AI-Generated Fake Reviews and the FTC Rule
How the FTC rule on fake reviews treats AI-generated reviews, who is liable, what platforms do about them, and how to document and report suspected fakes.
The FTC rule on fake reviews and testimonials covers AI-generated reviews: a review written by software that claims a reviewer who does not exist, or an experience nobody had, is a fake review under the rule, and a business that writes, creates or sells it violates the rule, as does one that buys or spreads it knowing or having reason to know it is fake, and knowing violations can bring civil penalties. The rule is a United States federal rule, so it does not decide what the UK, the EU or Australia allow, and it gives private parties no right to sue under it.
That answer matters because AI changed the economics of fake reviews, not the legal test. The FTC said in its rule notice that AI-generated reviews are covered, and its stated hope was that the rule would deter their use. What the rule asks is simple: did the reviewer exist, did they have the experience described, and is the review what it claims to be? Whether a human or a language model typed the words does not change the answer.
This guide is for two readers. One is the owner of a business or a professional practice who suspects a competitor, a vendor or an ex-employee is flooding a rating with machine-written reviews. The other is a business that wants to use AI tools around reviews without crossing a line. Both need the same things: the exact text of the rule, the platform policies that apply in practice, a sober way to judge whether a review is machine-written, and a documented path to report it.
Related readBought a Business With Bad Reviews: What to Do- The FTC rule, 16 CFR Part 465, prohibits fake or false reviews, including reviews that falsely claim the reviewer exists or had a real experience. The FTC has said AI-generated reviews are covered.
- Writing, creating or selling a fake review is prohibited outright; buying or spreading one, or procuring one from insiders, is prohibited when the business knew or should have known it was false. Merely hosting reviews is exempt.
- Platforms act on their own policies much faster than any regulator. Yelp and Trustpilot treat AI-written reviews as violations; Google's policy turns on whether a review reflects a genuine experience.
- No single sign proves a review is AI-written. Document patterns, then report through the platform.
- The UK, EU and Australia each have separate rules with different tests and different enforcers.
Three facts frame everything that follows: when the rule applies, what a knowing violation can cost, and who can enforce it.
Effective date from the Federal Register notice; penalty figure from the FTC's January 2025 inflation adjustment, which the FTC said in a September 2026 notice it will keep applying, and which can change; private right of action from the FTC's rule question-and-answer page.
What the FTC rule says about AI-generated fake reviews
The rule bans, among other things, a business writing, creating or selling a review that materially misrepresents that the reviewer exists, used the product or business, or had the experience described. That is the core of Section 465.2, and it is the provision that reaches machine-written reviews. A synthetic persona has no experience to describe.
The FTC announced the final rule on August 14, 2024, by a 5-0 vote, and it took effect on October 21, 2024. The Commission's statement of basis explains why it matters for AI: tools now let bad actors produce large numbers of realistic reviews quickly and cheaply and spread them across platforms. In a footnote, the Commission states that AI-generated reviews are covered by the final rule. The rule needs no separate AI clause because it regulates the false claim, not the tool.
The parts of the rule that matter for AI reviews
Part 465 has more sections than the headline suggests, and several touch AI-driven schemes.
- Fake or false reviews (465.2). Prohibits a business from writing, creating or selling a misrepresenting review, from buying one or spreading one it knew or should have known was misrepresenting, and from procuring such reviews from its officers, managers, employees or agents. Generalized solicitations to purchasers and mere review hosting are exempt from the buying, spreading and insider provisions.
- Buying positive or negative reviews (465.4). Prohibits compensation or incentives given in exchange for, or conditioned on, a particular sentiment.
- Insider reviews (465.5). Requires clear disclosure of a material relationship when officers or managers review their own business, and limits soliciting reviews from relatives or staff.
- Company-controlled review sites (465.6). Prohibits misrepresenting a site the business controls as an independent source of reviews or opinions, other than consumer reviews.
- Review suppression (465.7). Prohibits using unfounded legal threats, physical threats, intimidation or knowingly false accusations to stop or remove a review.
- Fake social media indicators (465.8). Prohibits buying or selling bot followers or fake views that the person knew or should have known were fake, when they misrepresent influence for a commercial purpose.
Section 465.3 is reserved. The rule defines a consumer review as a consumer's evaluation, or a purported consumer's evaluation, of a product, service or business, and "purported" is the word that reaches invented reviewers; the definition's remaining conditions were not re-read for this guide. This summary follows the Federal Register text of the final rule; check the current text of 16 CFR Part 465 before acting.
Who is liable when an AI wrote the review
Liability follows the act and the knowledge, not the software. The rule reaches the person who writes or sells the fake, the business that buys or spreads it while knowing or having reason to know, and the manager who has insiders post it. It does not make a platform liable for merely hosting reviews, and it does not make a business liable for a fake posted by a stranger with no connection to it.
Related readGoogle Reviews Disappeared: Why and What to DoThe vendor who generates the reviews
The person or company that produces and sells fake reviews is the clearest target. In 2024 the FTC settled a case against an AI writing-assistant company whose service let subscribers generate false and deceptive reviews; the final order of December 18, 2024 barred it from selling any service dedicated to or promoted as generating consumer reviews or testimonials. On December 22, 2025, the Commission voted 2-0 to reopen and set that order aside, saying the facts pleaded did not support a Section 5 violation and that the order unduly burdened AI innovation. The FTC said it will still pursue companies that use AI to break the law. The lesson is narrow: a general-purpose writing tool is not itself a violation, while a campaign that uses it to post fake reviews still is.
The business that benefits
A business that commissions a review farm, or pays an agency that quietly does, falls under the buying and spreading provisions if it knew or should have known the reviews were false. The FTC says businesses should look for red flags that a testimonial is likely fake, though the rule imposes no general duty to investigate every review. We read a promise of hundreds of reviews in a week as such a flag, though the FTC page does not use that example.
The customer who uses AI to write a real review
A real customer who describes a real experience and uses a tool to polish the wording is not a fabricated persona. The rule asks whether the review falsely represents the reviewer, the experience or its nature, and a true review with help on phrasing does not obviously do that. That is our reading, not an FTC statement. Platform policies can differ: Yelp says reviewers may not use third-party AI tools to write reviews.
Related readRemove a Clutch Review: What Agencies Can DisputeWhat a business may and may not do with AI around reviews
A business may use AI to analyze reviews, draft replies that a person approves, and prompt real customers to leave honest feedback. It may not use AI to invent reviewers, invent experiences or filter which genuine reviews appear in a way that misleads. The table separates the common cases by the rule's own tests.
| Practice | Position under the FTC rule | Why | Main risk |
|---|---|---|---|
| Generate reviews from invented customers | Prohibited | The reviewer and the experience do not exist (465.2) | Civil penalties, platform bans |
| Buy a batch of reviews from a vendor | Prohibited if you knew or should have known they were false | 465.2 and 465.4 cover buying and spreading | Liability for the buyer, not only the seller |
| Have staff post reviews without disclosure | Prohibited if undisclosed or false | 465.2 and 465.5 cover insiders and their relatives | Manager liability for pressure on staff |
| Email all purchasers asking for a review | Allowed | Generalized solicitations are exempt | Sending only to happy customers may breach the FTC Act |
| Offer a reward for a review of any rating | Allowed if not conditioned on sentiment | 465.4 bars conditioning on positive or negative | Platform bans on incentives, such as Trustpilot and Google |
| Use an AI avatar in an ad | Not banned as such | The FTC says there is no blanket prohibition on AI avatars in marketing | Implying a real customer said it |
| Draft a public reply with AI, then review it | Not a review, so outside 465.2 | Replies are separate, but 465.7 limits false accusations and threats | A reply that falsely accuses a reviewer |
Replies, suppression and hiding reviews
Two points need care. First, the FTC's question-and-answer page says a business may respond publicly to a negative review as long as it does not make knowingly false accusations against the reviewer or use threats or intimidation. An AI-drafted reply that accuses a reviewer of being a competitor, with no evidence, is an accusation in the rule's sense. Second, sorting reviews by helpfulness or rating is not suppression, but a business may not misrepresent that the reviews it shows are most or all of those submitted when it has suppressed reviews on the basis of rating or negative sentiment. The rule lists exceptions for content that is unlawful, fake or unrelated to the product, and for criteria applied equally regardless of sentiment.
How platforms treat AI-written reviews
Platforms enforce faster than regulators, and their policies are the practical route for removal. They differ on AI specifically, so read the policy of the platform where the review sits. The statements below are as of the pages read for this guide.
- Yelp. Yelp's 2025 Trust and Safety Report says automated systems filtered nearly 500,000 reviews that showed characteristics of AI-generated content, and that AI-generated reviews violate its Content Guidelines. It also says reviewers may not use third-party AI tools to write reviews.
- Trustpilot. Its guidelines for reviewers list AI-generated content among the promotional material and misinformation that is not allowed in reviews. They also prohibit incentivized reviews and bar owners, family, employees, contractors, shareholders and competitors from reviewing the business. Reviews pass through automated detection before posting, and businesses can flag reviews they believe are not genuine.
- Google Business Profile and Google Maps. The prohibited and restricted content policy says contributions should reflect a genuine experience and that fake engagement is removed. It names paid reviews, multiple accounts controlled by one person and content posted through emulators or device tampering. The page read for this guide does not mention AI-written reviews, so the test is whether the content reflects a real experience.
- Amazon. Amazon says it proactively blocked more than 200 million suspected fake reviews across its stores worldwide in 2022, and that it uses large language models and graph analysis to look for anomalies. The page read does not say how customers report abuse.
Yelp's report also gives the outcome for all reviews it received, not only suspected AI ones.
Source: Yelp's 2025 Trust and Safety Report, released February 25, 2026. The figures are not limited to AI-written reviews.
How to tell whether a review was written by AI
You cannot prove it from the text alone, and neither can the detection tools sold for the purpose. What you can do is document patterns that point to inauthentic activity, which is what platforms act on. Google's policy itself lists unusual volumes or patterns of review activity as a sign of rating manipulation.
Related readGoogle Seller Ratings: Remove a Review or Fix the StarsPatterns worth recording, none of them conclusive alone:
- Timing. Many reviews in a short window, especially after a period with none, or arriving at regular intervals.
- Shared structure. The same sentence shapes, the same praise or complaint in different words, or the same unusual phrase across reviews of the same business.
- No specifics. No product, staff member, date or detail that a customer would know, or details that do not match what you actually offer.
- Account profile. New accounts with one review, accounts that reviewed many unrelated businesses in one day, or profile photos that are generated or reused.
- No matching customer. A reviewer who does not appear in your orders, bookings or records, where your records would normally show them.
A missing customer record is not proof, since people book through partners or use another name. Ongoing monitoring helps here because a dated record of normal activity makes an abnormal week obvious; our reputation monitoring work includes that kind of baseline.
How to document and report a suspected AI review wave
The aim is a report a platform reviewer can act on in one read: evidence organized by policy, not an argument about the reviewer. The sequence below usually fits a two-week window; platform responses vary from days to weeks.
- Capture before anything changesOn day one, take dated screenshots of every review, the reviewer profile, the rating and the review count, and save the page addresses. Reviews can vanish or be edited before you report them.
- Check your own recordsOn days one and two, match each reviewer against orders, bookings and messages, and note which have no match. Record the result as a plain table with dates.
- Map the patternOn days two and three, write down timing, shared phrases and account details in one sheet. Group the reviews by the policy they appear to breach, such as fake engagement or conflict of interest.
- Report through the platformOn days three and four, use the platform's own report route for each review and cite the specific policy, with the evidence sheet summarized in a few sentences. Trustpilot, for example, lets businesses flag reviews they believe are not genuine.
- Report the broker, if you have oneIf the reviews point to a vendor or a ring of accounts, say so in the report. Trustpilot has a community reporting form for businesses that ask for fake reviews or pressure reviewers.
- Follow up at day tenCheck what was removed, record what remains, and resubmit anything left with the new evidence. Do not report the same review repeatedly with identical text.
- Decide on escalationBy day fourteen, decide whether the pattern justifies counsel. The rule gives you no private right of action, so a lawyer would rely on other law.
Mistakes that make it worse
Do not reply by accusing the reviewers of being fake: the rule limits knowingly or recklessly false public accusations, and a wrong one against a real customer becomes its own story. Do not offer a reward for removing or revising a review, which Google's policy names as an incentive it does not allow. And do not answer a fake wave with one of your own.
How the US rule compares with the UK, EU and Australia
Each jurisdiction reaches fake reviews through a different instrument, enforced by a different body. The table summarizes what each source read says. It is information, not legal advice, and local counsel decides which rule applies to your case.
| Jurisdiction | Instrument | What it says about fake reviews | Enforcer and penalty |
|---|---|---|---|
| United States | FTC rule, 16 CFR Part 465 | Bans fake or false reviews, including AI-generated; hosting exempt | FTC; civil penalties for knowing violations; no private right of action under the rule |
| United Kingdom | Digital Markets, Competition and Consumers Act 2024 regime, in force April 2025 | Bans submitting or commissioning fake reviews and misleading publication; publishers must take reasonable and proportionate steps | CMA can decide infringements directly; fines up to 10% of global turnover |
| European Union | Unfair Commercial Practices Directive, as amended by Directive 2019/2161 | Lists false reviews and endorsements as unfair; requires information on whether and how reviews are verified | National authorities, under each member state's law |
| Australia | Australian Consumer Law | ACCC says creating or arranging fake reviews is against the law and that suppressing or editing negative reviews may mislead | ACCC; no AI-specific guidance found in the pages read |
The UK guidance treats buying reviews generated by software, such as bots, as commissioning banned reviews, and requires traders to take reasonable and proportionate steps to prevent and remove banned reviews. Its only reference to AI concerns AI-generated review summaries based on banned reviews, and it allows incentivized reviews if disclosed and genuine. In the EU, how the directive reaches a case depends on the member state's implementation, so EU cases need local advice. A December 2023 ACCC release found concerning conduct in 37 percent of 137 businesses reviewed, and the ACCC's business page on online reviews, last modified in July 2026, does not mention AI, so we state no AI rule for Australia. The UK fake reviews guidance is the one to read first for UK sellers.
Related readRemove a RateMDs Review, and Vitals or WebMD CareWhat to do next
If the pattern points to a competitor or vendor, start with the evidence sheet and the platform report, and leave legal escalation until the platform route has run. If you are considering AI tools for reviews, keep the line the rule draws: real customers, real experiences, disclosed relationships, no invented voices. Our own work follows that line, as set out on the ethics page.
If a wave of suspicious reviews is hitting your ratings, Reputation Partners can review the evidence, map it to each platform's policy and advise on the report order. Start with a private audit.
Questions people ask
Does the FTC rule apply if I only use AI to help write replies?
The rule's fake-review provisions concern reviews and testimonials, not business replies. But the suppression section prohibits threats, intimidation and knowingly or recklessly false public accusations when used to stop or remove a review. An AI-drafted reply is held to the same standard as one typed by a person, so someone should read it before it posts.
Can I sue a competitor under the FTC rule for AI fake reviews?
No. The FTC states the rule creates no private right of action, so enforcement belongs to the Commission. A business may have claims under other laws, which depend on the facts and the jurisdiction, so speak to counsel before assuming any route exists.
Is it legal to ask customers for reviews and give them a reward?
Under the FTC rule, an incentive is allowed if it is not conditioned on the review being positive or negative. Platforms are stricter: Trustpilot does not allow incentivized reviews, and Google's policy prohibits merchants from offering incentives for reviews. Check the platform before you send the offer.
Do platforms remove a review just because it looks AI-written?
Not on appearance alone. Platforms act on policy: Yelp and Trustpilot treat AI-generated reviews as violations, while Google's policy turns on a genuine experience. A report with dated evidence of patterns is usually stronger than one saying a review reads like AI.
Does the rule apply outside the United States?
No. It is a US federal rule. The UK, the EU and Australia have their own laws with different tests and enforcers, as the comparison above shows. A business selling in several countries must meet each market's rules and each platform's policy.



