How to Spot Fake Reviews — and Read the Real Ones Properly

Nobody reads reviews for entertainment. You read them because the photograph tells you nothing, the specification sheet tells you nearly nothing, and a few hundred strangers have already found out what happens after six months. That makes the review section the most useful part of a listing — and the part with the most to gain from manipulation.

The good news is that manipulation is now unlawful in the UK, which has pushed the more sophisticated operators out and left the crude ones visible. The better news is that reading reviews well is pattern recognition, and most of it takes under a minute.

What UK law now says about reviews

The Digital Markets, Competition and Consumers Act 2024 replaced the earlier unfair trading regulations, and its consumer provisions came into force on 6 April 2025. Schedule 20 of the Act lists 32 commercial practices that are unfair in all circumstances — banned outright, with no need to show that any particular shopper was misled. Paragraph 13 is the one about reviews, and it is unusually specific.

It bans submitting a fake review, or commissioning someone else to write or submit one, where "fake" means a review presenting itself as based on genuine experience when it is not. It bans reviews that conceal the fact they were incentivised. It bans publishing reviews, or the information derived from them, in a misleading way. And it puts a positive duty on anyone publishing reviews to take reasonable and proportionate steps to prevent fake and concealed-incentive reviews appearing, and to remove them.

Usefully for readers, the Act spells out what "misleading way" includes: suppressing negative reviews while publishing positive ones, giving greater prominence to one over the other, and omitting information about the circumstances in which a review was written — such as the fact that someone was paid to write it. Because the definition covers "consumer review information" as well as reviews, it reaches the aggregate star rating, not just the text beneath it.

Enforcement changed as much as the rules. The Competition and Markets Authority can now act directly rather than through the courts, with penalties reaching up to 10% of global turnover, and it has opened investigations into fake and misleading reviews across several sectors. The practical consequence is that overt fakery is riskier, so what remains tends to be subtler — pressure applied to genuine reviewers rather than reviews invented wholesale.

Five ways review pages get manipulated

• Bulk fabrication. Batches of invented reviews posted over a short window, often before the product has plausibly reached buyers.
• Incentive without disclosure. A refund, voucher or free unit in exchange for five stars, arranged off-platform so nothing appears on the listing.
• Negative suppression. Buyers contacted privately and offered a refund to withdraw a poor rating, so the published distribution no longer reflects the real one.
• Listing hijacking. A listing built up under one product is edited to sell a different one, inheriting thousands of ratings for something else.
• Aggregate framing. A high average calculated across variants, bundles or a whole brand rather than the item you are looking at.

Only the first is what most people picture when they think about how to spot fake reviews. The other four leave the individual reviews genuine and distort the summary — which is why the average is the least informative number on the page.

Signals in the writing itself

Fabricated reviews describe buying rather than owning. They reproduce marketing language, list features in the order the listing does, and praise the product for existing. What they rarely contain is friction: the awkward first assembly, the setting that took a fortnight to find, the thing that turned out to be fine but irritating.

Look for specificity that would be hard to invent. Genuine reviewers mention their kitchen, their commute, their previous model, the day the courier came. They also mention duration — "three months in", "second winter" — because ownership has a timeline and a purchase does not.

Two weak signals to discount: poor spelling proves nothing, and neither does fluent prose. Judge the presence of lived detail, not the polish of the sentences.

Signals in the rating distribution

Open the histogram before you read a single review. A genuine product with thousands of buyers produces a recognisable shape: a large five-star block, a solid four-star block, a thinner middle and a persistent tail of one-star ratings from units that failed or buyers who bought wrong. Real populations are messy.

Three shapes deserve caution. A near-empty tail across a large volume is statistically odd, because some fraction of any manufactured run fails. A spike of five-star ratings clustered into a few days, especially early in a product's life, suggests coordination. And two peaks with nothing between them often means the listing covers two different things, or a specific failure mode splits owners into delighted and furious.

Check volume against plausibility too. A product launched last month with four thousand ratings has either had an extraordinary launch or inherited them from somewhere else.

Signals in the reviewer profile

Where the platform allows it, look at the reviewer rather than the review. Accounts that reviewed twelve unrelated products in a day, that have never awarded fewer than five stars, or that review only within one seller's catalogue are worth discounting — as are clusters of accounts created in the same period reviewing the same narrow set of items.

Photographs help in both directions. Genuine review photographs are badly lit, taken on a worktop, and show the product in a house. Studio images in a review are usually the manufacturer's, which tells you the reviewer never opened a box.

Verified purchase badges confirm a transaction was recorded — not that the reviewer paid full price, kept the item, or was uninfluenced. Treat the badge as one input among several.

How to mine genuine reviews for useful information

Once you have decided a review section is broadly honest, stop reading it for verdicts and start reading it for patterns.

1. Filter to three and four stars. This is where the trade-offs live. Five-star reviews say it arrived; one-star reviews often describe delivery.
2. Sort by recent. Manufacturing changes, components get substituted, firmware gets worse. A rating average is a historical document.
3. Search the review text for failure words — "stopped", "cracked", "leak", "battery", "replaced", "after". Repetition of the same component across unrelated reviewers is the strongest quality signal a listing can give you.
4. Ignore expectation mismatches. "Smaller than I thought" is a measurement error by the buyer, and exactly the failure mode how to shortlist products is designed to prevent.
5. Note what nobody mentions. If no one in two hundred reviews mentions noise in a category where noise matters, either it is fine or nobody has used it long enough.

Repeated, specific, mechanical complaints are evidence. Everything else is anecdote — and both are separate from the construction evidence covered in how to judge product quality online.

Cross-checks that take under a minute

• Compare the rating profile of the same product on two retailers. Populations differ; distributions of that different a shape usually mean something.
• Check whether the listing title matches the reviews. Reviewers describing a different colour, size or category means the listing has been repurposed.
• Search the product name with "problem", "fault" or "spare part" and read owner forums rather than shopping pages.
• Read the two-star reviews the retailer has left visible. Suppression is rarely complete.
• Compare reviews on the brand's own site with those on a marketplace it does not control.
• Where a claim is quantified — clinically proven, dermatologically tested — check who ran the study and on how many people, the habit developed in beauty marketing claims explained.

Reporting a suspicious listing in the UK

Start with the platform's own reporting tool: publishers now have a legal duty to take reasonable and proportionate steps to remove banned reviews, and a report creates a record. If nothing changes, report it to the Citizens Advice consumer service, which passes information to Trading Standards, or to the Competition and Markets Authority, which collects reports on unfair commercial practices.

If the problem is a paid social post or an influencer promotion presented as an ordinary opinion, that is an advertising matter and the Advertising Standards Authority takes complaints. The same principle applies to publishers as to platforms, which is why how to read a buying guide starts with disclosure rather than with picks.