Introduction

Fake reviews are a growing problem for online platforms, with a staggering 43% of Amazon reviews now likely to be fake, and 67% of consumers believing fake reviews are a common issue. 

As AI tools enable fraudsters to produce deceptive reviews at scale, detecting fake reviews has become more complex. This is a huge concern since nearly 60% of consumers heavily rely on reviews to make purchasing decisions, affecting not only individual choices but also the credibility and reputation of platforms.

As a result, detecting fake reviews now requires an  analytical approach using a combination of AI and machine learning, and focusing on behavioural patterns at the account level. 

‍

So, why do people create fake reviews?

Understanding the motivation behind fake reviews is key to addressing the issue effectively:

• Reputation enhancement: Businesses often create or purchase fake reviews to enhance their reputation quickly. This is common among new businesses, which may encourage friends and family to leave positive reviews. Others may purchase reviews from "review farms" or brokers.

• Undermining competitors: Known as "review bombing," this tactic is used by competitors to damage a business’s reputation by overwhelming their page with negative reviews. This aims to tarnish a brand's image and deter potential customers.

• Incentive-driven feedback: Companies sometimes offer discounts or freebies to customers in exchange for positive reviews. While these reviews may stem from actual experiences, the added incentives can lead to exaggerated positive feedback that doesn't reflect the true quality of the product or service.

‍

‍

What are the consequences of fake reviews?

The dangers of fake reviews extend far beyond misleading consumers, highlighting the importance of taking proactive measures to safeguard your platform.

• Consumer safety risks: Fake reviews on your platform can lead to hazardous purchasing decisions, particularly in critical sectors like health care or childcare, where trust is essential for safety. Such practices can put consumers at risk of actual harm.

• Reputation damage: The presence of fake reviews erodes trust in your platform as a trusted source of reliable information, diminishing brand loyalty and potentially inflicting long-term reputational harm. This could result in decreased user engagement and loss of market share.

• Unfair competition: Fake reviews disrupt the market's level playing field, making it more difficult for honest businesses that depend on legitimate customer feedback. This can distort competitive dynamics and harm the overall ecosystem.

• Legal and financial risks: Increasingly strict laws against fake reviews place platforms at risk of significant legal and financial consequences. For example, in the US, the FTC proposes penalties that could potentially amount to $50,000 per fake review per viewer. Learn about the legality of fake reviews here. 

‍

Analysing content alone can no longer detect fake reviews effectively

As generative AI tools become more mainstream, creating convincing copy is easier than ever. Within seconds, fraudsters can now generate hundreds of fake reviews that effectively mimic humans across multiple languages and contexts.

The following example demonstrates the power of ChatGPT in rapidly producing fake reviews:

‍

‍

Fraudsters having free and ready access to such tools means that simply analyzing the review content no longer effectively detects fake reviews. We need to look deeper at the behaviour of the accounts behind the reviews.

‍

How we detect fake reviews with AI & Behavioural Analytics

Themis's approach combines AI and behavioural analytics to understand and interpret patterns of account behaviour:

• Review patterns: We look for odd patterns in the frequency and timing of posted reviews, such as spikes in review activities -often suggesting a batch of fabricated reviews. 

• Connected accounts: Our system can spot groups of users working together to manipulate ratings across different platforms. 

• Behavioural footprints: We analyse metadata such as where the review came from, what device was used, and other digital signals that offer clues beyond the text.

By continually monitoring both the content and the context - who is writing, from where, how often, and their connections - we can provide you with a clearer picture of review authenticity on your platform. We don’t just catch lone fakes, but uncover entire networks of review fraud and review seller activity across multiple platforms.

And the integration of our system with platforms automates the detection process, reducing the need for manual moderation and letting you focus on providing genuine and trusted user experiences. 

‍