You’re about to buy something, you check the reviews, and something feels… off. They’re all glowing, all vague, all weirdly similar. Your instincts might be right. So how can I tell if a product review online was written by AI? Look for the tells: generic praise with zero specifics, suspicious patterns across reviews, and language that sounds polished but says nothing. Here’s your field guide to spotting the fakes.

The Tells: Language Patterns

AI-generated reviews have recognizable stylistic fingerprints:

  • Vague enthusiasm, zero specifics: “This product exceeded my expectations in every way!” — but never mentions what the product does, how it was used, or any concrete detail. Real reviewers mention specifics: “the battery lasted through my 14-hour flight,” “the medium fits my 6’1″ frame with room to spare.”
  • Suspicious polish: perfect grammar, balanced structure, marketing-adjacent vocabulary. Real reviews are messier — typos, fragments, ALL CAPS excitement, weird tangents.
  • Generic feature lists: the review reads like the product description reworded. “Its ergonomic design and premium materials make it a must-have” could describe literally anything.
  • Emotional flatness: AI struggles with genuine idiosyncratic voice. Real reviews have personality — humor, frustration, oddly specific complaints (“the zipper sticks when it’s humid”).
  • Absence of negatives: even beloved products have quirks. A wall of pure five-star praise with no “wish it came in blue” or “takes getting used to” is suspicious.

The Tells: Pattern Clues

Zoom out from individual reviews and look for structural red flags:

  • Review velocity spikes: 200 five-star reviews appearing in three days for an obscure product. Organic reviews accumulate gradually.
  • Reviewer profiles: accounts with only one review, generic names, no profile history, or dozens of five-star reviews posted the same week across unrelated products.
  • Repetitive phrasing: multiple reviews using the same unusual phrases or sentence structures — a sign of batch generation from similar prompts.
  • Mismatched details: reviews mentioning features the product doesn’t have, or describing a different product entirely — classic AI hallucination.
  • Extreme rating distribution: all five stars and a few one stars, nothing in between. Real products earn a bell curve of opinions.

Tools and Tactics for Verification

  • Read the 3-star reviews first. They’re the most honest — balanced enough to mention pros and cons, detailed enough to be useful, and rarely faked (nobody pays for mediocre fake reviews).
  • Check third-party review analysis: tools like Fakespot (now part of Mozilla) analyze review patterns for manipulation signals. Browser extensions can flag suspicious listings automatically.
  • Look for “verified purchase” badges — not foolproof, but a basic filter.
  • Search the product on Reddit or forums for unfiltered human opinions. Enthusiast communities are brutally honest.
  • Check multiple retailers. A product with 5,000 glowing Amazon reviews and crickets everywhere else deserves skepticism.
  • AI detectors exist but are unreliable — don’t treat a “likely AI” score as proof. Use them as one signal among many.

Why This Problem Is Getting Worse

AI review generation is industrializing: sellers buy bulk fake reviews, competitors buy fake negative reviews, and the tools get better monthly. Platforms are fighting back with detection systems and lawsuits, but it’s an arms race. The deeper issue is economic — reviews drive the majority of purchase decisions, so the incentive to fake them is enormous. Until platforms solve verification at scale, consumer skepticism is the main defense. (The same synthetic-content flood is hitting news articles too.)

The Fake Review Economy: Who’s Behind It

Understanding the incentives helps you spot the manipulation. Sellers buy positive reviews through review farms, “rebate for review” schemes (buy the product, get refunded for a 5-star review), and AI generation services that produce hundreds of unique-sounding reviews for pennies each. This is most common with new products fighting for initial traction and with sellers in hyper-competitive categories.

Competitors buy negative reviews — fake 1-star reviews targeting rivals’ products. Harder to spot because they mimic genuine complaints, but they often cluster in time and lack the specific detail of real grievances. Platforms fight back unevenly: Amazon sues review brokers and uses AI detection, but the sheer volume overwhelms enforcement. Smaller marketplaces often lack meaningful defenses entirely. And review influencers complicate the picture further — “honest reviews” from creators who received free products are technically disclosed (when they are), but the incentive to stay positive for future freebies is obvious.

The FTC has strengthened rules against fake reviews, with real penalties for sellers caught buying them. But enforcement lags the problem’s scale. Until verification improves structurally, your skepticism remains the most effective filter — which is why learning the tells matters more than waiting for platforms to fix it.

Building your personal review-reading system

Make this a habit, not a chore. Step 1: check the overall rating distribution — healthy products show a curve, not a cliff. Step 2: read 5–10 three-star reviews for the balanced picture. Step 3: skim the most recent 1-stars for deal-breaker patterns (is everyone complaining about the same defect?). Step 4: check one independent source — Reddit, a forum, a trusted reviewer — for unfiltered opinion. Step 5: for expensive purchases, wait 48 hours after research before buying; manufactured urgency (“only 3 left!”) is often paired with manipulated reviews. Total time: 10 minutes. That small investment has saved me from bad purchases more times than I can count — and it’s the difference between shopping and being shopped.

How to Write Reviews AI Can’t Fake

The best defense against fake reviews is more real ones — and writing genuinely helpful reviews is a skill. Be specific: “the battery lasted 11 hours of video calls” beats “great battery life.” Specifics are hard to fake and easy to verify — they’re the hallmark of real experience. Mention your use case: “I’m a 6’2″ runner using these for marathon training” gives readers context AI can’t invent convincingly.

Include the negatives: no product is perfect, and balanced reviews are inherently more credible. “Love the sound quality; the app is buggy and the case scratches easily” reads as real because it is how real people experience products. Pure five-star gushing triggers everyone’s fake detector — including the human reader’s. Update over time: “six-month update: the zipper broke, customer service replaced it free” — longevity information is the most valuable and least faked review content.

Photos help enormously: a real photo of the product in your home, showing wear or scale, is nearly impossible to fake convincingly and instantly boosts credibility. And write for the next buyer, not the seller: what do you wish you’d known before purchasing? That question produces the details that matter — sizing quirks, setup difficulty, the thing that broke first. Every honest, specific review you write makes the ecosystem slightly more trustworthy. In the age of AI-generated everything, your genuine experience is a small act of public service.

The Takeaway

Spot AI-written product reviews by looking for vague generic praise, suspicious polish, pattern anomalies (velocity spikes, thin reviewer profiles, repetitive phrasing), and mismatched details — then verify with 3-star reviews, third-party analysis tools, and independent forums. No single tell is definitive, but three or four together are damning. When in doubt, trust the detailed, slightly grumpy, imperfectly written reviews — those are the humans. Consumer Reports and the FTC’s endorsement guidelines are worth knowing as a baseline for what honest reviewing looks like.

AI-generated text is everywhere now — it helps to know the biggest misconceptions regular people have about AI so you can judge new tools clearly.