Research and Advances
HCI Research Highlights

Technical Perspective: A Rare Glimpse of Tracking Fake Reviews

"Leveraging Social Media to Buy Fake Reviews," by Sherry He et al., represents a breakthrough in our empirical understanding of fake reviews on Amazon.

question marks on five stars of a review on a smartphone, illustration

The success of marketplaces such as Amazon is inextricably tied to the third-party sellers who compete on these platforms. As e-commerce platforms have grown in complexity, ratings and reviews have played a crucial role in consumer decision making. According to a Pew Research survey, 93% of Americans read online reviews (sometimes) before purchase. However, the prominent role played by reviews has an unintended consequence: namely, it incentivizes sellers to purchase fake reviews to gain a competitive edge. These fake “5-star reviews” not only deceive consumers into buying low-quality products, they also trick the platform’s algorithms into providing poor search results. In the long run, this can erode public trust in online markets.

Although academics and practitioners both agree on the dangers posed by fake reviews, there is little consensus on their severity and how to address the problem. For example, while Amazon claims fewer than 1% of reviews on the platform are inauthentic, independent monitors argue the number is larger than 20%. Gaining a better understanding of the ecosystem in which fake reviewers operate is a crucial step toward tackling fraud. Efforts in this direction, however, have been stymied by a lack of transparency (and data) into how platforms operate and how bad actors solicit fake reviews.

The accompanying paper represents a breakthrough in our empirical understanding of fake reviews on Amazon. The authors track Facebook groups (with thousands of users) where sellers purchase fake reviews. By matching the posts in these groups to products on Amazon, they obtain several key insights on the type of products receiving fake reviews, the short-term and long-term gains due to fake reviews, and their negative impact on consumers.

What sets this paper apart is its novel methodology, which circumvents the need for any proprietary data. By directly documenting Facebook groups, the authors provide a rare glimpse into a marketplace whose sole purpose is to manipulate Amazon’s review system. Particularly surprising is the sophistication of the players involved. For instance, sellers ask fake reviewers to purchase the product before posting a review to ensure the coveted “Verified Purchase” tag. Contrary to popular belief, the paper also finds fake reviews are not unique to only new products on the platform. This suggests sellers view fake reviews not as a one-time lever to gain traction but as promotional tools that can be used repeatedly.

Another factor contributing to the paper’s success is how it constructs an unobstructed timeline from the point when the fake review is solicited, and then posted on Amazon, all the way up to its deletion. Such a comprehensive picture allows us to verify a causal relationship from fraudulent reviews to product sales and, eventually, consumer dissatisfaction. Consequently, the accompanying paper can answer important questions about how fake reviews impact different stakeholders:

  1. How harmful are fake reviews to consumers?

One popular theory argues that as fake reviews become ubiquitous, their adverse effects would be minimized since consumers would simply treat them as advertisements. However, the paper refutes this hypothesis by showing fake reviews lead to a 70% increase in one-star reviews from dissatisfied buyers in the future.

  1. Is the platform doing enough to address review fraud?

Although a significant portion of fake reviews are indeed deleted by Amazon’s filters, the lag between posting and deletion is rather large (100 days on average). Therefore, one cannot simply ignore the harm caused to consumers before deletion.

  1. Why do sellers persist with fake reviews?

It is surprising that fake reviews are so prevalent given the risk of detection and one-star reviews. A concerning, emerging theory is the short-term sales boosts due to fake reviews are so profitable they outweigh immediate costs and any long-term risks.

How can the insights from this work be leveraged to fight online manipulation? One immediate prescription that follows is to combat fake reviews, platforms and regulators must be proactive and not just reactive. In recent years, Amazon has doubled down on punitive measures to disincentivize fake reviews. For example, after the publication of this paper, Amazon filed a lawsuit against the administrators of over 10,000 “fake review” Facebook groups.

A parallel strand of research has focused on ensuring consumers reach good outcomes despite the presence of fake reviews. This includes, for example, making search algorithms less sensitive to bursts of reviews, and de-emphasizing numeric reviews in favor of textual content. However, it is necessary to complement these efforts by also studying how unreliable review systems alter consumer behavior, for example, they may turn to third-party websites and friends for recommendations.

Finally, it also is worth cautioning that any research on fraud has a finite shelf life. As manipulators evolve and adopt increasingly sophisticated strategies, it is important that one does not rely on outdated understanding to combat fake reviews.


Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More