SAN FRANCISCO – On Amazon, customer comments can help a product surge in popularity. The online retail giant says that more than 99 percent of its reviews are legitimate because they are written by real shoppers who aren't paid for them.
But a Washington Post examination found that for some popular product categories, such as bluetooth headphones and speakers, the vast majority of reviews appear to violate Amazon's prohibition on paid reviews. They have certain characteristics, such as repetitive wording that people likely cut and paste in.
Many of these fraudulent reviews originate on Facebook, where sellers seek shoppers on dozens of networks, including Amazon Review Club and Amazon Reviewers Group, to give glowing feedback in exchange for money or other compensation. The practice artificially inflates the ranking of thousands of products, experts say, misleading consumers.
Amazon banned paying for reviews a year and a half ago because of research it conducted showing that consumers distrust paid reviews. Every once in a while, including this month, Amazon purges shoppers from its site who it accuses of breaking its policies.
But the ban, say sellers and experts, merely pushed an activity that used to take place openly into dispersed and harder-to-track online communities.
There, an economy of paid reviews has flourished. Merchants pledge to drop reimbursements into a reviewer's PayPal account within minutes of posting comments for items such as kitchen knives, rain ponchos or shower caddies, often sweetening the deal with a $5 commission or an $10 Amazon gift card. Facebook this month deleted more than a dozen of the groups where sellers and buyers matched after being contacted by the Post. Amazon kicked a five-star seller off its site after an inquiry from the Post.
"These days it is very hard to sell anything on Amazon if you play fairly," said Tommy Noonan, who operates ReviewMeta, a website that helps consumers spot suspicious Amazon reviews. "If you want your product to be competitive, you have to somehow manufacture reviews."
Sellers say the flood of inauthentic reviews makes it harder for them to compete legitimately and can crush profits. "It's devastating, devastating," said Mark Caldeira, owner of the baby products company Mayapple Baby. He said his product rankings have plummeted in the last year and a half, which he attributes to competitors using paid reviews. "We just can't keep up."
Suspicious or fraudulent reviews are crowding out authentic ones in some categories, the Post found using ReviewMeta data. ReviewMeta examines red flags, such as an unusually large number of reviews that spike over a short period of time or "sock puppet" reviewers who appear to have cut and paste stock language.
For example, the first ten products listed in a search for "bluetooth speakers" had a total of 47,846 reviews, two thirds of which were problematic, based on calculations using the ReviewMeta tool. So were more than half of the 32,435 reviews for the top ten listed bluetooth headphones.
Diet pills and other supplements also generated large numbers of problematic reviews. Just 33 percent of the reviews for the top ten testosterone boosters listed on Amazon appeared legitimate, like 44 percent of reviews for the top listed weight loss pills, according to data crunched from ReviewMeta.
Incentivized reviewers give higher ratings than non-paid reviewers, according to ReviewMeta. The result is that consumers could unknowingly be purchasing poorer quality products.
"I don't like to be taken advantage of," said Eric Hall, 53, a research chemist in Minneapolis and an Amazon Prime customer who no longer trusts 5-star reviews. He sees them as a marker of likely fraud rather than excellence.
Amazon says it aggressively polices its platform for incentivized reviews. Amazon has filed five lawsuits since 2015 against people who write paid reviews and companies that solicit them.
"We know that millions of customers make informed buying decisions everyday using Customer Reviews. We take this responsibility very seriously and defend the integrity of reviews by taking aggressive action to protect customers from dishonest parties who are abusing the reviews system," an Amazon spokeswoman, Angie Newman, said in a statement. "We take forceful action against both reviewers and sellers by suppressing reviews that violate our guidelines and suspend, ban or pursue legal action against these bad actors."
"The issue with fake or unreliable reviews has not subsided at all but likely is worsening," said Ming Ooi, chief strategy officer for Fakespot, a review auditing site that analyzes comments, similar to ReviewMeta.
Problems with the authenticity of Amazon reviews come at a moment of broad public concern over the accuracy of information on platforms built by Silicon Valley. The spread of Russian disinformation and hoaxes on YouTube and Facebook has raised questions about the role of technology platforms in displaying and amplifying falsehoods, contributing to a climate of distrust and social division.
Amazon has escaped the scrutiny of its peers. But the same network effects that enable misinformation also increasingly distort online commerce. Social media has accelerated the practice of online reviewing because of its power to bring together groups of people who gather for a specific purpose, such as rating Uber drivers.
"We are committed to increasing the good and minimizing the bad across Facebook," a spokesperson, Rebecca Maas, said in an email. "There are many legitimate groups on Facebook related to online commerce, but the groups identified misuse our platform." Facebook would not disclose which groups it removed.
Sellers say that Amazon's position as the top e-commerce destination has spawned a race to master – and game – the company's systems. More than half of all online product searches start on Amazon, according to survey data by the digital marketing firm BloomReach. Landing among the first ten results on an Amazon search can drive an explosion in sales, according to sellers.
Amazon chief executive Jeff Bezos first championed the idea of showcasing customer reviews on his online bookstore in 1995, helping spread the notion that consumers make smarter choices shopping online than in brick-and-mortar stores because tech platforms enable them to instantly access the opinions of fellow shoppers. That paved the way for such review-based sites as Yelp, Uber and Airbnb.
But the vision Bezos popularized, of a review and ratings system that serves as a guide for consumers to make smarter choices, has given way to a system where some consumers are manipulated and misled. (Bezos is the owner of the Post.)
Amazon rankings are the new "battlefield" for online manipulation, said Renee DiResta, policy lead for the non-profit Data for Democracy, a group of technology researchers dedicated to promoting integrity online. She has conducted research on paid Amazon reviews by joining some of the Facebook groups. "There's a dark side to the race for the stars."
How Amazon comes up with its star ratings is a closely-guarded secret. On its website, Amazon says it uses a machine learning model that takes into accounts many factors, including the age of a review, helpfulness votes by customers, and whether reviews are from verified purchasers. Reviews are a minor factor in the overall rating, Amazon said, but it wouldn't quantify their weighting.
The company said it uses artificial intelligence to analyze "hundreds of thousands" of Amazon customers who have been banned from leaving reviews and uses the data collected to build computer models of their behavior in order to predict future techniques.
The auditing sites use a software algorithm that scrapes Amazon's website for suspicious patterns or attributes of the review or the reviewer. ReviewMeta then gives the product a new star rating based only on the reviews its system deems likely to be authentic. For example, deleting the suspicious reviews on a pair of wireless headphones from Atgoin dropped its star rating from 4.4 stars to 2.6.
ReviewMeta and Fakespot both say the ease of detecting potentially fraudulent reviews makes them wonder why Amazon isn't more stringent.
For two decades, Amazon permitted incentivized reviews, so long as reviewers disclosed that they had received a free or discounted product. But it began cracking down on the practice in 2015, acknowledging its struggles to control it.
"Despite substantial efforts to stamp out the practice," company lawyers wrote in a lawsuit, "an unhealthy ecosystem is developing outside of Amazon to supply inauthentic reviews."
Around the same time, Amazon began courting foreign sellers to sell products directly on its site. The move drastically worsened the problem of scamming, experts say, because competing on price and review-padding was a way for Chinese manufacturers to go up against entrenched brands that are well-known to American consumers.
Atgoin, an electronics company based in Shenzen, China, was one such company that leapfrogged to the top of Amazon rankings. In November, its $30 headphones had just a handful of reviews. Then, over a five-day period in December, the product received nearly 300 reviews, almost all of which gave five stars.
The ReviewMeta analysis found that more than 90% of all the reviews for the Atgoin headphones were suspicious. Many featured repeat phrases, including "I'll be using this for my gym workout going forward," and "comfortable to wear." By early February, the Atgoin headphones, which had 927 reviews, appeared at the top in organic search results.
It's unclear how Atgoin obtained the flood of positive reviews. Atgoin on its website says it offers consumers "free and exclusively discounted samples" in return for "your valued, honest feedback" – language that Amazon said broke its policies. Atgoin did not respond to a request for comment through its Amazon seller page or its website. After the Post's queries, Amazon removed Atgoin as a seller.
Amazon's ban has not cut down on paid reviews so much as push it into the shadows, say sellers and researchers.
In February, there were nearly 100 different Facebook groups, split up by geographic region and by product categories, where Amazon merchants actively solicited consumers to write paid reviews. One such group had over 50,000 Facebook members until Facebook deleted it after the Post's inquiry. There are also Reddit boards and YouTube tutorials that coach people on how to write reviews. Websites with names like Slickdeals and JumpSend let merchants give out discounted products, using a loophole to get around Amazon's ban.
Merchants seeking to defraud Amazon have flocked to Facebook in particular, said DiResta.
Last year, DiResta began studying and joining Amazon reviewer groups on Facebook. Her first act in the groups was to write "interested" next to post describing a pair of bluetooth headphones for $35.99. Almost immediately, a Facebook user purportedly named SC Li sent her a direct message, calling her "dear" and asking for a link to her Amazon profile. If she reviewed the headphones, SC Li said, he would reimburse her via her PayPal account.
Within an hour of getting SC Li's message, DiResta got a slew of direct messages from other sellers, asking her to review tea lights, containers, shower caddies, badge holders, sanding discs, rain ponchos, pocket-size vanity mirrors, and butterfly knives. The messages came in so quickly she said she barely had time to respond.
DiResta spent three months monitoring the groups. She observed the sellers using tactics to avoid detection by Amazon, because they focused on reviewers who have a long history of writing Amazon reviews. The sellers asked her for screenshots showing when she started her profile.
DiResta found that many of the Facebook accounts had no friends on the social network. Their only Facebook posts were about cheap products, and their profile pictures included stock photos. A reverse image search on SC Li's profile photo, of a man on a beach, for example, revealed a stock photo called "seaside man" that appeared on various on Chinese-language lifestyle websites, an indication of a fake profile.
"[Reviewers] just see it as a way to make extra money," DiResta said. "The question is why doesn't Amazon crack down more? These communities are not a secret."