AI Wrote That 5-Star Review: How to Spot Fake Reviews Before You Buy

Have you ever bought a product with hundreds of glowing five-star reviews, only to find it nothing like the praise?

You probably cannot tell a fake review from a real one by reading it — but you can catch it by reading the pattern around it.

Generative AI made fake reviews cheap to produce and hard to spot. The Transparency Company analyzed 73 million reviews across home, legal, and medical services and found nearly 14% were likely fake — including 2.3 million it identified with high confidence as partly or entirely AI-generated. Trust in online reviews fell from 76% in 2019 to 46% in 2022. This guide gives you six checks that still work, and a clear rule for what to do when a listing fails them.

A smartphone showing five gold stars with one star dissolving into digital pixels, examined under a magnifying glass

Why AI Changed Fake Reviews — and Why the Old Clues Stopped Working

Fake reviews used to be easy to spot. AI removed the tells.

Before generative AI, fake reviews were often easy to catch: broken English, the same phrase repeated across accounts, praise with no detail. AI now writes reviews that are grammatically clean, varied in style, and packed with plausible specifics. The scale is industrial. Amazon says it blocked more than 250 million suspected fake reviews in 2023 and more than 275 million in 2024 — and those are only the ones it caught.

The way review farms work now was laid bare in September 2026, when Singapore's competition watchdog announced its largest fake-review probe. A marketing firm called Reputifly had supplied fake reviews to about 100 businesses. Generative AI wrote the review text, varying sentence structure and adding local details and small imperfections so it read as human. The firm recruited people through Telegram to post the reviews and gave clients a tool that calculated how many five-star reviews they needed to reach a target rating. In one case, three five-star reviews appeared within ten minutes.

The law has caught up, at least on paper. The U.S. Federal Trade Commission's rule banning fake reviews and testimonials took effect on October 21, 2024. It covers reviews written by AI and reviews from people who never used the product, and it lets the FTC seek civil penalties of up to $51,744 per violation — the figure when the rule took effect, adjusted for inflation each year. In September 2024, the FTC sued Rytr, the company behind an AI writing tool, accusing it of offering a service that could flood the market with fake reviews.

Even so, fakes do not disappear quickly. A study published in Marketing Science found that roughly half of a product's fake reviews were eventually deleted — but only after an average of more than 100 days online. That is more than three months in which a fake review can steer your purchase.

Why Reading the Review Will Not Save You

In tests, people — and detectors — get fooled.

Yale School of Management professor Balázs Kovács tested whether people could tell real Yelp restaurant reviews from ones written by ChatGPT-4. The participants were often wrong, even when they were paid for correct answers. They mistook AI reviews for human ones more often than the reverse — and they were more confident about the AI-written reviews than the real ones. Kovács also tested AI detection software, which was fooled by the AI-written reviews too. The same weakness showed up in student essays: AI detectors flagged 61% of real student essays as AI-written, while missing plenty of actual AI text.

Another experiment makes the point even sharper. Researchers at the University of South Florida told participants that only 10 of 20 restaurant reviews were authentic. The participants still classified an average of 11.38 reviews as real. We are biased toward believing reviews, and AI exploits that bias.

The lesson is simple: if your method is "does this sound human?", you have already lost. Judge the pattern around the review instead.

The 6 Checks That Actually Work

No single check proves a review is fake. Two or more failing together is your signal to walk away.

1. Read the distribution, not the average

A 4.6-star average can be a worse sign than a 4.3.

Click into the rating histogram before you trust the average. Genuine products usually follow a J-curve: mostly five stars, a meaningful band of four stars, and a smaller tail of one- and two-star reviews from genuinely disappointed customers. Research from Northwestern University's Spiegel Research Center found that purchase likelihood peaks when a product's average rating sits between 4.2 and 4.5 stars — and falls as the rating approaches a perfect five. Shoppers sense, correctly, that a near-perfect score is often manufactured.

The math shows how little it takes. A product with 20 reviews averaging 4.0 stars holds 80 stars in total. Add 20 fake five-star reviews and the average becomes 180 divided by 40 — 4.5 stars. Twenty fake reviews can lift a mediocre product straight into the zone buyers trust most. Treat a wall of five stars with almost no two- or three-star reviews as a warning, especially soon after launch.

2. Read the 3-star reviews first

Three-star reviews are the closest thing to ground truth.

There is almost no economic incentive to fabricate them. Brands buying reviews want five stars; competitors planting reviews want one star. The three-star band is where real friction lives: the zipper that broke after a month, the size that runs small, the delivery that took two weeks. Read five to ten of them before you read the praise. If the complaints repeat the same defect, believe the repetition.

3. Check the dates, not just the stars

A sudden burst of five-star reviews is the oldest fake-review signal — and it still works.

Sort the reviews by most recent and look at the calendar. A healthy product gains reviews steadily. A burst of five-star ratings within a few days, especially after a long quiet period, deserves a closer look. Remember the Singapore case: three five-star reviews within ten minutes. Real customers do not coordinate like that.

4. Click the reviewer, not just the review

One review tells you nothing. A reviewer's history tells you almost everything.

Tap the reviewer's name and read their other reviews. Warning signs include little or no history, dozens of reviews posted within a short period, or glowing praise for unrelated businesses in distant places — a plumber in Ohio, a restaurant in Bangkok, a phone case in the same week. The FTC's own consumer guidance lists exactly this: an account created just to review one product may be fake.

5. Compare the same product on another platform

A 4.9 on the seller's own site and a 3.8 somewhere independent is a gap worth investigating.

Fake reviews are usually concentrated where they are easiest to place. Check the same product or business on a second platform — Google, Amazon, Trustpilot, or an independent expert review — and compare. A large gap between the seller's site and independent sites is one of the strongest signals available, because faking every platform at once is expensive.

6. Look for "Verified Purchase" and disclosure badges

Verification is not proof — but its absence on a wall of praise is a flag.

On Amazon, a "Verified Purchase" badge means the platform confirmed a purchase happened. It does not guarantee honesty, but a listing full of five-star reviews with almost no verified purchases deserves doubt. Also look for disclosure labels: when a reviewer received a free product or payment, that must be disclosed. One distinction matters here. A real customer who uses AI to tidy up spelling is not the problem. The FTC rule targets reviews that pretend to come from someone who does not exist, or from someone who never used the product.

What to Do When a Listing Fails the Checks

Do not try to prove a review is fake. Change what you do next.

If two or more checks fail, stop using that listing as evidence. Find the same product on another platform, or look for an independent expert review that tested it. Read the one-star reviews for repeated complaints and check how the business responds — a seller that answers complaints with specifics is more credible than one that only collects praise.

Report what you find. On Amazon and Google, use the "Report" link on the review itself. In the United States, you can also report fake reviews to the FTC at ReportFraud.ftc.gov. If you already bought the product, check the return window first — then leave an honest review describing what actually arrived. That helps the next buyer more than a report does.

Match your effort to the stakes. For a $15 phone case, a quick pass through the first three checks is enough. For a $300 appliance, a hotel booking, or a service contract, run all six. The cost of a fake review is not the review itself — it is the purchase you would not have made with honest information.

Key Takeaway

AI made fake reviews read like real ones, and no detector — human or software — reliably catches them one by one. What gives them away is the pattern: the shape of the ratings, the timing of the posts, the reviewer's history, and the gap between platforms.

Before your next purchase, spend two minutes on the rating distribution and the three-star reviews — that is where the truth usually survives.

Sources

  • U.S. Federal Trade Commission, "FTC Announces Final Rule Banning Fake Reviews and Testimonials," August 14, 2024 — ftc.gov
  • Amazon, "How Amazon uses AI to combat fake reviews" — aboutamazon.co.uk
  • Associated Press, "The internet is rife with fake reviews. Will AI make it worse?" December 23, 2024 — techxplore.com
  • UNSW BusinessThink, Nitika Garg, "Can you trust a five-star review written by AI?" November 11, 2026 — businessthink.unsw.edu.au
  • Scam.sg, "Singapore's Largest Fake Review Probe: How to Spot AI-Generated Reviews" — scam.sg
  • UK Department for Business and Trade, "Fake online reviews research: executive summary," April 25, 2023 (11–15% fake in three product categories)

Comments

Popular posts from this blog

AI and Jobs: Which Tasks Change First, According to Researchers

What Marketing Tasks Will AI Replace Next — and Which Still Need a Human?

How AI Screens Your Resume Before a Human Sees It — and How to Write for It