Public protection groups and researchers say that artificial intelligence (AI) tools that allow people to write online reviews have put sellers, service providers and buyers in unfamiliar territory.
Fake reviews have long been a feature of many popular websites, such as Amazon and Yelp. These reviews are often traded on private social media groups between dealers of fake reviews and businesses willing to pay for them. Sometimes, businesses offer buyers rewards such as gift cards in exchange for good reviews.
However, AI tools, popularised by OpenAI’s ChatGPT, enable people to produce reviews more quickly and in greater numbers, according to experts in the technology industry.
Fake reviews can be found in a wide range of industries, including e-commerce, travel, and services such as home repairs, medical care, and music lessons.
The Transparency Company is a technology company and public protection group that uses software to detect fake reviews. The company said it started to see AI-generated reviews appear in large numbers in mid-2023. These reviews have increased quickly ever since.
In a recently published report, The Transparency Company examined 73 million reviews in three categories: home services, legal services, and medical services. Nearly 14 per cent of the reviews were likely to be fake. The company expressed a ‘high degree of confidence’ that 2.3 million of the reviews were partly or entirely AI-generated.
Last September, the Federal Trade Commission (FTC) took legal action against the company behind an AI writing tool called Rytr. The FTC accused Rytr of offering a service that could pollute the marketplace with fake reviews.
The FTC, which banned the sale or purchase of fake reviews in 2024, said that some Rytr users had used the tool to produce hundreds, or perhaps even thousands, of reviews. These reviews were posted in support of garage door repair companies, sellers of counterfeit designer handbags, and other businesses.
What are companies doing?
Major companies are developing policies to determine how AI-generated content fits into their systems for removing fake reviews. Some companies already use special programmes and investigative teams to identify and remove fake reviews. Nevertheless, companies are allowing users to use AI to some extent.
For example, spokespeople for Amazon and Trustpilot said they would allow buyers to post AI-assisted reviews as long as they represent the buyer’s genuine experience. Yelp has taken a more cautious approach, stating that its rules require reviewers to write their own reviews.
The Coalition for Trusted Reviews, launched last year by Amazon, Trustpilot, Glassdoor, Tripadvisor, Expedia and Booking.com, said that while people may misuse AI, the technology also presents ‘an opportunity to push back against those who seek to use reviews to mislead others’.
The FTC’s rule banning fake reviews, which came into effect in October, allows the agency to fine businesses and individuals involved in fake reviews. However, tech companies hosting such reviews are protected from the penalty. This is because, under U.S. law, they are not legally responsible for the content posted by outsiders on their websites.
Tech companies, including Amazon, Yelp and Google, have sued fake review dealers whom they accuse of selling fake reviews on their sites. The companies claim that their technology has blocked or removed a significant number of dubious reviews and accounts. However, some experts say they could be doing more.
‘Their efforts thus far are not nearly enough,’ says Kay Dean, a former federal criminal investigator who runs a public protection group called Fake Review Watch. ‘If these tech companies are so committed to eliminating review fraud on their platforms, why can I, one individual working without automation, find hundreds or even thousands of fake reviews in a day?’
Finding fake reviews
Researchers say that consumers can try to spot fake reviews by watching out for a few possible warning signs. Reviews that are overly positive or negative are suspect. Reviews that use highly specialised terms or repeat a product’s full name or model number are another possible clue.
When it comes to AI, research by Balázs Kovács, a professor at Yale University, has shown that people cannot tell the difference between AI-generated reviews and those written by humans. The study also said that some AI detectors may be fooled by shorter texts, which are common in online reviews.
However, there are some AI-related details that online shoppers and service seekers should be aware of. According to Panagram Labs, reviews written with AI are often longer and more structured, and include ’empty descriptors’. Empty descriptors include general phrases, attributes, and characteristics. The writing often includes overused phrases and opinions such as ‘the first thing that struck me’ and ‘game-changer’.
