Fake AI Reviews Are Flooding Amazon and Google

The five-star review has stopped being a reliable signal. Originality.ai analyzed Google reviews and found 19% were machine-generated in 2024, up from 5.01% in 2019 and 12.21% in 2023, a 279% jump in five years (Originality.ai, 2026). Amazon fights the same flood with moderation teams and detection sweeps, but the economics run the wrong way: a prompt costs pennies and a believable paragraph passes most readers, so the bots outproduce the firepower meant to stop them.

Key takeaways

• AI-written Google reviews roses from 5.01% in 2019 to 19% in 2024, a 279% increase.
• 2024 was 12.21%, up from 5.98% in 2020 and 8.56% in 2021.
• The FTC final rule bans buying and selling fake reviews, including AI-generated ones.
• Platforms delete millions of reviews a year, but detection is a yield game.
• Sellers keep going because automated moderation cannot keep pace with generated volume.

How AI reviews took over the rating economy

The incentive chain is straightforward. Rankings on Amazon and Google surface the products with the most and fastest reviews, so sellers came to see seeded ratings as reputation insurance. Originality.ai’s study of heavily-shopped categories found the machine-written share climbed fastest in electronics, home goods, and impulse purchases, exactly the sections where a 4.8 star average converts a casual browser (Originality.ai, 2026).

Aggregators accidentally compound the problem. Ratings platforms pull from the same polluted pools, so a fake-heavy Amazon listing can inflate the score it shows on sites that quote Amazon stars, spreading the contamination to review ecosystems the seller never touched. The review is now a vector, not a document, and its reach is measured in embedded widgets, not in actual reads.

What the FTC rule actually prohibits

In August 2024 the Federal Trade Commission finalized a rule banning the sale or purchase of fake reviews. It covers buying, selling, and propping up reviews, including AI-generated ones, and it allows civil penalties of up to tens of thousands of dollars per violation (FTC, 2024). The rule also bars review suppression and review hijacking, where a seller swaps an existing listing to inherit its five-star history.

The rule is real regulation, but enforcement runs through detection, and detection lags generation. Its practical value so far is quiet: reputable sellers point to it when their categories are flooded, and the platforms cite it as cover for their own takedowns. What the rule does not do is put a moderator at every listing page, which is where the volume problem lives.

How Amazon and Google actually respond

Amazon says it blocked or removed hundreds of millions of suspicious reviews in the years after it started investing in automated detection, using signals like review velocity, purchase history, and language patterns. Google similarly purges policy-violating reviews and tightened enforcement after the FTC rule. Both publish their billion-figure takedown counts, and neither publishes a false-positive rate for the innocent review it deleted.

The hard truth is that moderation is a yield game. Platforms catch the obvious floods and the reviewers who post one a minute, but slow-drip fake reviews written to read like human praise slip through at scale. Every wrapper that looked at the last three reviews will tell you the system works, until a product with a clean 4.9 and zero real owners shows up in your search.

How to spot an AI review in the wild

• Near-identical phrasing across dozens of reviews, the same sentence skeleton, different nouns.
• Universal praise with no negatives, no tradeoffs, and no product-specific detail.
• Review dates clustered in a suspicious burst, sometimes dozens on one day.
• Overwritten adjectives like "unbelievable," "game-changing," and "lifesaver" in every fifth review.
• Accounts with one review, no photo, no other activity, and a generic handle like "Customer" or a name plus random digits.

None of these alone proves a review is fake, which is exactly what the flood exploits. The tell that matters most is the combination: hundreds of near-identical five-star blurbs from sparse accounts on a product that launched three weeks ago. When the pattern appears, treat the rating as manufactured and weigh the negative or neutral reviews you can actually read.

How fake reviews hurt the sellers who behave

Decent products lose the most. A fake-review factory gives a mediocre listing a wall of five stars and a top-of-page position, so the honest product with thirty genuine reviews sits on page two. That sales shift is real money, and it is why established brands now monitor their own categories for suspicious entrants and file copyright-style takedowns against identity thieves.

The pattern also degrades trust in the platform itself. When shoppers cannot tell a genuine rating from a generated one, more buyers stop reading stars and start filtering on price, and the review system quietly becomes a tie-breaker nobody believes. That erosion is harder to measure than a lost sale, but it changes every other metric in the store.

What the FTC announces in its new rule

The Commission announced the final rule on August 14, 2024, describing it as banning the sale and purchase of fake reviews and testimonials and giving the agency the power to seek civil penalties. The announcement specifically notes the rule addresses reviews that misrepresent they were written by real users, which covers the output of AI systems presented as authentic customer experience (FTC press release, 2024).

The news hook is the same for every enforcement milestone: a rule that starts with positive reviews, negative reviews, and everything between is hard to litigate at scale. Experts who followed the rule expect the first cases to target the largest, most provable operations, not the long tail of sellers posting a hundred reviews a week. That is where the venue is, and where the public pressure is.

Why platforms can’t just moderate it away

The core problem is asymmetry. Review generation costs seconds and pennies, while human review costs minutes and dollars. Every time a platform sharpens its detector, the generators adapt their prompt, and the cycle repeats on the next product launch. The economics have no equilibrium where the good guys win, only a slower slide.

Platforms respond by shifting the flag onto verification instead of detection: verified purchase badges, seller ratings, and "bought this product" confirmation. Those help honest reviewers, but fake purchases can be manufactured too, which is why the most durable fix is demand-side. Shoppers who learn to judge the listing, not the stars, remove the reward that funds the whole industry.

The buyer’s playbook for AI review noise

Treat the star average as a hypothesis, not a verdict. Open the one-star and critical reviews first, since fake operations rarely seed those, and read the middling three-star reviews for the tradeoffs real owners describe. Cross-check the same product on two platforms: if a product is 4.9 on Amazon but 3.2 on a site with stricter moderation, the high score is probably manufactured.

Buy from the channel with accountability. Products with verifiable sellers, customer service numbers, and return policies are easier to fix when the review lied. And remember the fastest tell of all: the products with a thousand reviews and a two-month history did not get them from human patience.

What the numbers mean for 2026

The trendline has no off switch. Generators got cheaper and better through 2025, and the share of synthetic text in the ratings pool followed. The only realistic scenarios for 2026 are more of the same or a consumer backlash that makes authenticity itself a selling point, which some direct brands already market as "reviews from verified buyers only."

If you are a shopper, the practical move is unchanged: ignore the average, read the worst reviews, and buy from sellers with skin in the game. If you are a seller, the only durable defense is a review policy that is visibly human, because a wall of perfect AI prose marks you as either lazy or dependent on the same factory you claim to fear.

Why the big brands are suing over this

The enforcement picture took a sharp turn as brands started pursuing fake-review operations directly. Across 2024 and 2025, household names filed suits against sellers who bought credible-sounding reviews, and platforms began cooperating with law enforcement to trace review farms to their payment rails. The result is that a seller who once treated fifty AI reviews as cheap insurance now faces a civil complaint with the screenshots as evidence (FTC rule context, 2024).

The litigation wave matters because it changes who pays. Under the FTC’s final rule, each fake review can be treated as a separate violation, which turns a month of automated posting into hundreds of thousands of dollars in potential civil penalties. That asymmetry is what finally moves sellers who ignored takedown notices, a compliance reality the 2026 market is only beginning to price in.

What verified-purchase badges actually prove

Verified purchase is the platform’s best-worst answer. It proves a transaction happened, which blocks the cheapest bots, but it does not prove the review is honest or human. AI can generate a plausible review from a real purchase history, and sellers still supply product to reviewers by writing off the cost as marketing. Treat the badge as a minimum bar, not a guarantee.

The counter-signal people skip is recency. A reviewer who bought the product yesterday and reviewed it today cannot report on durability, so their praise is inherently shallow no matter how well it is written. Longest-horizon reviews, the kind written weeks after use, are far harder to fake at scale, which is why the ratio of same-day to month-later reviews is one of the better screens a shopper can run.

Frequently asked questions

Related coverage

• AI slop is eating the web
• Why AI chatbots hallucinate and how to catch them
• How to verify AI answers before you trust them
• The dead internet is not a conspiracy theory anymore
• A practical guide to digital minimalism
• Originality.ai: AI content in Google reviews up 279% since 2019
• FTC announces final rule banning fake reviews
• Amazon: how we keep customer reviews trustworthy