the Ideal Online Review Process.

Note. White boxes symbolize actors within the process. Gray boxes symbolize cultural artifacts. Arrows symbolize relations between actors and artifacts. Dotted lines frame the core of the online review process. Note that the model indicates relations only for the ideal case in which all actors act in a way that facilitates the best online review quality.

From the perspective of the opinion seeker, the involvement of other actors poses risks due to each actor’s own selectivity: The opinion seeker cannot know whether these actors act in a way that ensures good review quality and thus might disappoint expectations directed toward them. For example, platform providers might sell specific editing options to option providers, option providers might pay reviewers to write tendentious reviews (Gössling et al., 2018), reviewers might be unconsciously biased toward premium brands (de Langhe et al., 2016), and community members might have different views on what constitutes a helpful review. Because of these risks, the opinion seeker needs to trust in one or more of the involved actors if they consult online reviews as a source of information in the decision-making process. This conclusion implies that trust in online review contexts usually goes beyond the sole trust in the reviewer to also involve other actors as trustees. The model of the online review process provides a framework for organizing research on trust in online reviews. It allows for attribution of the examined factors to specific actors in the online review process, by indicating to whose actor’s selective actions the factor refers.

Considering the various actors also helps to reflect on four distinctive features of trust in online review contexts that, in combination, distinguish this setting from offline settings. First, while pooling recommendations in offline settings usually includes familiar trustees such as friends, family members, and colleagues, trust in online review contexts is directed at anonymous strangers such as reviewers and platform communities (Borchers, 2021). This mechanism allows opinion seekers to benefit from the experiences of other internet users beyond local, temporal, and social constraints, yet it comes at the price of drastically reduced familiarity with the trustees. Second, review platforms allow their users to share experiences and thus facilitate trust between strangers. However, they provide not only the technical infrastructure for the online review process, but also the terms and conditions for publishing and accessing online reviews (van Dijck, 2013). These terms and conditions reflect the commercial interests of platform providers. Third, quality management is delegated to the community and usually takes place only after the publication. This is different from other (offline) sources to which opinion seekers could turn such as travel books when looking for accommodation or consumer safety groups when looking for a new smoothie maker. Fourth, the online review process is expected to be a peer-to-peer communication process (Borchers, 2021). Peer-to-peer communication implies that the roles of reviewer and opinion seeker are generally interchangeable. Consequently, all opinion seekers could also act as reviewers and vice versa. Such role flexibility does not exist in most other social contexts, such as traditional advertising or journalism.

Research Questions

Trust in online reviews has emerged as an eminent topic in the research on eWOM. This review article aims at providing an overview of the state of research. First, it addresses the research designs that researchers apply. Research designs determine what researchers can see. For example, standardized surveys make exactly those attitudes, intentions, behaviors etc. visible that the surveys ask for. In contrast, observations do not predefine what is to see and thus expand the perspective of the researcher, yet they usually fail to see the broader picture beyond individual cases because they focus on only a few participants.

RQ1: Which research designs are applied in studies on trust in online review contexts?

Second, this review examines the theoretical conceptualizations that inform studies on trust in online review contexts. Like methods, theoretical conceptualizations allow researchers to see specific aspects of a phenomenon because they direct the view of the researcher.

RQ2: How do studies on trust in online review contexts conceptualize trust?

While the RQ1 and RQ2 provide information on how researchers produced their findings, the third research question aims at the actual findings and accounts for the factors that explain trust in online review contexts.

RQ3: Which factors have which effects on trust in online review contexts?

Method

Methodological Framework: Integrative Literature Review

To answer the research questions, I conducted an integrative literature review (Whittemore & Knafl, 2005). The integrative literature review “reviews, critiques, and synthesizes representative literature on a topic in an integrated way” (Torraco, 2005, p. 356). It aims at generating a summary of research trends as well as new perspectives and frameworks on the reviewed topic (Khoo et al., 2011; Torraco, 2005). The integrative literature review supports my objectives in going beyond the description of existing research and allowing for the application of a new conceptual framework and the development of future research directions.

Data Collection

An appropriate and comprehensive literature search strategy is important for enhancing the rigor of literature reviews. The strategy should allow identification of all relevant articles to ensure that the review is based on an adequate corpus and can yield accurate results (Whittemore & Knafl, 2005).

Databases

Online reviews are studied in different disciplines, such as marketing, communication, psychology, information systems, and tourism. To take the diversity of the research field into account, I considered a variety of established international academic databases: (1) Business Source Premier, (2) Communication & Mass Media Complete, (3) PsycARTICLES, (4) PsycINFO, (5) PSYNDEX, (6) Library Information Science and Technology Abstracts, and (7) Web of Science.

Search Term

The research literature discusses online reviews under various labels. I therefore included alternative labels in the search term by varying (a) the term “online” with “internet,” “digital,” and, as a large share of existing studies is interested in consumer behavior, “consumer;” and (b) the term “review” with “rating” and “recommendation.” In addition, I searched for the term “eWOM” or “electronic word-of-mouth” (in different variants) because eWOM is an umbrella concept that includes online reviews. Furthermore, research on trust relies on two theoretical concepts, trust and credibility (Kohring & Matthes, 2007). As the two terms are closely connected, I decided to include both in the search term. These considerations resulted in the search term displayed in Table 1.

Table 1. Final Search Term for Integrative Literature Review.

Selection Criteria

I included an article in the corpus of this review if it met the following content criteria: (1) The article explains the perception of trust (or credibility, respectively) in an online review or the reviewer (see section 2.1), i.e., it conceptualizes trust in online review contexts as a dependent or mediating variable. Articles that focused on trust in other social actors, such as brands or platforms, were not included. I also did not consider articles that examined trust exclusively as an independent or moderating variable. (2) The article focuses on online reviews. Articles on other eWOM formats, including blog posts, forum postings, or social media commentaries, were excluded. Articles were also excluded if they studied online reviews posted outside of review platforms. If in doubt whether an article studied online reviews, for example if the article addressed eWOM and did not specify the eWOM format, I excluded it to ensure a clean data set. (3) The article presents the results of an empirical study.

Furthermore, I introduced three formal criteria: (4) The article is published in a peer-reviewed scientific journal. (5) The article is published in English. (6) The article is published before January 1, 2021, the cutoff date of the query.

Article Selection

The full text query yielded 1,418 records in the seven databases. The combined Ebsco databases (databases 1–6) produced 626 hits, and Web of Science produced 792 hits. As the first step, I removed all articles that did not meet the formal criteria. 510 records were published in other formats than in peer-reviewed scientific journals, another 22 records were published in languages other than English, while all remaining articles were published before the cutoff date of the query. Applying the formal criteria thus led to the exclusion of 532 records. I then controlled the remaining 886 records for duplicates. This procedure identified 275 duplicates, reducing the number of records to 611. As the next step, I used the content criteria to decide whether the article should be included in the sample. To do so, I reviewed the title, keywords, and abstract. If this information did not suffice to make a decision, I examined the full text of the article. This procedure yielded to the exclusion of another 541 records. The final sample thus comprised 70 articles (see Appendix A). Figure 2 displays the selection process.

Data Analysis

For data analysis, I followed Whittemore and Knafl’s (2005) instructions for integrative literature reviews. They propose the following procedure: (1) Data reduction: I coded the articles for concepts of trust, theories, methods, examined factors, and results. (2) Data display: I synthesized the data from the individual articles and organized it into subgroups. This process was guided by Kohring’s (2004) concept of trust and the online review process model. I put particular attention to the identification of different factors discussed under the same label and similar factors discussed under different labels. (3) Data comparison: I examined the synthesized findings to identify strengths, shortcomings, and desiderata of the field. I placed emphasis on a critical analysis as described by Torraco (2005). (4) Conclusion drawing: I critically assessed the current state of research to develop future research directions.

Figure 2. Overview of Article Selection Process.

Note. Adapted from Page et al., 2020 (modified).

Results

Corpus Characteristics and Bibliometrics

I identified 70 articles on trust in online review contexts that met the selection criteria. The first article on the topic was published in 2005, but more profound scholarly interest in the topic started growing only in 2011 (see Figure 3). Articles were published in 43 individual journals (see Table 2). The large number of journals indicates a high fragmentation of the research field. Only five journals published more than two articles on the topic.

Figure 3. Number of Publications per Year of Articles on Trust in Online Review Contexts.

Note. The database search for this literature review was conducted in early 2021 and covers research activities before January 1, 2021. The database search yielded two articles (Bartosiak, 2021; Zhang et al., 2021) that were registered as online first versions for 2020. These articles have been published in a journal issue in the meantime and are thus displayed as 2021 in this figure.

Table 2. Journals That Have Published Articles on Trust in Online Review Contexts.

Research Designs

Research on trust in online reviews applies several methods (see Table 3). Experiments are the most employed method by far, followed by standardized surveys. Accordingly, there is a predominance of articles that adopt quantitative approaches (64 articles) over articles that adopt qualitative (2 articles) or mixed quantitative-qualitative designs (4 articles). Using student samples is a common procedure in the field (30 studies), although most studies rely on general population samples (49 studies). The samples demonstrate geographical diversity (see Table 4). Most studied review objects are hotels (21 studies), electronics (e.g., cameras, TV sets; 16 studies), and restaurants (15 studies). Many studies, especially surveys, do not specify a review platform, but adopt platform-independent approaches, while some focus on existing platforms. Platforms studied most often are TripAdvisor (12 studies), Amazon (6 studies), and Yelp (5 studies), which clearly indicates a Western bias.

Table 3. Methods Applied in Studies on Trust in Online Review Contexts.

 Table 4. Sample Origins in Studies on Trust in Online Review Contexts.

Conceptualizing and Measuring Trust

Few studies are informed by more comprehensive concepts of trust. Although it is admittedly hard to determine what is a “more comprehensive concept” and when such a concept “informs” a study, I identified eight articles that related to a conceptual work on trust at least on the level of a definition of trust and not only in passing. The works by Mayer et al. (1995; 6 references) and McKnight and Chervany (2001; 5 references) found some resonance in the field. This resonance might result from the fact that the first and widely cited study in the field (Smith et al., 2005) is informed by these works.

Trust is more often conceptualized on the level of its operationalization for empirical inquiries. I therefore examined reported scales and items that were used to measure trust. I identified a total number of 46 scales that have been used in the field. 38 scales draw on items that originate from previous research, i.e., studies either adopting a complete scale or combining different scales, while eight studies used their own scales. Although the scale presented by Ohanian (1990, 1991) was used much more frequently than others (informing 15 articles), the variety of scales again suggests a great heterogeneity of the field. Qualitative studies used wordings that are similar to the wordings of items in quantitative studies. Where articles reported questions from interview guides, there was a tendency to ask respondents for trust or trustworthiness directly rather than breaking the concept down to its various dimensions.

Examined Factors and Their Effect on Trust

Research has examined the impact of 77 factors on the emergence of trust in online review contexts. I used the model of the online review process to systemize these factors by organizing them according to the actors upon whose own selectivity they touch. For most factors, this process should be self-evident. For example, I assigned the experience of the reviewer in writing reviews to the reviewer since experience will help a reviewer to compose a sound review. Experience thus indicates that trusting this particular reviewer might be less risky. Yet for some other factors, this process may appear to have less face validity. For example, I sorted status badges that reviewers can earn on some platforms in the category “platform-related factors” and not “reviewer-related factors.” I did this although one might think that the badge indicates that a reviewer is trustworthy. However, it is the platform provider who decides whether the platform awards such badges, who determines what the criteria are for acquiring a badge, and who ensures a robust award process. The same logic applies to some of the factors that I systemized as “community-related.” For example, whether a particular review is consistent with other reviews of the same review object depends on the decision that the other reviewers made when writing their reviews. These other reviewers constitute the community. Obviously, the particular reviewer can also tune in the review to the community’s voice. From the perspective of the opinion seeker, however, the other reviews constitute the background against which to assess the consistency of a particular review. The opinion seeker thus has to determine how to respond to this background: Should they base their own assessment of the particular review on these other reviews, or should they discard them? In other words: The opinion seeker has to decide whether to trust in the community. I therefore categorized the consistency of a particular review with other reviews as a community-related factor.

Reviewer-Related Factors

The most exhaustively studied actor in the field is the reviewer. The opinion seeker’s perception of the reviewer’s states and traits is informed by the review that the reviewer wrote as well as by other information that the reviewer provides on the platform, usually when adding information to the user profile. Table 5 presents the reviewer-related factors that have been examined. Research literature often treats review characteristics independently from the reviewer. However, the reviewer is the author of the review and the review depends on their own selectivity. For example, Bannerjee and Chua (2019) examined how the attractiveness of review titles impacts trust but essentially, it is the reviewer who concocts this title. Table 6 presents the findings on review-related factors as a subset of reviewer-related factors.

Table 5. Examined Reviewer-Related Factors^a.

Table 6. Examined Review-Related Factors^a.

Option Provider-Related Factors

The impact of the provider of the reviewed option has been examined only rudimentarily (see Table 7). Notably, Bae and Lee (2011) found that the type of review object influences trust relationships. This factor should be treated with consideration. While the option provider decides which type of options they offer at a market, the review object itself is not an actor in its own right and thus does not possess own selectivity. According to the trust concept informing this literature review, the type of review object should thus rather be theorized as moderator variables than as independent variable.

Table 7. Examined Option Provider-Related Factors^a.

Platform-Related Factors

Research has examined factors that refer to (1) the platform provider as an actor in the online review process and (2) to the information that a platform offers about other actors. For example, platforms might use algorithms to create meta-information, such as marking suspected non-authentic reviews and reporting the status of a platform user. An understanding of studied platform factors can be gained from Table 8.

Community-Related Factors

With regard to the community, research has focused on two different types of factors: (1) meta-information provided by the community that relates to the quality assessment of the specific review via comments or recommendation ratings (e.g., usefulness and helpfulness), and (2) context information that is derived from considering the specific review and its author within the platform environment, such as the consistency of the specific review with other reviews on the same object or the total number of reviews on the object. Table 9 provides an overview on examined community factors.

Table 8. Platform-Related Factors^a.

 Table 9. Examined Community-Related Factors^a.

External Actors-Related Factors

External actors have only occasionally been considered in the field. The little interest in these factors is not surprising, given that external actors remain outside the core online review process. Table 10 summarizes the findings on this factor category.

Table 10. Examined External Actor-Related Factors^a.

Opinion Seeker-Related Factors

Individual characteristics of the opinion seeker can be associated with their willingness to trust. Studies examined the effects of knowledge, states, and, most often, traits on trust in online review contexts (see Table 11).

Table 11. Examined Opinion Seeker-Related Factors^a.

Discussion

Assessment of the State of Research

Research on online reviews has examined a considerable number of factors and how they influence trust. Figure 4 provides an overview of examined factors and their frequency of their consideration in research. For most factors, the findings are relatively distinct, whereas the impact of some others is less clear. In general, however, the field appears to be rather fragmented and heterogeneous. Research uses many different operationalizations of trust and is spread over various journals and disciplines.

The Role of Theory

This literature review set out to synthesize and systemize the existing findings. At the same time, however, the review also brings to light some of the field’s theoretical shortcomings and empirical gaps. Based on this analysis, the main shortcoming of previous research is a certain lack of awareness and reflection of the research’s formal object “trust” (or “credibility,” respectively). In general, the reviewed studies do not refer to comprehensive concepts of trust (e.g., Barber, 1983; Giddens, 1990; Kohring, 2004; Lewis & Weigert, 1985; Luhmann, 1968, 1975/2018; Möllering, 2001) that could more accurately capture the complexity of trust relationships and help interpret the results. This finding echoes a longstanding dissatisfaction with the state of empirical trust research in general. Already in 1991, Wrightsman (1991, p. 411) cautioned that “the general concept of trust deserves much more theoretical analysis. Measurement has advanced more rapidly than conceptual clarification.” Ten years later, McKnight and Chervany (2001, p. 38) emphasized that in the absence of such theoretical analysis, “the plethora of empirical studies (…) has brought trust research to so confusing a state.” For the sake of this review article, I adapted the trust concept advanced by Luhmann (1968, 1975/2018) and other authors following in his footsteps (e.g., Kohring, 2004; Meyer & Ward, 2013; Morgner, 2018) to the online review context. This is not to say that the line of reasoning on trust that I drew on is the only possible line for informing research on trust in online review contexts. Every theoretical perspective enables researchers to see certain aspects while obscuring others, as does this. Nevertheless, my conceptual choice permits the identification of some unfortunate consequences that result from the scarce engagement with trust concepts.

Figure 4. Considered Factors in Research on Trust in Online Reviews.

Note. Size of bubble corresponds to frequency of consideration (number of studies) of the factor in research on trust in online reviews.

This review brings to light some of the field’s theoretical shortcomings. First, its scarce engagement with trust concepts leads to a fuzzy understanding of who or what the trustor trusts in. Two trust objects are referenced in the studies, the reviewer and the review itself, and sometimes both. The theoretical foundation proposed here could help to resolve this issue. For example, I argued above that trust refers to the own selectivity of actors and that therefore, trust in online reviews should analytically be attributed to their actions. Unsurprisingly, studies consistently report significant correlations between trustworthiness of reviewer and trust in reviews (e.g., M. Y. Cheung et al., 2009; Clare et al., 2018). Furthermore, the online review process model revealed that there are more actors involved in the process than only reviewer and opinion seeker so that the emergence of trust relationships becomes more complex.

Second, the scarce engagement impacts the possibilities to interpret data in meaningful ways. On the one hand, many scales are imported from research on trust in social contexts other than online reviews. From the perspective of the trust concept adopted in this article, trust depends on expectations that are specific for a social context so that dimensions of trust potentially differ from context to context. This perspective thus increases sensitivity for the risk that imported scales might result in neglecting the peculiarities of trust in online reviews. For instance, the scale used most frequently in the field (Ohanian, 1990, 1991) measures trustworthiness of celebrity endorsers in an advertising and marketing context. It seems at least debatable whether trustors’ expectations about celebrity endorsers are similar to their expectations about online reviewers. For example, different from celebrity endorsers, online reviewers are expected to be independent from the option provider. On the other hand, I found that there is a great heterogeneity of items to measure trust. This heterogeneity makes it difficult to compare findings across studies. The trust concept adopted here can inform the operationalization of trust and protect research from inconsistencies when composing scales of items that relate to trust in different ways, for example to dimensions of trust, reasons to trust, and abstract synonyms of trust (e.g., Ayeh et al., 2013; Dou et al., 2012).

Third, the scarce engagement sometimes obscures how factors relate to one another. For example, I found that in some reviewed studies, manifestations of certain dimensions of trust are actually conceptualized as independent variables, i.e., antecedent of trust. For example, Banerjee and Chua (2019) study the credibility of descriptions as an antecedent of trust. Following Kohring’s (2004) concept of trust, this factor can be identified not as an antecedent but rather as an element of trust because it directly refers to a particular trust dimension that he conceptualizes as “trust in the accuracy of descriptions.”

Fourth, this shortcoming restricts the generalization of research findings. Without an appropriate theoretical foundation that helps clarify the conditions for legitimate generalizations, the validity of findings remains limited to the specific context in which the data was generated. For instance, it remains unresolved whether it is possible to transfer findings from a study on online reviews of hotels to reviews of physicians or electronics.

Empirical Gaps

Besides these conceptual challenges, this literature review identified some empirical gaps. First, many factors and factor combinations have only been examined in one or a very limited number of studies, with regard to single review objects (e.g., restaurants, physicians, films), on one platform, and in one cultural context. Moreover, a substantial number of studies uses student samples. In the face of a lasting replication crisis in social and life sciences (Open Science Collaboration, 2015), a more stable empirical validation of some findings appears preferable.

Second, research has examined the factor categories with different degrees of diligence. Whereas the reviewer and the opinion seeker received much attention, the situation is different for the other actors who are involved in the process. The focus on reviewer and opinion seeker is understandable because these two actors can be said to constitute the basic dyad in online review contexts. As the model of the review process illustrates, however, the other actors also play a role in the emergence of trust.

Third, little is known on how different factors interact with one another. While there are studies that examine interactions (e.g., Craciun & Moore, 2019; Flanagin & Metzger, 2013), a more systematic, comprehensive, and coordinated approach to determining interactions is missing.

Fourth, most studies adopt a deductive approach to identify factors that might possibly be related to trust in online reviews. Exploratory approaches do exist in the field (e.g., Clare et al., 2018; O’Reilly et al., 2016), yet are rare. If accepting the theoretical foundations of this review, trust should be considered as directed toward expectations about other actors’ behaviors that are specific to the social context in which trust relationships emerge (Kohring, 2004). A lack of exploratory studies might consequently result in overlooking relevant factors that are specific to online review contexts and thus cannot be derived from research on trust in other contexts.

Future Research Directions

Based on the analysis of strengths, shortcomings, and gaps, I want to propose three endeavors that I think are worthwhile for advancing the field.

Corroborating the Body of Knowledge

Research findings for most factors are unambiguous. The fact, however, that many factors or factor combinations have been examined only once and in specific online review environments, on specific platforms, and in specific cultural contexts calls for replication studies. Moreover, more comparative studies are needed. Researchers can corroborate initial insights by examining if, for example, the relevance of factors that determine opinion seekers’ willingness to trust differs between decisions to buy a plant detergent and to visit a physician or between Northern American and Eastern Asian cultures.

There are some factors for which the evidence is not clear. These include: non-benevolent motives of the reviewer, homophily (reviewer-related factors), correctness of orthography, quality of review, two-sidedness of review, emotionality of review, trustworthiness of descriptions in the review (review-related factors), trustworthiness of platform, reviewer status indication (platform-related factors), receptive attitude of opinion seeker, and experience of opinion seeker with reviews (opinion seeker-related factors). There are many possible reasons why studies yield diverging results. For example, different review objects, platform environments, or cultural backgrounds might explain differences. In some cases, diverging results may also be due to study designs. In still other cases, diverging results might follow from different operationalizations of trust. The fact that the studies usually differ in more than just one aspect makes it difficult to pinpoint the exact reasons that led to diverging results. Moreover, in the cases of the ambiguous factors, the number of studies—and thus the available data basis—is simply too small to identify patterns and reach informed conclusions. Here, further studies are needed to more precisely determine the influence of these factors. A promising starting point is to explore these factors in qualitative studies that allow to address the respective factors in more detail. For example, adopting thought elicitation techniques such as the think aloud method allows researchers in the field to explore the considerations of opinion seekers while engaging with online reviews.

Interestingly, it is sometimes the qualitative studies that yield contradictory findings. For example, while the majority of studies finds that positive valence increases trust, the qualitative studies by Filieri (2016) and Prendergast et al. (2010) demonstrate that positive valence can also have negative effects on trust if reviews become overly positive. While qualitative studies usually do not claim generalizability of their findings, they still can yield detailed and differentiating insights into the effects of specific factors. In particular, this example shows that it might be instructive to reconsider whether identified effects of factors such as valence are indeed linear.

Integrating the Body of Knowledge

It is necessary to integrate the existing findings from different disciplines that are interested in trust in online review contexts to overcome the fragmentation of the research field. Respective attempts might help to better pool resources because the integration of findings makes it easier to identify true research gaps. To tackle the fragmentation, the framework presented in this article might be further refined and adjusted to more specific research interests. Another promising approach to integrate findings might be meta-analyses because this method allows researchers to determine the effect strengths of trust factors across various quantitative data sets. A key challenge in such an undertaking will, however, be the great heterogeneity in operationalizing trust.

Extending the Body of Knowledge

Extending knowledge on trust in online review contexts is a more extensive task. Regarding the theoretical rigor of research, I suggest that engaging with comprehensive concepts of trust (e.g., Barber, 1983; Giddens, 1990; Kohring, 2004; Lewis & Weigert, 1985; Luhmann, 1968, 1975/2018; McGeer & Pettit, 2017; Möllering, 2001) beyond the level of measures and scales provides a more solid foundation for understanding trust in online review contexts. Researchers need to adjust these concepts to online reviews as a specific context. Specifically, the online review process does not only include reviewer and opinion seeker, but also option provider, platform provider, community, and external actors. All of these actors have their own selectivity and, consequently, they influence the emergence of trust relationships.

With regard to empirical directions, I identify three worthwhile directions for extending the body of knowledge. First, experiments and standardized surveys dominate the field. While these two methods are perfectly apt for studying trust in online reviews, I nevertheless want to argue for a greater methodical variety and a stronger consideration of qualitative approaches, such as problem-centered interviews, Q methodology, think alouds, and virtual ethnographies. Due to their open and exploratory character, qualitative approaches allow identifying factors that are specific to trust in online reviews (Clare et al., 2018). Furthermore, by allowing for a greater consideration of individual rationales, qualitative approaches might yield “thick descriptions” (Geertz, 2017) that help contextualize findings. Another promising approach to advance the field is the use of cognitive neuroscience. Various commentators have highlighted the potential of cognitive neuroscience for trust research (Dimoka, 2010; Gefen et al., 2008). This potential has not yet been tapped in the context of online reviews. The approach rests upon the finding that trust, as well as distrust, spans distinct brain areas (Dimoka, 2010). Researchers have successfully applied neuroscientific procedures to general trust research (Tzieropoulos, 2013), and Krueger and Meyer-Lindenberg (2019) proposed a trust model based on neuroscientific findings. These pioneer works hold the potential to add an innovative angle to the study of trust in the context of online reviews.

Second, it is likely that there exist various types of opinion seekers that differ in the importance they attach to specific factors. While one opinion seeker’s willingness to enter a trust relationship may largely depend on, for example, review valence, emotionality, and prior experience with other reviews, these factors might be of minor importance to other opinion seekers. Duffy (2017) suggests opinion seekers’ trust in their own benevolence and integrity as a possible distinguishing feature. There might also exist various types of opinion seekers regarding different usage situations, e.g., decisions on different review objects or decisions made on mobile devices and on-the-go. I encourage researchers to pay more attention to specific opinion seeker types and usage situations. To tackle this task, explorative approaches should prove instructive as a first step. For instance, the Q methodology allows researchers to reconstruct individual perspectives on trust in online reviews and, on this basis, determine what unites opinion seekers who share common viewpoints.

Third, trust relations emerge and grow over time as they unfold within interaction histories. Since the body of reviewers is multitudinous and reviewers often remain anonymous beyond some basic profile information, it seems unlikely that interaction histories unfold between individual peers in general. Nevertheless, opinion seekers gather experience with platforms and platform communities. I propose that these experiences will have a noticeable impact on the emergence of trust. The opinion seeker’s familiarity with a platform has been examined by three studies (Casaló et al., 2015; Lim & Van Der Heide, 2015; Van Der Heide & Lim, 2016), but in general we know little about how prior experiences with platform and platform communities impact trust. This situation motivates two suggestions: First, future study designs should control for prior experiences of opinion seekers. Second, given the dearth of longitudinal studies in the field, researchers should apply designs that allow tracing the formation of trust around specific platforms and their communities over time.

Limitations

The quality of literature reviews crucially depends on the corpus on which they are based (Whittemore & Knafl, 2005). Although I carefully searched for relevant studies, using a broad search term and databases from various disciplines, I cannot guarantee that I identified every academic publication on trust in online reviews. The formal exclusion of non-peer reviewed journal articles and articles published in languages other than English constitutes an obvious limitation. Furthermore, I synthesized the findings on a rather high level of abstraction. It could be argued that the individual factors that constitute categories in this review should be less condensed. I opted for this level of abstraction because the operationalization of most factors was sufficiently similar to group them into one category. However, a more granular systematization might yield a more detailed picture of the state of research. Finally, this review adopted a strict focus on articles that study trust (or credibility) as a dependent or mediating variable. I did not, however, consider the outcomes of trust relationships like purchase intentions (e.g., S. Wang et al., 2015), review adoption (Lis, 2013), attitude toward the review object (e.g., Grabner-Kräuter & Waiguny, 2015), or willingness to re-visit specific online review platforms (Baker & Kim, 2019). Likewise and to avoid “scope creep” (Booth et al., 2016, p. 98), I did not consider concepts that are or might be related to trust, most notably usefulness and helpfulness of reviews (Clare et al., 2018). Future literature reviews on these concepts might help to draw a more comprehensive picture of the role of trust in the online review process.

Conflict of Interest

The author has no conflicts of interest to declare.

Acknowledgement

The author thanks Sevda Can Arslan (Paderborn University) for her involvement with an early outline of this mansuscript that we presented at the 2016 Annual Conference of the Digital Communication division of the German Communication Association in Braunschweig, Germany. The author also thanks the editorial team of Cyberpsychology: Journal of Psychosocial Research on Cyberspace as well as five anonymous reviewers for sharing their helpful advice and support

Appendix

Full appendices are available in the pdf version of the manuscript.