Explainers / FRAUD
Are AI-generated reviews allowed, and how many are being removed?
A review that misrepresents a real person's experience is unlawful under the FTC's 2024 rule whether a person or a language model wrote it. Yelp and Tripadvisor go further and remove AI-written reviews as a category. Yelp reports filtering nearly 500,000 in 2025.
Writing a plausible paragraph about a meal you never ate used to cost a few dollars of someone's time. It now costs nothing. That change in price is the whole story, and both the law and the platforms have responded to it.
The federal position
The FTC's final rule on reviews, announced in August 2024, bars creating or selling reviews that misrepresent they are by someone who does not exist or who did not have the experience described. The announcement names AI-generated fake reviews explicitly as covered.11 Federal Trade Commission, Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (August 2024)
The rule text does not mention software at all, and does not need to. Section 465.2 prohibits a business from writing, creating or selling a consumer review that materially misrepresents "that the reviewer or testimonialist exists", that the reviewer "used or otherwise had experience with the product, service, or business", or what that experience was.22 Code of Federal Regulations, 16 CFR 465.2, Fake or false consumer reviews, consumer testimonials, or celebrity testimonials A fabricated experience is a violation regardless of the tool.
What the rule does not do is ban assistance. A real customer who uses a chatbot to tidy the grammar of a true account is describing a real experience. Nothing in section 465.2 reaches that.
The Rytr case: liability for the tool
A month after the rule was announced, the FTC went after the supply side. Rytr sold an AI writing assistant that included a "Testimonial & Review" use case. According to the FTC, paid subscribers could generate unlimited reviews from minimal input, and the output contained specific, often material details that had no relation to what the user typed in. Some subscribers produced "hundreds, and in some cases tens of thousands" of reviews.33 Federal Trade Commission, FTC Announces Crackdown on Deceptive AI Claims and Schemes (September 2024)
The complaint charged Rytr with providing the means to deceive. The proposed order barred it from offering any service dedicated to generating consumer reviews or testimonials. The Commission vote was 3 to 2, and both dissenting commissioners wrote separately, so this theory of tool-maker liability should be read as contested inside the agency.
The alleged flaw is instructive for readers. A generator asked for a review of "a plumber" will invent the leak, the arrival time and the friendly technician's name. Those inventions are exactly the specifics that make a review persuasive.
Platform rules go further than the law
Yelp prohibits AI-drafted reviews outright. Its 2025 report says reviews "must reflect genuine, firsthand experiences", and may not be "content drafted or revised by third-party AI tools". Its automated systems filtered out nearly 500,000 suspected AI-generated reviews during 2025.44 Yelp Trust and Safety, 2025 Trust and Safety Report
That is a stricter line than the FTC's. A true account polished by a chatbot is lawful and still against Yelp's policy.
Tripadvisor reports that 214,000 AI-generated reviews were flagged and removed in 2024. Its head of trust and safety framed the policy around sameness: removing AI-written reviews "helps our users avoid a 'sea of sameness'".55 Tripadvisor, 2025 Transparency Report press release (March 2025)
Thumbtack addresses AI on the other side of the transaction. Its content policy bans "digitally altered or AI-generated media" used to "create a fake portfolio, misrepresent a property's condition, misrepresent work quality, or fabricate evidence".66 Thumbtack Help, Content and Review Policy That covers a pro's project photos and a customer's photos of alleged damage alike.
Google and Amazon do not single out AI in the policy pages we read. Google's requirement is that contributions "reflect a genuine experience"; Amazon's is that a review be the reviewer's own authentic experience.77 Google Maps Help, Prohibited and restricted content, Maps user-contributed content policy88 Amazon Customer Service, Community Guidelines Under those rules the test is truthfulness, not authorship.
How detection works, as far as anyone will say
Neither Yelp nor Tripadvisor explains how it identifies AI text. Yelp says only that its systems filter reviews that "exhibited characteristics of AI-generated content". Neither publishes an error rate. Our guess is that platforms lean on the same signals they use for all fraud, such as account history, posting patterns, shared networks and bursts of activity. That inference is ours; the platforms do not disclose their methods.
Two consequences follow.
First, the published counts are counts of flagged reviews. They say nothing about how many AI-written reviews were missed, or how many honest reviews were caught by mistake.
Second, for a reader, style is a poor guide, and we would not try to judge authorship by voice. What still works is everything that does not depend on prose: the reviewer's history, the timing, whether the details match the business, and whether photos accompany the text.
Does AI change how much to trust ratings?
It lowers the cost of one kind of fraud and leaves the others where they were. The largest reported category of fake reviews is still insiders boosting their own business, which accounted for 54% of fraud Tripadvisor detected in 2024.55 Tripadvisor, 2025 Transparency Report press release (March 2025) An owner who would previously have written six clumsy fake reviews can now produce sixty fluent ones, but each still needs an account, and accounts are where detection concentrates.
Yelp points to one structural defense. Because it requires text with every rating, and the average recommended review ran 446 characters in 2025, each rating comes with material that can be checked against other signals.44 Yelp Trust and Safety, 2025 Trust and Safety Report Platforms that accept a bare star tap have less to examine. Amazon handles the same weakness differently: a star rating without Verified Purchase status is not counted until the customer adds text, an image or video.99 Amazon Customer Service, Understanding Customer Reviews and Ratings
AI summaries of reviews are a separate thing
Several platforms now display machine-written summaries of what reviewers say. Nextdoor, for example, describes AI-powered search that summarizes neighbor knowledge about local businesses.1010 Nextdoor Blog, Introducing Local Faves A summary is the platform's own speech about the reviews, not a review. Its accuracy depends on the reviews underneath and on a model you cannot inspect. If a summary matters to your decision, read ten of the source reviews and see whether you would have written the same sentence.
Questions readers ask
Is it illegal to use AI to write a review of a place I did visit?
The FTC rule targets reviews that misrepresent the reviewer's existence or experience. A truthful account of a real visit is not that. The platform may still forbid it; Yelp's policy excludes content drafted or revised by third-party AI tools.44 Yelp Trust and Safety, 2025 Trust and Safety Report
Can platforms reliably detect AI-written reviews?
They report large removal counts and do not publish error rates. Without a false-positive figure, "reliably" cannot be assessed from outside.
Did the FTC ban AI review generators?
No general ban exists. In the Rytr matter the FTC obtained a proposed order against one service that offered review generation as a feature.33 Federal Trade Commission, FTC Announces Crackdown on Deceptive AI Claims and Schemes (September 2024)
Method Rule coverage is taken from the FTC's announcement and the rule text. Removal counts are the platforms' own reported figures and reflect what their systems flagged. See how we source and check.
Sources
Every source below was opened and read for this page. Links go to the publisher's own copy.
- Federal Trade Commission. Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (August 2024)
- Code of Federal Regulations. 16 CFR 465.2, Fake or false consumer reviews, consumer testimonials, or celebrity testimonials
- Federal Trade Commission. FTC Announces Crackdown on Deceptive AI Claims and Schemes (September 2024)
- Yelp Trust and Safety. 2025 Trust and Safety Report
- Tripadvisor. 2025 Transparency Report press release (March 2025)
- Thumbtack Help. Content and Review Policy
- Google Maps Help. Prohibited and restricted content, Maps user-contributed content policy
- Amazon Customer Service. Community Guidelines
- Amazon Customer Service. Understanding Customer Reviews and Ratings
- Nextdoor Blog. Introducing Local Faves