How to spot fake Amazon reviews: a 13-point checklist (2026)

Fake reviews used to be easy to spot. Five-star rating from an account with no other reviews, written in broken English, posted within 24 hours of the product launching. The obvious tells are mostly gone — the industry has gotten sophisticated, and AI has made production trivial. What's left is a subtler problem: reviews that look completely normal, were written by actual humans (or convincing machines), appear on verified-purchase accounts, and are still commercially planted.

This is a method for finding them. It's the method I use before every review I publish on this site, because I need to know whether the existing reviews of a product are giving me useful information or whether they're a polished signal I need to ignore. The method is also useful for any reader who wants to stop getting misled by the dominant signal on a product page.

Why this matters more in 2026 than it did five years ago.

Three things have changed.

First, the volume. In 2023, Amazon reported that it had proactively blocked more than 250 million suspected fake reviews — a number that only represents the ones Amazon caught. The ones that got through are, by definition, not in that count. Independent research published in arXiv in 2024 documented that products buying fake reviews cluster together in the review-network graph, and that the actual rate of manipulation in some categories runs as high as 30 to 40 percent of reviews. Supplements and electronics are especially bad; home goods and apparel follow close behind.

Second, the quality. Large language models have made generating a plausible-sounding "I bought this and here's my experience" review a 20-second task for anyone with a ChatGPT subscription. The tells that used to distinguish human from machine prose — incongruous phrasing, grammatical oddities, repetitive structure — mostly don't apply anymore. Fake reviews now read like real reviews, because they're generated by systems trained on real reviews.

Third, the counter-tools are collapsing. Fakespot, the most popular consumer-facing fake-review-detection browser extension, shut down in early 2026 after being acquired and wound down. Other tools exist but most are either paywalled or rely on older detection methods that assume the obvious tells. The reader is increasingly on their own.

Which means the skill of reading reviews critically has moved from "nice to have" to "required for not wasting money." This article is the version of that skill I've written down so I don't have to reinvent it every time.

What counts as a "fake" review, legally and practically.

Before the method, a definition. The word "fake" gets used loosely. Legally, the FTC Final Rule on the Use of Consumer Reviews and Testimonials — codified as 16 CFR Part 465, effective October 21, 2024 — treats a review as fake or deceptive under several specific conditions:

• The review is written by someone who does not exist (AI-generated without disclosure, bot-generated, or attributed to a fictional person).
• The review is written by someone who did not actually use the product or experience the service.
• The review is written by a company insider — an officer, manager, employee, agent, or close relative of one — without clear disclosure of that relationship.
• The review was obtained in exchange for compensation conditioned on expressing a particular sentiment (positive or negative).
• The review was "hijacked" — originally written about a different product, then repurposed to inflate a new one.

Violations of this rule can result in civil penalties of up to $51,744 per violation, which is the FTC's current maximum. The rule doesn't create a private right of action — only the FTC can enforce it — but it represents the legal line between "acceptable marketing" and "unlawful deception."

Practically, for a shopper, the categories matter less than the outcome: a fake review is any review that was produced to influence your purchase rather than to honestly describe someone's experience. Below is how to recognize them.

The 13 tells I look for.

None of these is a single-point-of-failure detector. A real review can hit one or two of these and still be legitimate. But when a review hits four or more, the probability that it's planted rises sharply, and when an entire product page is dominated by reviews that hit these patterns in aggregate, the product's entire ratings distribution is suspect.

1. Excessive generic praise with no specific complaints. Real people who buy a product notice at least one minor thing that annoys them — the box was hard to open, the instructions were confusing, the color was slightly different than pictured. Reviews that are uniformly positive across every dimension, with no friction points at all, are statistically rare in the wild and extremely common in paid review factories.

2. Brand-name repetition in unnatural frequency. Normal reviewers refer to the product as "it" or "this" after the first mention. Fake reviews often repeat the brand or product name multiple times in a short review, because the reviewer was paid per-mention or because the reviewer was given a script that emphasized SEO-relevant keywords. If a 150-word review uses the full brand name five times, something is wrong.

3. The "I was skeptical but..." opener. An extraordinarily common template that begins "I was skeptical at first, but..." or "I had my doubts, but..." and then pivots to enthusiastic praise. The structure is psychologically engineered — the fake skepticism is meant to establish credibility before the endorsement. Real reviewers who were skeptical often stay partially skeptical even after the product worked.

4. Impossibly-specific backstories. "I'm a 42-year-old marathon runner training for my third Boston qualification, and I needed a product that could handle my 80-mile weeks..." The specificity sounds credible but the template is suspicious because it checks off "identifies as a credible reviewer" in a way that real reviewers usually don't. Real marathon runners writing reviews mention they run. They don't establish their credentials in the first sentence.

5. Perfect grammar with no personality. The paradox of modern fake reviews: the grammar is better than a real human would produce. No sentence fragments, no typos, no idiosyncratic punctuation, no regional vocabulary. Just clean, professional prose that reads like a product description. Real reviews have texture.

6. Overuse of the product name in place of pronouns. "The [Product Name] arrived on time. The [Product Name] was easy to set up. The [Product Name] worked as expected." This repetition is both a paid-per-mention artifact and an AI-generation artifact, because LLMs often over-reference the subject of a review to stay on-topic.

7. Praise that matches the sales page language almost verbatim. If the manufacturer's product description says "ergonomic design for all-day comfort" and the review says "I love the ergonomic design — so comfortable for all-day use," the review writer was either reading the product page while writing or was handed talking points. Real reviewers describe the experience in their own words, which rarely matches marketing copy.

8. No mention of any downsides whatsoever. Related to #1 but worth calling out separately: a review that explicitly states "there are no downsides" or "I can't think of anything I'd change" is almost certainly planted. Every real product has something the reviewer would improve. Real reviewers mention it even when the product is otherwise excellent.

9. Reviews that read like advertorials. If a review has a clear problem-solution-benefit structure ("I was tired of X. Then I found [product]. Now X is no longer a problem and I've saved Y"), it's probably following a template. Real user reviews are usually more chaotic — people describe what they bought, mention the one or two things that mattered to them, and move on.

10. Identical timing patterns across multiple reviews. If you look at a product's review feed and notice that fifteen 5-star reviews appeared within a 36-hour window, with no surrounding review activity before or after, that's a coordinated burst — usually from a paid review group or an incentivized email campaign that violated FTC § 465.4.

11. Reviewer profile that reads as artificial. If you click through to the reviewer's profile and they've only ever reviewed products from the same brand, or they post 5-star reviews across wildly disparate categories (a blender, a phone case, a supplement, a gardening tool, all rated 5 stars within the same week), the account is either fake or incentivized. Amazon identifies these patterns at scale and blocks hundreds of millions annually, but plenty slip through.

12. The photo that's too professional. When a reviewer uploads a photo of the product, real photos are usually mediocre: poor lighting, cluttered background, shot from an awkward angle on a kitchen counter. Fake review photos are sometimes suspiciously professional — white background, studio lighting, the product posed aesthetically. If the "customer photo" looks like a stock image, it often is one. Reverse image search is a 30-second confirmation.

13. The review that resolves every hesitation you might have. This is the most sophisticated tell and the hardest to articulate. A planted review is often written after the planter has read the critical reviews on the same product page, and it's designed to preemptively address those critiques. "I was worried about durability based on some other reviews, but after six months of heavy use, mine still works perfectly." This review exists to neutralize objections. Real reviews don't usually resolve objections; they sometimes raise them.

Platform-specific signals to check.

Different platforms surface different kinds of information about a review. The signals to check, by platform:

• Amazon. The "Verified Purchase" badge means Amazon's records show the reviewer bought the product through Amazon. It doesn't mean the review is honest — verified purchases can still be incentivized — but the absence of the badge on a pile of 5-star reviews is a strong signal something is off. Also check the reviewer's profile for "Top Contributor" badges with suspicious review histories, and look for reviews with the incentivization language Amazon explicitly prohibits ("I received this for testing," "I was asked to review this") outside the official Vine program.
• Google Reviews. Google doesn't verify purchases, which makes fake reviews easier to plant. The signal here is the reviewer's full history: click the reviewer's name and look at all their reviews. If they've only ever reviewed one business positively, or if all their reviews are identically-worded across businesses, the account is likely fake.
• Trustpilot. Trustpilot has the added complexity that businesses can solicit reviews from their own customer lists, which isn't illegal but skews the results upward. Check the review distribution on Trustpilot for anomalies: a site with 85% 5-star reviews and almost no 4-star reviews is usually either (a) genuinely excellent in an unusual way or (b) asking only satisfied customers for reviews.
• Direct-to-consumer sites. Reviews hosted on a merchant's own site are the least reliable by default, because the merchant controls what gets published and can suppress negative reviews (also prohibited by the FTC rule, but enforcement is spotty). Cross-reference any DTC site's reviews against reviews of the same product on third-party platforms — if the patterns don't match, the DTC reviews are filtered.
• Reddit and forum mentions. These aren't formal reviews but often contain the most honest information. If a product has glowing reviews everywhere except Reddit, and the Reddit discussion includes real-user concerns not reflected on the review sites, trust Reddit. If Reddit looks clean too, that's a stronger positive signal.

Tools that still work, now that Fakespot is gone.

The tool landscape shifted in 2026 when Fakespot — acquired by Mozilla and then wound down — stopped operating. What's left is a smaller, more fragmented set of options, none as polished as Fakespot was at its peak, but usable.

ReviewMeta (revmeta.com) still operates and produces a "report card" analyzing suspicious reviews on Amazon products. Their methodology weights factors like unverified purchase ratios, incentivized-review language, and unusual rating patterns. It's not perfect — no tool is — but it's free and it runs in a few seconds. For any product I'm considering reviewing on this site, I run it through ReviewMeta first to see what Amazon's patterns look like before I commit time to testing.

Independent Chrome extensions that replaced Fakespot include RateBud, Savinoo, and a handful of others. They use similar algorithms. I'd trust the numerical output of any one of them as a rough signal and not a definitive answer.

Reverse image search on review photos is a manual but powerful tool. Google Images, TinEye, or Yandex can reveal whether a "customer photo" was lifted from a stock library or from the manufacturer's press materials. If the photo appears on multiple unrelated websites, it's not a customer photo.

Your own eyes on the review graph. Most product pages let you filter reviews by rating and sort by date. A product with a healthy review distribution will show a natural mix — bursts around product launches and major retail events, but with consistent activity in between, and a realistic distribution across ratings. A product with a manipulated profile often shows all of the good reviews clustered in time, or an obvious gap between "early reviews that are all 5 stars" and "later reviews that are more varied."

When reviews cluster in suspicious ways.

Recent academic work — the arXiv paper referenced above, and subsequent research — has established that products buying fake reviews form tight clusters in the review graph. If Seller A buys fake reviews from the same network as Sellers B and C, the same reviewer accounts tend to show up reviewing all three. You don't need the academic method to use this insight practically:

Pick a suspicious review on a product. Click through to the reviewer's profile. Look at the other products they've reviewed. If they've reviewed a cluster of similar-category products from obscure brands you've never heard of, all at 5 stars, within a narrow timeframe, you've found a fake-review farm. Every one of those 5-star reviews across every product in the cluster is suspect, which means every product in the cluster has an inflated rating, which means the entire corner of the marketplace is polluted.

This is why supplements and electronics — categories where clusters of obscure brands compete on ratings alone — have the highest fake-review rates. The cluster effect compounds the deception.

What the FTC rule actually gives you.

A quick note on what the October 2024 FTC rule does and doesn't do for consumers.

What it does: It makes the practices above explicitly illegal for businesses, advertisers, and marketing agencies. Civil penalties up to $51,744 per violation give the FTC teeth it didn't previously have. The rule specifically calls out AI-generated reviews, incentivized reviews, insider reviews without disclosure, suppressed negative reviews, and legal threats designed to silence critics. Companies that previously could operate in the gray area now operate clearly on the wrong side of a bright line.

What it doesn't do: It doesn't give you, the consumer, a private right to sue anyone over a fake review. Enforcement is exclusively by the FTC, which is a federal agency with finite resources and big priorities. Your ability to get a specific fake review removed still mostly depends on reporting it to the platform (Amazon, Google, Trustpilot, etc.) and hoping their internal review systems catch it. The FTC's rule matters most as a deterrent on the supply side — making fake-review operations riskier at the industrial scale — and less as a tool for individual shoppers to use in the moment.

You can report suspected violations to reportfraud.ftc.gov, and you should if you spot something egregious, but for day-to-day shopping the practical approach is still: learn to read reviews critically and rely on your own judgment.

A worked example (anonymized).

Let me walk through a real product I researched recently — anonymized because the specifics don't matter — and show how the method lands in practice.

The product was a supplement marketed at men over 30 for energy and recovery. Amazon rating: 4.6 stars across 2,847 reviews. On the face of it, a strong product. I was considering it for a review on this site, so I applied the method.

Rating distribution check. 71% 5-star, 18% 4-star, 5% 3-star, 2% 2-star, 4% 1-star. This is not a natural distribution. Natural distributions for genuinely good products have a stronger 4-star band than this — usually 25–35% — because real users who like a product often pick 4 stars as "very good" rather than saving 5 stars for exceptional. A 71% 5-star rate with a collapsed 4-star band is a fingerprint of rating manipulation.

Temporal clustering. Filtered reviews by date. The product had received 400+ reviews in a 30-day window after launch, of which 340 were 5-star. Review volume then dropped to 20–30 per month, and the rating distribution of the post-launch window was more like 55% 5-star, 25% 4-star — much closer to natural. Conclusion: the launch window was a coordinated review push; the organic baseline was meaningfully lower.

Language patterns. Sampled 20 5-star reviews. Four of them used the phrase "I was skeptical at first" (tell #3). Six of them repeated the brand name more than three times in under 150 words (tell #2). Nine of them had zero mentioned downsides (tell #8). Three had photos that reverse-image-searched to the product's own press kit.

Reviewer profile check. Clicked through three of the most effusive 5-star reviewers. Two had reviewed exclusively health and supplement products from small brands, all 5 stars, within a short window. One of those reviewers had reviewed 14 products from the same apparent cluster of brands within a 45-day period.

Verdict. I didn't review the product. Not because it was definitely bad — I never actually tested it — but because the signal was polluted enough that I couldn't trust the baseline. If I'd reviewed it and scored it independently, my review would either confirm the manipulated consensus (looking credulous) or contradict it (looking contrarian), and I had no way to know which was honest without data I couldn't cheaply get. Skipping it was the right call.

What to do about it when you're shopping.

Beyond the method above, a short list of practical moves that compound your defense:

• Sort reviews by "Most recent" and "Most critical" — not the default. Platforms default to a "relevance" sort that surfaces the reviews the platform thinks you'll find helpful, which usually skews positive. The most recent reviews reveal current product state (did the manufacturer change the formula? Did quality drop after success?). The most critical reviews tell you the specific failure modes.
• Read the 3-star reviews. Not the 5-star (which includes the planted ones) or the 1-star (which includes angry outliers and competitor sabotage). The 3-star reviews are where honest people describe what the product does well and what it fails at, without either enthusiasm or resentment clouding the description. They're usually the most useful reviews on the page.
• Look for reviews from people who'd notice a problem. If you're buying a camera lens, find a review from a photographer. If you're buying a chef's knife, find one from someone who mentions their actual knife collection. Their criticisms are more informed than the generalist 5-star.
• Cross-reference across platforms. If a product has 4.7 stars on Amazon and 4.7 stars on the merchant's own site and 3.1 stars on Trustpilot, trust Trustpilot. The platform with the most complaint volume is usually the one closest to the truth.
• Trust Reddit and niche forums more than product-page reviews. Fake review operations don't usually penetrate organic community discussion as effectively as they penetrate product pages. A subreddit dedicated to the category will usually have one or two honest threads about any popular product, and those threads are more useful than all the 5-star reviews put together.

The method above is what I use before deciding whether to even test a product. If a product's reviews are too polluted for me to form a baseline, I don't review it — I pick a different product in the same category where the signal is cleaner. That's part of the selection process at /how-i-review.

The broader point is that review literacy — knowing how to read between the lines — is now a prerequisite skill for spending money online. The platforms aren't going to save you. The FTC isn't going to save you at the individual-shopper scale. What you have is your own ability to pattern-match, cross-reference, and apply some skepticism before you click Buy. This article is the version of that skill I've written down so I can point to it instead of explaining it each time.

If you catch me violating any of the tells above in one of my own reviews on this site, email me. I'll log the correction publicly at /corrections.

Frequently asked, briefly answered.

Are fake Amazon reviews illegal? Yes. The FTC's Final Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465), in effect since October 21, 2024, makes writing, selling, buying, and disseminating fake reviews unlawful — including AI-generated reviews without disclosure, incentivized reviews, insider reviews, and suppressed negative reviews. Civil penalties reach $51,744 per violation. Only the FTC can enforce; there's no private right of action for individual shoppers.

Does the "Verified Purchase" badge mean the review is real? No. It only confirms Amazon's records show the reviewer bought the product through Amazon. The review itself can still be incentivized via outside platforms, refunded after posting, or written by paid review-network participants who keep the product as part of the deal. One signal, not a verdict.

What replaced Fakespot now that it's shut down? ReviewMeta (revmeta.com) is the most established free option still operating. Independent Chrome extensions like RateBud and Savinoo use similar approaches. Treat any single tool's score as a rough signal, not a definitive answer, and combine it with the manual checks above.

Can AI tools detect AI-generated reviews? Unreliably. General-purpose AI-detection tools have high false-positive rates on short, conversational text — exactly the genre most reviews fall into. They flag genuine human reviews as AI-generated and miss skilled, prompt-tuned fake ones. Behavioral signals (account history, posting patterns, photo provenance) beat text-classification tools at this point.

Where should I report a fake review I find? Two places. The platform itself (Amazon, Google, Trustpilot) is the only party that can actually remove the review. The FTC at reportfraud.ftc.gov won't act on individual reviews but does build cases against operators and businesses running coordinated networks.