AI-generated text has made fake reviews harder to spot than ever. A 2024 Fakespot study found that AI-written reviews are virtually indistinguishable from human-written ones in blind testing. The old tells — broken English, repetitive phrasing, generic praise — no longer work when each AI-generated review is grammatically perfect and stylistically unique.
But you can still spot unreliable reviews. You just need to look at different signals. Here's how.
# The signals that still work
# 1. Check the verification source, not the badge
A "Verified Buyer" badge tells you the review was verified — but not by whom. The real question is: who attested that this person bought the product?
- "Verified" by the merchant (Level 3) means the business said this person was a customer. Most platforms operate here — Trustpilot (invited), Yotpo, Judge.me, most Shopify apps.
- "Verified" by an independent payment processor (Level 4) means a third party — Stripe — independently confirms the charge. Only a handful of platforms operate here.
A badge is not proof. The source of the verification is the proof. If the platform doesn't make this distinction clear, ask why. Full verification spectrum breakdown →
# 2. Look at the review distribution
Real review distributions follow predictable patterns:
- A mix of ratings, typically clustered around 4.0–4.5 for good products
- Some 1-star reviews (even excellent products have shipping mishaps, defective units, wrong expectations)
- Reviews spread over time, not clustered in bursts
Warning signs:
- Too many 5-star reviews. A product with 500 reviews and a 4.99 average is statistically improbable for anything with more than a few hundred customers.
- Review bursts. 50 reviews posted on the same day, then silence for three months. Organic review velocity is relatively smooth.
- No negative reviews at all. Even the best products have at least a few legitimate complaints.
# 3. Read the 2-, 3-, and 4-star reviews
Fake review operations overwhelmingly target 1-star (attack competitors) and 5-star (boost products). The middle ratings — 2, 3, and 4 stars — are disproportionately likely to be genuine. They reflect real customers with mixed experiences: "Product works but setup was confusing," "Good value but shipping was slow."
If a product has hundreds of 5-star reviews and zero 3-star reviews, the 5-star reviews are suspicious — not because they're individually fake, but because the distribution doesn't reflect how real customers behave.
# 4. Check reviewer history
On platforms that show reviewer profiles, look at:
- Review velocity: A reviewer who posted 15 reviews in one day across unrelated product categories
- Review similarity: The same review text posted for different products
- Rating pattern: All 5-star or all 1-star reviews — real customers have varied experiences
- Account age: An account created last week with 30 reviews is more suspicious than a 5-year-old account with 30 reviews
One or two suspicious signals isn't proof. A cluster of 3+ is a strong indicator.
# 5. Watch for AI language patterns
AI-generated reviews have gotten very good, but they still have subtle tells:
- Uniform sentence length. Humans vary sentence length naturally. AI tends toward consistent 15–25 word sentences.
- No specific details. "This product is great, highly recommend" vs. "The zipper broke after three weeks but customer service sent a replacement in two days." AI is bad at invented specificity.
- Perfect grammar, no typos. Real reviews contain occasional typos. A product with 200 grammatically flawless reviews and zero typos is statistically unlikely.
- Generic enthusiasm. "Changed my life!" "Best purchase ever!" "You won't regret it!" — real reviews are more measured, even when positive.
# 6. Cross-reference with return/refund rates
If you're a business evaluating a competitor's reviews: products with unusually high review counts relative to their market position, coupled with unusually high average ratings, often have manufactured review profiles. The economics of fake reviews make this pattern easy to produce and hard to hide.
# For businesses: how to show your reviews are real
Consumers are getting more skeptical. The best defense is transparency:
# Show your verification chain
Don't just say "verified." Show the chain: purchase → payment processor confirmation → invitation → review → cryptographic signature. The more links in the chain the consumer can verify independently, the more trustworthy each review becomes.
# Publish your review policy
Make it easy to find: how do you verify reviews? Who can leave one? What do you do about refunds? What's your incentive policy? A transparent review policy is a trust signal — and, under the FTC's 2024 rule, it's increasingly a compliance requirement.
# Display verified and unverified reviews separately
If you collect reviews from multiple sources, don't mix them. A verified review loses its trust value when lumped in with unverified ones. Make the distinction visually clear.
# Never remove negative reviews (unless they violate policy)
A 4.5-star average with some thoughtful 2-star reviews is more credible than a 5.0 average with no complaints. Negative reviews that are addressed professionally show that you stand behind your product and care about customer experience. They're a conversion asset, not a liability.
# The bottom line
AI is winning the detection arms race. The only durable solution is structural: verify reviews against an independent source at the point of collection, not after. As a consumer, ask who verified the review — not just whether it's verified. As a business, choose a verification method that makes fake reviews impossible, not just detectable.
Further reading: Fake Review Statistics 2026 · The Fake Review Problem · What Does "Verified Buyer" Actually Mean?