Fake reviews are now banned in the US, the UK and the EU, but bans do not clean up a product page overnight. Here is what counts as fake, which warning signs actually carry weight, and a short checklist to run before you trust a star rating.
The Review at NYU is not affiliated with New York University.
Short answer. You cannot reliably spot a fake review from its wording alone, so judge the context instead: look for sudden bursts of five star reviews, reviews that describe a different product, reviewer accounts with no real history, and any sign that rewards were tied to positive ratings. Then read the middle ratings, check dates, and compare with at least one independent source. As of 7 November 2026, fake reviews are illegal under the FTC rule in the US (since October 2024), the DMCC Act in the UK (since April 2025) and EU consumer law (since May 2022), but enforcement is uneven and detection is never certain.
Most "how to spot a fake review" lists mix good advice with folklore. Bad grammar, too many exclamation marks and a reviewer who says "I" a lot all turn up in these lists, yet plenty of honest people write that way and plenty of paid writers do not. So before ranking any warning sign we set out four criteria, and every signal in this article is judged against them.
Our method was desk research: we read the FTC rule and consumer guidance, the CMA's fake reviews guidance, the EU directive text, platform transparency reports and the academic studies cited below. We did not run our own detection experiment or buy fake reviews to test anything. Where a number comes from a platform about itself, we say so, because self-reported enforcement figures cannot be checked independently.
"Fake" covers more than a stranger inventing a five star rave. The current rules in the US, UK and EU all describe a family of practices, and knowing the family helps you recognise each member.
The core case: a review that claims to reflect a real experience that did not happen, or that misrepresents the experience. The FTC's final rule (16 CFR Part 465, section 465.2) covers writing, buying and selling such reviews, and the FTC states in the rule's preamble that AI generated reviews are covered. The CMA's guidance defines a fake review as one that appears to reflect a genuine experience but does not, and stresses that fakes can be negative as well as positive. A review is not fake just because the business disagrees with it.
Paying for a review is not automatically illegal. What crosses the line in the US is a reward that depends on the review being positive or negative (section 465.4 of the FTC rule). A card in the box that says "Leave us five stars and get a free gift" is the textbook example. In the UK, the problem is a concealed incentive: an incentivised review that is not clearly disclosed is a banned practice under the Digital Markets, Competition and Consumers Act 2024 (DMCC Act). Disclosed programmes, such as a platform badge that marks reviewers who received a free sample, are allowed, and the FTC's consumer advice says it is up to you how much weight to give them.
Reviews by owners, employees, relatives or agents that do not disclose the connection are covered by section 465.5 of the FTC rule. The same logic sits behind section 465.6, which bans company controlled "review websites" that pose as independent.
A seller attaches its new or weak product to an old listing with thousands of good reviews, or merges product "variations" that are really different items, so that reviews of one thing appear under another. The UK guidance calls this catalogue abuse and treats it as most likely misleading when the products differ materially. Note a gap: the FTC proposed a specific ban on this practice but dropped that section (the proposed section 465.3) from the final rule, so in the US it falls back on the general ban on deceptive practices rather than a dedicated provision.
Generative AI makes fake text cheap and fluent. In September 2024 the FTC brought a case against Rytr, an AI writing service with a "Testimonial and Review" feature, alleging that some subscribers used it to produce hundreds, and in some cases tens of thousands, of reviews that could contain false details. The proposed order barred Rytr from selling any service dedicated to generating reviews. The case was authorised by a 3 to 2 vote, with two commissioners dissenting, which tells you that where the line sits for tool makers is still argued.
A rating can be fake without a single fake review in it, if the bad ones never appear. The FTC rule (section 465.7) covers suppressing reviews through threats and misrepresenting that the reviews shown are all or most of those submitted. The CMA lists suppressing genuine negative reviews as a way of publishing reviews misleadingly. Related tricks include asking only happy customers for reviews and fake social proof such as bought followers or views (section 465.8 of the FTC rule).
These are the signs worth your time, ordered roughly from most to least useful. The examples are illustrative patterns, not quotations from real listings.
The FTC's consumer advice specifically tells shoppers to watch for a burst of reviews over a short period. Example: a product with a handful of reviews over a year suddenly gains forty five star reviews in one week, then goes quiet. Most platforms let you sort by newest, which makes this visible in seconds. Research supports the concern: in a study of fake review markets operating in Facebook groups, published in Marketing Science in 2022, He, Hollenbeck and Proserpio found that buying fake reviews was associated with a short term jump in average rating and review count, and that after sellers stopped buying, ratings fell and one star reviews rose.
This is the most damning sign and one of the easiest to check. If reviews on a pair of headphones talk about fabric softness, the smell of a candle or how well a phone case fits, you are probably looking at merged or hijacked reviews. In an April 2022 investigation, the UK consumer group Which? found that nine of the ten top rated Bluetooth headphones it checked on Amazon UK carried reviews for unrelated products such as umbrellas, jigsaw puzzles and cuddly toys.
The FTC advises checking whether a reviewer has written other reviews, because an account that exists to post one review of one product may be fake. The opposite pattern matters too: a profile with dozens of five star reviews for unrelated, little known brands, all posted within days, looks like a review for hire account. Click the reviewer's name when the platform allows it.
Look for reviews that mention a refund, gift card or free product "for my honest review", or complaints in critical reviews that the seller offered money to change or delete a rating. A reward conditioned on positivity is what the FTC rule targets, and its presence taints the whole rating, not just the reviews that mention it.
Fake negative reviews exist too. The FTC reminds shoppers that companies may post fake bad reviews to hurt competitors, and the Yelp research below found that restaurants facing more competition were more likely to receive unfavourable fake reviews. Look for one star reviews that are vague, posted in a cluster, or that recommend a rival by name.
Deceptive reviews do differ statistically from honest ones. In the most cited study on the question, Ott and colleagues at Cornell (ACL 2011) collected 400 fake positive hotel reviews written to order and compared them with 400 genuine ones. Fakes used more first person singular pronouns, more superlatives and positive emotion words, and fewer concrete spatial details such as room features or location. But the same study found three human judges were only 53.1% to 61.9% accurate, close to a coin flip, while a trained classifier reached about 90% on that dataset. The lesson for shoppers: style cues are real but too faint for a human to rely on, and AI tools now make polished, specific sounding text cheap to produce.
Before treating a pile of five star reviews as suspicious, it helps to know that genuine ratings are lopsided too. In a 2009 paper in Communications of the ACM, Hu, Pavlou and Zhang analysed Amazon ratings for books, DVDs and videos collected between February and July 2005 and found that 78%, 73% and 72% respectively were four stars or higher. The typical shape is a J: a tall bar at five stars, a smaller bump at one star and few ratings in between.
The authors traced this to two self selection effects. Purchasing bias: people who buy a product already value it more than those who skip it, so they rate it higher. Under-reporting bias: people with strong feelings write reviews to "brag or moan", while people with moderate views rarely bother. Their point is that the simple average of a J shaped distribution is a poor summary of quality even when every review is honest.
Who buys fakes? Research on Yelp by Luca and Zervas (Management Science, 2016) found that about 16% of Yelp restaurant reviews were filtered as suspicious, and that review fraud was more likely among businesses with weak reputations (few reviews or recent bad ones) and among independent restaurants rather than chains. The Marketing Science study reached a related conclusion for Amazon: rating manipulation was mostly used by lower quality products, not by good new products trying to get noticed. In other words, the products that most need fake reviews are often the ones you most want to avoid.
The table sorts common warning signs by how much weight we think they deserve, using the four criteria from the start of this article.
| Signal | Strength | Why | How to check |
|---|---|---|---|
| Reviews describe a different product | Strong | Almost never happens with honest reviews; a classic sign of merging or hijacking (CMA "catalogue abuse"). | Read 10 to 20 reviews, including the most helpful ones. |
| Reward offered for a positive review | Strong | Directly targeted by the FTC rule and the DMCC Act; skews every rating on the page. | Search reviews for "gift card", "refund", "free" or "contacted me". |
| Burst of five star reviews in days | Strong | Named by the FTC; matches the pattern of bought reviews in published research. | Sort by newest and scan the dates. |
| Reviewer profiles with one review, or many five star reviews of unrelated brands | Strong to medium | Named by the FTC; harder to fake than wording, though new genuine users exist. | Open three or four reviewer profiles. |
| Rating shape shifts sharply over time | Medium | Consistent with buying campaigns, but product updates can also change ratings. | Compare old and new reviews; check for a new model or formula. |
| No middle ratings on a big product | Medium | Possible suppression, but popular good products can look like this. | Look at the star breakdown and read any critical reviews. |
| "Unverified purchase" label | Weak to medium | Raises doubt, but verified purchases can be bought too, and some honest reviewers bought elsewhere. | Note the share of unverified reviews among the five star ones. |
| Generic, gushing, superlative language | Weak | Statistically linked to fakes in research, but humans judge it near chance and AI tools hide it. | Use only as a tiebreaker alongside stronger signals. |
| Poor grammar or spelling | Weak | Common in honest reviews and among non-native writers; no good evidence it predicts fakes. | Ignore on its own. |
| Very high average rating | Weak | Genuine ratings are J shaped and skew high anyway. | Look at the distribution, not just the average. |
The FTC announced its final rule on consumer reviews and testimonials in August 2024; it was published in the Federal Register on 22 August 2024 and took effect on 21 October 2024. It bans fake or false reviews and testimonials (including AI generated ones), buying reviews conditioned on sentiment, undisclosed insider reviews, company controlled review sites posing as independent, review suppression and fake social media indicators. Because it is a trade regulation rule, the FTC can seek civil penalties; when it announced warning letters to ten companies on 22 December 2025, the per violation maximum was reported as $53,088. The FTC's consumer alert from that month again told shoppers to check a variety of sources and to watch for sudden bursts of reviews.
The consumer protection part of the DMCC Act came into force on 6 April 2025. Its list of banned practices now includes submitting or commissioning fake reviews or concealed incentivised reviews, publishing reviews or review information (ratings, counts, rankings) in a misleading way, and offering services that facilitate any of this. Businesses that publish reviews must take reasonable and proportionate steps to prevent and remove banned reviews. The CMA published its fake reviews guidance (CMA208) on 4 April 2025 and can now fine businesses up to 10% of global annual turnover directly, without going to court first.
Directive (EU) 2019/2161, which EU countries had to apply from 28 May 2022, added review rules to the Unfair Commercial Practices Directive. A trader that gives access to consumer reviews must say whether and how it ensures those reviews come from people who actually used or bought the product. Two practices are banned outright: claiming reviews come from real buyers without reasonable and proportionate checks, and submitting or commissioning false reviews or misrepresenting reviews to promote products. How serious the gap was became clear in January 2022, when a European Commission coordinated sweep of 223 major websites found that at least 55% potentially violated the directive, 144 could not show they did enough to ensure reviews were authentic, and 176 gave no information on whether incentivised reviews were allowed.
Large platforms publish big enforcement numbers. These are self reported and cannot be audited from outside, but they show the scale.
Two cautions. First, "blocked" figures count what platforms caught, not what got through, so they cannot tell you how much fakery remains. Second, platform policies differ in what they call fake, so the numbers are not comparable with each other.
This takes about five minutes and works on most shopping and local review platforms. If two or more strong signals fire, stop trusting the star rating and rely on other sources.
With the criteria and signals in place, here is how we would act.
In the US, the FTC rule on consumer reviews and testimonials (16 CFR Part 465) has banned fake reviews, including AI generated ones, since 21 October 2024, with civil penalties per violation. In the UK, the DMCC Act 2024 made fake and concealed incentivised reviews a banned practice from 6 April 2025. In the EU, the Omnibus Directive has banned submitting or commissioning false reviews since 28 May 2022.
There is no single reliable sign. The strongest signals are about context rather than wording: a sudden burst of five star reviews in a short window, reviews that describe a different product, reviewer profiles that exist only to praise unrelated brands, and any offer of a reward in exchange for a positive review.
Usually not with confidence. Even before modern AI tools, a 2011 Cornell study found human judges identified fake hotel reviews only slightly better than chance. Treat polished, generic text as a weak hint and look for stronger signals such as timing, reviewer history and whether the details match the product.
Not automatically. Genuine ratings tend to pile up at five stars anyway, because people who buy are already inclined to like the product and people with strong feelings are the ones who write. Read the three star and two star reviews, check how recent the reviews are, and compare against at least one independent source.
Mozilla shut down Fakespot on 1 July 2025 and its Firefox Review Checker on 10 June 2025. Some other review grading tools exist, but none has published independent accuracy testing we could verify, so a manual checklist plus independent expert reviews is the safer approach as of 7 November 2026.
Report it to the platform first, using the report or flag option next to the review. In the US you can also report it at ReportFraud.ftc.gov with the words fake review in the comments, in the UK to the CMA or Citizens Advice, and in the EU to your national consumer authority.