Five-star reviews everywhere: how to judge a product

The short version

Five stars became wallpaper. On Amazon, Google, Booking, or the App Store, the average rating helps — and it’s gameable: fake reviews, discounts-for-stars, selective deletion, organized spikes. You don’t need to become an investigator. You need a reading method that works in two minutes.

Why ratings mislead

• Asymmetry: the thrilled and the furious write more than the middle
• Incentives: “leave 5 stars, get a coupon” (banned under many rules, still common)
• Bots and review farms: sudden volume, clone phrasing
• Selection: some sellers mainly solicit happy customers
• Product age: a great 2019 score says little about 2026 quality (especially electronics)

The FTC and EU consumer authorities have targeted deceptive reviews for years — a sign the problem is mainstream, not niche. Platforms remove some fakes; they can’t read your specific needs. That’s why a human skim of mid-range reviews still beats trusting the big yellow number alone.

The 5-signal checklist method

Before you buy, mark each signal green / orange / red. Three oranges or one strong red = pause.

#	Signal	Green (reassuring)	Orange / red (slow down)
1	Volume	Hundreds / thousands of reviews	48 reviews at 4.9★ on a “viral” product
2	Distribution	Slightly “normal” curve, real 3s and 4s	Mostly 5s and 1s, almost no middle
3	Recency	Useful reviews in the last 3–6 months	Great old score, recent silence, or sudden spike
4	Visual proof	Customer photos / videos (even blurry)	Studio-only; or images that don’t match the item
5	Mid-range (2–4★)	Usage detail, clear limits	Clone text, wrong product, zero detail

How to run it in 90 seconds

1. Average rating + review count (signal 1)
2. Check the star histogram (signal 2)
3. Sort by most recent (signal 3)
4. Open 2–3 customer photos (signal 4)
5. Read 3 mid-range reviews, not best / worst (signal 5)
6. Bonus: one off-site source + seller age / ship-from country

Fake-review cluster

• All posted in a 48-hour window
• Generic lines (“Great product! Fast shipping!”) with no detail
• Same phrasing / typos; profiles with no history
• Perfect stars + text about a different product
• Rating explosion right after a marketing price cut

No single tell convicts. A cluster does. If the product is expensive or safety-related (bike helmet, car seat, appliances): prioritize standards and tests, not stars. For clothing and furniture, prioritize fit/size complaints over “love it!!!” praise. For software, scan for billing and cancellation pain before feature wishlists.

What stars will never tell you

• Whether you, with your use case, will be happy
• Whether support will exist in 18 months
• Whether the “sale” price is actually low
• Whether the listing is a marketplace / refurbished disguise

Reviews are a signal. Not a verdict. Use them to narrow options, then confirm with specs, return policy, and — when it matters — a trusted test. Same caution when a price looks too good or a sale banner shouts: see deal scam red flags and how Black Friday pricing works.

Be stricter for edge cases: health and child safety (standards and recalls first), gadgets (exact generation/model), cosmetics and supplements (miracle photo transformations), and apps (high store scores can hide aggressive ads or trap billing — read reviews that mention charges). Choosing between two close products? Pick the one whose mid-range reviews describe your use case, not the one with an extra 0.1 stars.

Going further

• FTC — consumer reviews: rules and enforcement on fake reviews.
• EU — consumer rights: European framing.
• DGCCRF: French commercial-practice resources.
• Which?: regular consumer investigations on reviews.