Start with the distribution, not the average
The average rating is the least informative number on the page. The distribution is where the truth lives, and two shapes matter. The first suspicious shape is the barbell: clustering at five stars and one star with a thin middle. A polarized distribution can be legitimate — some products genuinely divide people — but it is also the signature of review campaigns, because paid reviewers write five stars and angry victims write one star, while the quiet satisfied middle never bothers. The second shape is the wall: a product with hundreds of reviews, every one of them four or five stars, and nothing else. Real products disappoint someone. Uniform delight across hundreds of buyers is not a product quality; it is a curation policy.
Also read the volume over time. Reviews arriving in a tight burst — thirty in one week for a product with no history — usually means a campaign, a relaunch, or a seller who just changed the listing. Steady trickles over years are harder to fake and mean the product has actually been in circulation.
And check the ratings count against the sales reality of the category: a specialized tool with 40,000 reviews deserves a raised eyebrow. Category context is the cheapest lie detector there is.
The language of fakes
Fake reviews have a grammar, and once you have read it a few times you cannot unsee it. They praise the product category rather than the specific product: "this is the best purchase I have ever made, amazing quality, highly recommend" could describe a lamp or a lawn mower. They speak in adjectives instead of facts: amazing, perfect, incredible, love it — with no detail that could only come from ownership. They are often oddly formal or oddly generic about use: "I use this every day in my home." And they frequently review the speed of delivery, the packaging, or the seller's service when the question is supposed to be about the product — because the person never opened the box.
Real reviews have texture instead. They include numbers and circumstances: "the zipper snagged on the third week," "worked fine for video calls, mic is weak in a noisy room," "bought it for my mother who is 5'2"." They contain mild annoyances, because nothing is perfect and the people who lived with a product know exactly where its seams are. And they often say who the product is not for, which is the single most useful sentence a review can contain and the one fake reviews never write.
One more tell, easy to check: identical or near-identical phrasing across reviews, especially for the same product from different "buyers." Sellers who buy reviews buy them in batches, and batch writing repeats.
The sweet spot: why 4.2 beats 4.9
Counterintuitive but reliable: products with ratings in the 4.1 to 4.6 band are usually more trustworthy than products at 4.9. A 4.9 means either a genuinely exceptional product or a managed review section — and exceptional products are rare enough that the odds favor the management. The 4.2 product with a visible mix of five, four and three stars, whose negative reviews describe understandable limits ("great for beginners, too basic for daily professional use"), is showing you honest data about a real product.
Read the three-star reviews first, deliberately. They are the most honest rating on any platform: the reviewer was neither angry enough to distort nor delighted enough to gush. Three-star reviews read like field reports — what worked, what did not, who the product suits. If a product's three-star reviews sound reasonable and its failure modes are ones you can live with, you have learned more from thirty seconds of them than from an hour of five-star prose.
Incentives change what people say
Every review comes from someone with an incentive, and the incentives form a spectrum. At the honest end: the unasked reviewer, who spent their own money and is writing from memory and mild obligation. In the middle: the incentivized reviewer — discounts, free products, contest entries for reviews. These reviews are not worthless, but they skew positive because the incentive arrives after the product does. Platforms mark many of them; take the marking seriously and mentally downgrade praise, while noting that negatives from incentivized reviewers are unusually credible — nobody's incentive favors criticism.
At the other end sit the professionals: paid reviewers, affiliate writers, and sellers managing their own listings. The affiliate tell is a link or a glowing comparison that always ends with one product winning. The seller tell is the deleted review: sort by most critical and see whether the critical reviews exist at all. A listing where every one-star review is suspiciously about shipping is a listing with a filter.
The takeaway is not cynicism — it is weighting. Unincentivized negative reviews and detailed positive reviews both count for a lot. Generic praise counts for almost nothing, no matter how numerous.
Where honest feedback actually lives
If you want reviews that are hard to fake, go where there is no payment at stake. Niche communities and forums, where the same faces talk for years and a fake account gets noticed in an hour. Video reviews that show the product being used by hands that are not in a studio. Long-term updates — "six months later" comments are almost never faked, because fakes are written the week the campaign runs. And friends who own the product: one honest sentence from someone who knows you beats a hundred stars, because they know your actual life, not the average buyer's.
This is also why we show community votes alongside every product in our catalog — repeated, small, unsolicited signals are harder to buy than a single polished review. When you can, look for the accumulated pattern rather than the perfect testimonial.
A ten-minute review audit
Compressed into a routine: check the distribution, not the average. Sort by most critical and confirm critical reviews exist and were not filtered. Read five three-star reviews. Scan for adjective-only praise and identical phrasing. Check when reviews arrived. Search the community web for the product name plus "disappointed" or "broke." Ten minutes, and it converts a wall of stars into actual information.
Fake reviews are not going away, but they do not need to worry you. They follow scripts, and scripts are easy to recognize once you have decided to look. The honest signal was never hiding — it was just quieter, which is, in the end, how honest things usually sound.