The fake review checker below reads a single review's text and scores the risk signals it finds: repetitive phrasing, superlative density, generic praise with no concrete details, authenticity claims embedded in the text. Paste any review from Amazon, Trustpilot, Google, Shopify, Tripadvisor, or your own store. The instant signals run in your browser. If you use the AI cross-check, the review text (and the website URL, if provided) is sent to our server and scored by an AI model. We do not store the pasted text: no account, no review archive. We keep only a usage record per check (token count and cost), with no text or IP address. For a manual method, see how to check if reviews are fake.
Check any review
Instant pattern signals in your browser. Optional AI cross-check compares the review against the business website.
Heuristic signals, not proof. A clean score doesn't guarantee a real review. See below.
Adding the site enables the strongest check: the AI compares what the review claims against what the website actually sells.
Instant heuristics run in your browser. The optional AI cross-check sends the pasted text (and URL, if given) to our server for scoring by an AI model; the text is not stored. Automated estimates, not proof. See how to spot fake reviews and the verification spectrum.
How the fake review checker works
The checker scores six writing-pattern signals and credits two opposite ones:
- Generic praise: "great product", "highly recommend", "fast shipping" with no supporting specifics.
- Superlative density: "amazing", "perfect", "game changer", "best purchase" as a large share of total words.
- Repeated phrasing: the same three-word sequence repeated within one review, a hallmark of spun or templated text.
- No concrete detail: no product specifics, dates, numbers, or first-person purchase language ("I ordered…").
- Authenticity claims in the text: "100% real", "not a bot", "verified buyer here".
- Exclamation density and ALL-CAPS emphasis: classic paid-review over-performance.
It also credits the opposite signals: concrete detail, and balanced tone: a real review usually carries at least a small reservation ("only", "slightly", "but").
The AI cross-check
Optional, and the stronger half of the tool. The pasted review text, plus the website text when you add a URL, is scored by Jev (TypeSafe's System One model) on three questions: does the review describe what the site sells (0 to 4), does the writing read like a real customer (0 to 4), and how likely is the text machine-written (0 or 1). The three combine into a single 0 to 100 score.
An honest limit: an AI cross-check is evidence, not proof. Determined humans can write reviews that score clean, and a genuinely enthusiastic customer can score suspicious. Verification at the source remains the only structural defense.
What a text checker can't prove
Writing analysis is a filter, not a verdict. AI-generated fake reviews are now virtually indistinguishable from human writing in blind tests, and a genuinely concise happy customer can look identical to a seeded one. The checker answers "does this text look like typical fake-review writing?". It cannot answer "did this person actually buy the product?"
That second question is the one that matters, and it has a structural answer: a review can only be certain to come from a real customer when an independent party attests the purchase. Under the FTC's Trade Regulation Rule on Consumer Reviews and Testimonials (16 CFR Part 465), fake reviews carry civil penalties, and the only reviews that make fabrication structurally irrational are transaction-verified ones: every fake attempt costs a real payment-processing fee. See transaction verified reviews for the full mechanics.
Frequently asked questions
How accurate is this fake review checker?
It surfaces the writing patterns most common in fake and template-generated reviews, and it is deliberately conservative: it reports signals with explanations rather than a definitive yes/no. Treat a high score as a reason to look closer (at reviewer history, timing, and the verification source) not as proof the review is fake.
Does this work for Amazon, Trustpilot, Google, or Shopify reviews?
Yes. The checker analyzes pasted text, so it is platform-agnostic: copy the review text from any listing, platform, or email, paste it, and run. It does not read live pages or fetch reviews itself. The instant signals run in your browser; if you use the AI cross-check, the pasted text is scored on our server and not stored.
Can it detect AI-written fake reviews?
Partially. AI-generated review text often trips the generic-praise and no-concrete-detail signals, but modern AI can also write specific, balanced prose that passes a text check. The AI cross-check adds a machine-written likelihood score, which catches many templated and LLM-written reviews, but no text-based method is proof. Verification at the source (who attests the purchase) is the durable defense.
What's the difference between checking text and verifying a purchase?
Text checking estimates likelihood from writing style. Purchase verification confirms fact from a payment record: an independent processor such as Stripe attests that a real charge occurred between the reviewer and the business. The former is a screen; the latter is proof. Verified vs unverified reviews explains the levels in detail.
Is the fake review checker free?
Yes: completely free, no signup. The instant signals run in your browser. The AI cross-check is also free (rate-limited per address). We do not store what you paste.
What happens to the text I paste?
We do not store what you paste. The instant signals never leave your browser. If you run the AI cross-check, the review text and optional website URL are sent to our server, scored, and discarded. We do not keep a copy or use it for training. We do keep a usage record per check (token count, cost and duration), which holds no text, website or IP address.
Collect reviews from real, paying customers →
Related: How to check if reviews are fake (7-point method) · Fake review statistics 2026