Amazon uses artificial intelligence (AI) tools, including machine learning models, advanced linguistic models (LLM), and graph neural networks, to analyze, detect, and remove fake reviews, as well as reassure users that those posting them are honest.
The e-commerce company believes that an essential part of the shopping experience on its platform is making it easier for customers to share their opinions about products, thus helping other customers make informed purchasing decisions.
However, among all the reviews posted on the platform, there are also others that are false, which some users share in order to “profit” from the shopping experience on Amazon. These reviews intentionally mislead customers by providing unbiased information. Techniques such as hijacking reviews are also implemented to sell fake products at competitive prices.
In this sense, in order to put an end to this type of fraudulent publication, the company led by Jeff Bezos uses a series of tools powered by artificial intelligence with which it tries to identify and eliminate fake reviews, as well as confirm which ones are genuine. .
As explained in a post on its blog, before publishing a review, Amazon analyzes the information mentioned for risk indicators that help detect the possibility of it being false. Therefore, if it finds signs, it acts quickly to “prevent or remove” them and, if necessary, “take additional measures.”
For example, if it is determined to be a fake review, action is taken such as terminating the “privilege” to post reviews, blocking the accounts of the offending individuals, or even taking legal action.
On the other hand, if a review is suspicious but can’t be clearly confirmed as fake, Amazon uses specially trained investigators and experts to “identify abusive behavior.” Specifically, this employee is responsible for “looking for and reviewing other risk signals before making a decision.”
Likewise, Amazon is also using “the latest advances” in artificial intelligence to detect these fake reviews, as well as manipulated reviews, fraudulent customer accounts and other violations “before customers even see them.”
As the company explains in detail, it uses machine learning models to analyze a large number of data. In this case, issues such as whether the seller has invested in the ad are taken into account, which allows him to create a number of additional revisions. Likewise, aspects such as submission of abuse reports related to the mentioned seller or history of reviews are also analysed.
Following this line, Amazon also uses large language models, along with other natural language processing techniques, to “analyze anomalies” that might indicate that a review is fake or that it is being incentivized with a reward like a gift card or free product. vs.
Another tool used by the platform is graph neural networks (GNN), through which it can understand complex relationships and behavioral patterns. As detailed, this is a “crucial” technique because to properly evaluate the review, both information related to the type of customer writing it, and the type of product.
Thanks to the combination of these AI techniques, Amazon claims to be able to identify fake reviews “more accurately” because it is a deeper analysis than that achieved using superficial abuse indicators. In fact, he asserts that with AI it is possible to “identify deeper relationships between potentially abusive behaviors.”
“Maintaining a reliable and safe shopping experience is our top priority,” said Rebecca Mund, head of external relations and trustworthy reviews at Amazon, noting that it will continue to develop new ways to prevent the publication of fake and misleading reviews. “Buy with complete confidence.”
For his part, head of Amazon’s fraud and abuse prevention team, Josh Meek, added that preventing fake reviews is important because not only “millions of customers trust the authenticity of reviews to make purchasing decisions,” but also “as millions of brands and business partners trust” Amazon “to accurately identify fake reviews and prevent them from reaching customers.”
Finally, Amazon highlighted that during 2022, it was able to “proactively” identify and block more than 200 million suspected fake reviews globally.
