How AI is tackling beauty’s fake reviews problem

Today, sources of product reviews are nearly as infinite as beauty retailers. Shoppers can turn to Amazon, Sephora, a brand’s website, an influencer’s Instagram post, a vlogger’s “first impressions” video, Reddit and other sites to see users’ satisfaction with a product before making a purchase. But as more brands and platforms partner with influencers, the featured reviews are becoming less trustworthy.

As a result, a number of technology companies have cropped up, offering shoppers solutions for sifting through the noise. While their services aren’t limited to reviews of beauty products, the category has been among their most popular. It makes sense. According to a 2017 report by TABS Analytics, 57 percent of beauty buyers check online reviews when making an online purchase.

Data analytics company Fakespot, for one, uses artificial intelligence to provide an instant analysis of the reliability of a product’s reviews. When a user pastes a review onto the site from Amazon, Yelp, TripAdvisor and the App Store, a letter grade is generated and an analysis is provided, stating something to the effect of, “Our engine detects that, in general, the reviewers have a suspiciously positive sentiment.” What’s more, individual reviews that are least likely to be reliable are pulled out, so customers can concentrate on the reviews worth paying attention to.

Beauty is the second most popular category on the site, following electronics, said the company’s chief strategy officer, Ming Ooi. To date, Fakespot has graded an average of 30 percent of beauty reviews unreliable.

The Fakespot algorithm “considers both the review and the reviewer’s background to determine if a review is unreliable,” said Ooi, who went on to offer tips for customers looking to identify fake reviews. In short, she said shoppers should consider it a red flag if a product has many reviews in a short period of time, or if multiple reviews contain similar language or “brand speak.”

Review platform Influenster hosts 24 million user-generated reviews from 3 million “social savvy” members, some of whom receive products from brands through Influenster in exchange for providing honest reviews on the platform. The brands pay Influenster to get products in influencers’ hands, but the company is adamant that all featured reviews are legit.

“We have taken many steps to adhere to the FTC’s regulations and best practices,” said Elizabeth Scherle, Influenster’s co-founder and president. “Every review originating from an Influenster product trial campaign is auto-tagged with a clear label reading ‘Received free from Influenster’ and a box icon drawing attention to the disclosure.” Scherle said members are free to love, hate or feel ambivalent about a product, and negative reviews don’t affect their standing on the platform.

For beauty company Proven, which launched this year and uses machine learning to offer customers personalized products, fake reviews are a big concern. The company uses an AI engine to crawl millions of reviews, then uses the data paired with individual customer surveys to create products optimized for each shopper. But before formulating products based on reviews, weeding out fake reviews with a fraud detection algorithm is crucial.

The Internet is filled with blog posts and tutorials on how to exchange positive reviews for free products. 

According to Ooi, there are a many ways a fake review come to be. Some are obvious: Companies ask friends and family to leave positive feedback, and influencers suck up to brands, hoping to maintain relationships and keep receiving free products. But there are also groups Ooi terms “review clubs,” or companies that encourage high reviews in exchange for swag. And brands can hire “review farms,” or companies set up to proliferate positive reviews for products.

As long as reviews equal capital for brands, shoppers will need to take them with a grain of salt — or, at the very least, consider the source before before adding to cart.

Get news and analysis about fashion, beauty and culture delivered to your inbox every morning.