How do you spot AI generated reviews?
You spot AI generated reviews by zooming out. Knowing how to spot AI generated reviews is less about catching one suspicious sentence and more about seeing a pattern: many reviews that sound alike, appear together and come from accounts with little history.
Keep two questions apart. Was this review written with AI help? And is it fake, meaning not based on a real customer’s experience? A real buyer can use AI to tidy up their grammar. A fake review can be written by a human. What matters to you as a shopper is the second question.
| Method | What it catches | Main limit |
|---|---|---|
| Reading for language signals | Generic, template-like AI written reviews | Many honest reviews are generic too |
| Pattern analysis | Bursts, duplicates and rating spikes | Needs many reviews to see a pattern |
| Reviewer history | Accounts that only post five-star reviews, or post in batches | Some platforms show little history |
| Off-site comparison | Ratings that disagree sharply across sites | Takes time; not every product is reviewed elsewhere |
| AI text detector | Statistical traces of generated text in longer reviews | Needs at least 300 characters; short reviews are hard |
What language gives away AI written reviews?
AI written reviews tend to sound polished, balanced and oddly empty. They describe the product in general terms anyone could write without owning it, and they rarely include the small, specific details that come from real use.
- Generic praise: “exceeded my expectations”, “a game-changer”, “highly recommend to anyone looking for quality”.
- Product-page language: the review repeats feature names and marketing claims from the listing.
- Tidy structure: an introduction, three neat points and a summary sentence, like a short essay.
- Balanced hedging: a mild drawback mentioned and immediately dismissed, which reads as staged rather than felt.
- No checkable specifics: no sizes, dates, how long they used it, what went wrong or how support responded.
- Leftover artifacts: phrases like “as an AI language model”, placeholder brackets or a product name that does not match the listing.
Compare that with what real reviews usually contain: a reason for buying, a concrete use, a surprise, a photo taken at home, sometimes a typo or a complaint about delivery. None of these signals is proof alone. Our guide on how to tell if text is AI generated covers the language tells in more depth.
Example: a generic review versus a specific one
These two invented reviews of a coffee grinder show the difference. The first could have been written by anyone who read the product page. The second could only come from someone who used the grinder.
| Generic, possibly AI written | Specific, likely from real use |
|---|---|
| “This grinder exceeded my expectations. Its sleek design fits seamlessly into any kitchen, and the consistent grind elevates my morning routine. While it is slightly louder than expected, the quality more than makes up for it. Highly recommend!” | “Bought it in March to replace a blade grinder. Setting 12 works for my French press, but anything finer than 6 clogs the chute and I have to tap it. Loud enough to wake my partner at 6 a.m. Still using it daily after four months.” |
The specific review names dates, settings, a real annoyance and how long the writer has used the product. Those are details a reader could disagree with, and that is what makes them believable.
What patterns reveal fake reviews AI tools produce in bulk?
Bulk fake reviews leave patterns across the review section that no single review shows. Sort reviews by newest, scroll through several pages and look for clusters.
| Pattern | What it may mean |
|---|---|
| Many reviews posted on the same day or week, then silence | A paid or automated batch |
| Several reviews share an unusual phrase or the same structure | One template or one prompt reused |
| Ratings jump from mixed to almost all five stars | A campaign started, or negative reviews were buried |
| Reviews praise features the product does not have | Written from a description, not from use |
| Reviews mention a different product, color or model | Reviews merged from another listing or generated without care |
| Very few three-star reviews, mostly fives and ones | Real buyers often land in the middle; check the ones |
Read the one- and two-star reviews closely. If several unhappy buyers mention the same defect while the five-star reviews never address it, trust the specific complaints over the generic praise.
Then compare across sites. If a product or shop has glowing reviews in one place and a stream of complaints about delivery or refunds on independent forums, the gap itself is a signal. Look for the same model sold by other sellers, too: reviews that only appear on one listing, and nowhere else, deserve less weight.
How do you check a reviewer’s history?
You check a reviewer’s history by opening their profile and reading what else they have reviewed and when. A real customer reviews a scattered mix of things over months or years; an account built for fake reviews tends to look narrow and recent.
- 01Open the reviewer’s profile, where the platform allows it.
- 02Look at how many reviews they posted and over what period. Many reviews in a few days is a warning sign.
- 03Check the ratings they give. Almost only five stars, across unrelated products, is suspicious.
- 04Compare their reviews with each other. The same style and length on everything suggests a template.
- 05Check for a verified purchase badge or equivalent where the platform offers one.
- 06Look at their profile photo; a generated face is another signal, and our guide on how to spot a fake profile picture shows what to check.
Are Amazon reviews AI generated?
Some text on Amazon review pages is AI generated on purpose, and that is different from fake reviews. Amazon shows AI-generated review highlights, a short summary of what customers often mention, which it introduced in 2023. Amazon says the highlights draw only on reviews from verified purchases.
Individual reviews are another matter. Amazon says it strictly prohibits fake reviews and uses machine learning and investigators to stop them, but no system catches everything on any large marketplace. Use the same checks: patterns, reviewer history, verified-purchase badges and the low-star reviews.
Sellers and platforms facing this problem at scale will find a fuller workflow in our page on marketplaces and e-commerce.
Are AI generated fake reviews illegal?
Fake reviews, including AI-generated ones, are now explicitly banned in the United States and the United Kingdom. The rules target businesses that create, buy or sell fake reviews; they do not stop a real customer from using AI to word an honest review.
The FTC rule on fake reviews (United States)
In August 2024, the Federal Trade Commission announced its Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, which took effect in October 2024. The FTC’s announcement says the rule aims to deter AI-generated fake reviews. It bans creating, selling or buying fake reviews, including reviews that misrepresent that the reviewer exists or used the product.
The rule also covers paying for reviews conditioned on a positive or negative sentiment, undisclosed insider reviews, suppressing negative reviews with threats, and buying or selling fake social media indicators such as bot followers.
The DMCC Act ban on fake reviews (United Kingdom)
In the UK, the Digital Markets, Competition and Consumers Act 2024 lists fake reviews as a banned practice from 6 April 2025. It defines a fake consumer review as one that claims to be, but is not, based on a genuine experience. It also requires businesses that publish reviews to take reasonable steps to prevent and remove fake ones. The Competition and Markets Authority explains the rule in its guidance CMA208.
What this means for you
If you suspect a business is buying fake reviews, report the listing to the platform first. In the US you can also report to the FTC at ReportFraud.ftc.gov. A detector result can support a report as one signal among several, but it is not proof on its own.
How do you batch-check reviews with an AI text detector?
You batch-check reviews by grouping them into longer texts the detector can work with, then reading which passages it highlights. Short reviews are hard for any detector, because there are too few words to measure reliably.
The AnyFact AI text detector needs at least 300 characters in English, French or Arabic. It uses the Fast-DetectGPT method with an open-source language model and scores each passage of two to three sentences. A text that is only partly generated shows up as mixed, with the passages highlighted.
- 01Collect the reviews you suspect, each with its date, rating and reviewer name.
- 02Check longer reviews (300 characters or more) one at a time.
- 03For shorter reviews from the same suspected batch, paste several into one text, separated by line breaks, and read which passages are highlighted.
- 04Run a few reviews you trust, such as detailed critical ones, as a comparison.
- 05Record the verdict and the signals for each check, not just the label.
Read a combined check carefully. A “mixed” result tells you which passages look generated, which helps you see whether one template runs through several reviews. It does not tell you who wrote them or whether the experience was real.
AI Text Detector · Free plan
Check longer reviews for AI writing: each passage is scored, and partly generated text shows up as mixed with highlights.
What are the limits of detecting AI written reviews?
The main limit is that style is not honesty. An AI text detector can suggest that wording looks generated, but it cannot tell whether the writer bought the product, and it can be wrong in both directions.
| Case | Why it is hard |
|---|---|
| One-line reviews | Too short for any reliable text analysis |
| Honest reviews polished with AI | Real experience, generated wording |
| Non-native writers | Simple, regular phrasing can look machine-made to detectors |
| Human-written fake reviews | Fake content, human style |
| Translated reviews | Machine translation changes the statistical profile of the text |
So weigh the detector with the pattern and reviewer checks, and never sanction a reviewer on a detector result alone. Our guide on whether AI detectors are accurate explains the limits in detail. To check whether a whole shop looks generated, see how to tell if a website is AI generated.
AI Website Checker · Free plan
Not sure about the shop itself? Check up to 20 pages for AI text and images, plus domain age and Safe Browsing where available.
How to spot AI generated reviews: a step-by-step checklist
Run this checklist before you trust a rating enough to buy, book or recommend. It works on any platform, and most steps need nothing more than a few minutes of scrolling.
- 01Sort reviews by newest and look for bursts of posts on the same days.
- 02Read five or six five-star reviews for generic praise, product-page language and missing specifics.
- 03Look for repeated phrases or identical structure across different reviewers.
- 04Open two or three reviewer profiles and check their age, volume and rating habits.
- 05Read the one- and two-star reviews for specific, repeated complaints.
- 06Compare the rating with reviews on independent sites and forums.
- 07Check longer suspicious reviews, or a group of short ones, with an AI text detector.
- 08Decide on the overall pattern. If the signals disagree, give the reviews less weight rather than none.
Write down what you found if you plan to report the listing. A note such as “twelve five-star reviews posted in two days, four with the same opening sentence, reviewers created this month” helps a platform act far more than a complaint that the reviews “seem fake”.
Frequently asked questions
Can you detect AI written reviews reliably?
Not one by one, especially short ones. Patterns across many reviews, reviewer history and a detector on longer texts together give a solid indication. No method gives proof, and detectors can be wrong in both directions.
How short is too short for an AI text detector?
The AnyFact text detector needs at least 300 characters. Below that, there is too little text to measure reliably. Group several short reviews from a suspected batch into one text, or rely on pattern and reviewer checks instead.
Is using AI to write a review illegal?
Using AI to word your own honest experience is not what the rules target. The FTC’s 2024 rule and the UK’s DMCC Act ban fake reviews, meaning reviews not based on a genuine experience or that misrepresent who wrote them.
Are Amazon review summaries AI generated?
Yes. Amazon shows AI-generated review highlights that summarize what customers often mention, and says they draw only on verified-purchase reviews. That summary is a feature, not a fake review.
What is the fastest way to spot fake reviews AI tools wrote?
Sort by newest and look for bursts of similar five-star reviews, then open two or three reviewer profiles. If they are new, review only in batches and use the same phrasing, the reviews deserve little weight.
Where can I report fake reviews?
Report them to the platform first, using its report option on the review or listing. In the US you can also report to the FTC at ReportFraud.ftc.gov; in the UK, the Competition and Markets Authority enforces the DMCC Act rules.