Amazon's one-tap ratings drowned fake reviews in a flood of real ones

Fake reviews had become a first-order problem for the dominant online shopping platform. The FTC brought its first case over paid fake reviews against an Amazon seller of a weight-loss supplement, Amazon itself had filed at least five lawsuits over fake-review schemes in five years, and in some categories a flattering review of a faulty product could be outright dangerous.

In late 2019 Amazon quietly introduced one-tap ratings: shoppers select one to five stars for a product without writing anything. The feature is available only to customers who actually bought the item — 'verified' buyers — a hurdle written reviews don't have, since Amazon allows those from non-verified buyers too. Amazon's overall star score is not a simple average but a machine-learned model weighing recency and verification, though the company doesn't disclose how one-tap ratings weigh in.

The arithmetic did the work. Amazon said overall customer feedback increased and new products generated feedback sooner. As Patrick Miller, co-founder of agency Flywheel Digital, put it: 'fake reviews as a percentage of legit reviews should decrease', because schemers now have to buy far more phony ratings to break through the noise. The tradeoff: only 0.5–3% of buyers typically write anything, so some of that flow migrated away from descriptive reviews, and shoppers now had to work to see how much of a product's feedback was actually written.

Don't police fakes — drown them: make rating effortless and verified-only, so authentic feedback grows faster than scams can buy it.

Writing a review is high-friction — only 0.5–3% of buyers do — so Amazon moved feedback to a one-tap action any verified buyer can afford.

Restricting one-tap ratings to verified purchases adds a hurdle fake-review farms don't clear, unlike written reviews.

The fake-review fight shifted from detection to arithmetic: as authentic feedback grows, bought reviews shrink to a smaller fraction.

Amazon's rating model weighs recency and verification, so a flood of current verified one-taps moves scores more than stale written praise.

When you can't stop bad actors from adding noise, change the economics: make honest participation so cheap that noise loses by arithmetic.

Industry observers expected the authentic-feedback flood to keep raising the cost of buying influence, and Amazon — known for iterating on the behavior its changes produce — to adjust further. Schemes persisted: private Facebook groups recruited buyers with PayPal refunds in exchange for five-star write-ups. Sellers and brands lost some descriptive signal, and shoppers had to dig past the combined ratings-and-reviews number to find written context.