Amazon is upping its use of artificial intelligence to tackle the burgeoning problem of fake reviews, which bamboozle consumers into buying inferior products.
Customer reviews were once considered reliable and trustworthy, representing the real-world experience of consumers who had paid good money for products, rather than receiving one free of charge for a professional review.
Even if an individual review could be poor quality, from someone using a product incorrectly, or with an axe to grind, there was a certain safety in numbers. More than 100 five-star reviews can’t be wrong, surely?
Unfortunately, they very much can. Trust in customer reviews has plummeted on Amazon and around the web, as companies pay for five-star ratings, muddying the waters of what is a genuinely good product, and what has just paid for rave reviews.
Last year, Amazon targeted more than 10,000 Facebook groups generating fake reviews with legal action. Since then, the consumer rights group Which? has found plenty more.
While Amazon has used AI to tackle fake reviews for years, the company told the BBC that it is investing in “sophisticated tools” to help flag suspicious behaviour. Everything from sign-in activity and review history to an author’s relationship with other online accounts is fair game for the algorithm.
“We use machine learning to look for suspicious accounts, to track the relationships between a purchasing account that’s leaving a review and someone selling that product,” Dharmesh Mehta, head of Amazon’s customer trust team, explained.
“Through a combination of both important vetting and really advanced machine learning and artificial intelligence— that’s looking at different signals or behaviours— we can stop those fake reviews before a customer ever encounters it,” he added.
Some, such as Chris Downie, CEO and co-founder of the trust and safety platform Pasabi, welcomed the news. “Fake reviews directly influence $152 billion of online spending every year, damaging businesses, eroding consumer trust and poisoning the market,” he said.
“Harnessing the power of AI and other sophisticated technology is essential in the fight against this growing threat, especially with so many easy-to-use tools out there making it easy to produce fraudulent accounts and reviews at the touch of a button.”
Is AI the answer?
Which? is sceptical this can actually make a dent in such a mammoth problem, however.
“Amazon has been trying all sorts of technology to crack down on fake reviews and by all accounts that’s having some success,” the consumer group’s Harry Kind told the BBC. “But as far as we’re concerned, it’s still nowhere near enough to solve this huge problem.”
The problem is ultimately very hard to fix — with AI or any other means. As long as there’s big money to be made on Amazon, then companies are going to use every trick to try to take advantage.
One thing Amazon could do today is to block reviews from anybody who hasn’t purchased the product directly from the site. Currently, anybody can tap the “write a review” button on any product and type away (though, in Amazon’s defence, it’s not clear whether these reviews are given identical weighting to those from verified buyers).
But such a block wouldn’t be foolproof either. The so-called “brushing scam”, where companies ship free goods to unsuspecting consumers to artificially inflate sales volumes, is often accompanied by a fake review.
The Standard has also seen an example of a genuinely purchased ‘Amazon’s Choice’ product that came bundled with an incentive for a positive review.
“Write a review to get £15 Amazon gift card [sic],” a slip of paper enclosed with the product read. “Email us with your review screenshot. Thank you.”
This does not appear to be an isolated example.
While AI may be able to sniff out patterns of behaviour and suspicious wording, this kind of ad-hoc incentivisation is harder to snuff out. And it’s made significantly worse by Amazon’s gradual transformation into a third-party marketplace where anybody can sell their goods, with questionable oversight.
Without dedicated buyers picking out quality products for its virtual shelves, more than ever, shoppers are at the mercy of a customer reviews system that even Amazon itself accepts is flawed.
If artificial intelligence can’t address that problem, then Amazon’s reputation as the go-to online shop will surely suffer — though perhaps too late for its rivals to take advantage.
