4 September 2026

AI-generated fake reviews: how to spot them and what to do if your reputation is attacked

For years, a fake review gave itself away immediately: four badly spelled words, an invented name and no detail about the business. Not any more. With any AI tool you can generate twenty different reviews, well written, with plausible detail, in under a minute. And that changes the problem: what used to be an occasional nuisance can now arrive as an avalanche.

For a small business, reviews are a good part of the first impression. Many people decide which garage to take the car to, or where to have dinner, by looking at three things: the rating, the photos, and the two or three most recent comments. A batch of fake negative reviews does real damage, and the damage lands before any complaint does.

A case that keeps repeating

A restaurant receives eight one-star reviews over the same weekend. They are well written, they talk about "slow service" and a "careless menu", but none mentions a specific dish. The profiles are accounts with no photo, created recently, with a single review published. Two days later an email arrives: somebody offers to "clean up the venue's online reputation" for 300 euros a month.

It is not a coincidence. It is the classic pattern: first the attack, then the paid remedy. There is also the less aggressive version, where a competitor orders half a dozen negative comments and leaves it there.

The warning signs

Before worrying, check whether several of these signs coincide:

An odd concentration in time. Several negative reviews within a few hours, when you normally get one a fortnight.

No verifiable detail. No dish, service, date or team member is named. AI writes well, but it does not know what you served on Tuesday.

Empty or brand-new profiles. No photo, no history, with a single review published.

Text that is too polished and too neutral. A genuinely annoyed customer writes with typos, capital letters and very specific things. Generated text reads like a school essay.

Repeated structure. If you read five in a row and they all follow the same order of arguments, five different people did not write them.

Reviews of something that does not exist. Comments about a venue, an opening time or a service you do not offer.

The "rescuer" email that arrives right afterwards. Any offer to remove reviews in exchange for money is, at the very least, suspicious.

One sign on its own proves nothing: there are legitimately unhappy customers who write short and without detail. What sets off the alarm is a combination of three or four.

What to do if it happens to you

In the first hours: document. Take screenshots of every review, with the profile name and the date visible. If they disappear later, you will need evidence.

Do not reply while angry, but do reply. A brief, polite, factual comment ("We have no record of a booking or order under that name on that day; if you have been our customer, write to us and we will look into it") does two things: it invites clarification and it makes clear to the next reader that something does not add up.

Report them to the platform. Google, TripAdvisor and the rest have a process for reporting fake content. It is slow and does not always work first time, but a report showing the whole pattern (dates, empty profiles, repeated text) carries far more weight than an isolated complaint.

If there is extortion, report it. Asking for money to remove negative reviews that you planted yourself is not marketing: it is a crime. Keep the full emails, with headers, and take it to the police.

Dilute the noise. The best medium-term defence is a steady flow of real reviews. Asking happy customers naturally — with no incentives and without filtering by rating — means ten fake comments weigh much less on a healthy average.

And the flip side: do not buy reviews

AI also makes it cheap to generate positive reviews for your own business. Do not do it. Beyond the ethics, publishing or commissioning fake reviews is an unfair commercial practice, sanctionable in the European Union, and platforms are getting better at detecting the patterns: when they fall, they often take the whole business listing with them. A profile suspended for bought reviews does far more damage than the eight stars you were trying to gain.

Online reputation is a lot like the ordinary kind: it is built slowly and defended with evidence, not by shouting.