A listing can carry reviews written about a different product

Everything here earned its place by changing an outcome. Nothing about review provenance on listings is included to round the number up.

How reviews and listings come apart

Marketplaces commonly group variants under one listing so that colours, sizes and capacities all share a single pool of reviews. That is reasonable when variants differ trivially and misleading when they differ substantially, and the platform rarely distinguishes between the two cases.

Sellers can sometimes edit a listing extensively after it has accumulated reviews, replacing the product while keeping the accumulated history intact. Listings are also merged and split by platforms for catalogue reasons, which can carry reviews across items that were never related. None of this requires fake reviews to be involved, which is precisely why review-fraud detection does not catch it.

Repurposed listings and inherited history

A listing with a long history and a high rating is a commercial asset, so there is a standing incentive to reuse one rather than start again. Where a seller changes what they sell but keeps the listing, the older reviews describe a product that is no longer available anywhere. Product images and titles update instantly while review text does not, so the mismatch sits in the text of the oldest entries.

The pattern is most common in categories with rapid model turnover and generic manufacturing, where one listing can serve many similar products. A sudden change in what reviewers describe, at a specific point in the timeline, is the signature of this having happened.

Reading reviews for provenance

Sort by oldest rather than by most recent, since the earliest reviews reveal what the listing originally sold. Look for descriptions of features, colours, dimensions or accessories that the current product does not have at all.

On the bench, check whether reviewers refer to a different category entirely, which happens rather more often than the review count would suggest. Where a platform marks which variant a reviewer bought, use that marker and disregard reviews attached to variants you are not considering. Reviews that quote a specific model designation are worth more than those that do not, because they can be matched to the item in front of you.

Incentivised and solicited reviews

Reviews obtained in exchange for a free or discounted item are lawful in many jurisdictions when disclosed, and disclosure rules differ by country. Solicitation timing shapes content, because a request sent days after delivery collects first impressions rather than any durability information. Follow-up requests are often sent only to buyers who did not report a problem, which biases the pool before anyone writes a word.

Insert cards offering something in exchange for a review are prohibited by several platforms and still appear regularly in packaging. None of these necessarily produce false statements, but all of them shift the distribution of who writes and when they write.

Patterns that suggest manipulation

Clusters of reviews arriving in a short window after a long quiet period suggest coordination rather than a sudden surge of buyers. Repeated phrasing across accounts, unusual grammar consistent between reviewers, and generic praise without product-specific detail are all recognised signals.

On the bench, reviewer histories consisting entirely of five-star entries across unrelated categories are a stronger signal than any individual review's content. A rating distribution with almost nothing in the middle can indicate two different populations rather than one product performing inconsistently. Third-party analysis tools exist for this and their methods are proprietary, so their conclusions are indicative rather than authoritative.

Using reviews for what they are good at

Reviews are strongest as a source of failure descriptions, since people describe specific breakages in specific words that are hard to fabricate convincingly. They are weakest as a measure of quality, because the rating scale compresses and the population who writes is not the population who buys. Filtering to reviews written some months after purchase, where the platform shows dates, gives access to the only durability evidence available.

In the small print, cross-checking a claimed fault against a repair forum or a parts supplier confirms whether it is a pattern or an isolated incident. Reading twenty detailed reviews carefully tells you more than an average computed from several thousand of them.

Everything above, in order of what to do first

1. How reviews and listings come apart. Marketplaces commonly group variants under one listing so that colours, sizes and capacities all share a single pool of reviews.
2. Repurposed listings and inherited history. A listing with a long history and a high rating is a commercial asset, so there is a standing incentive to reuse one rather than start again.
3. Reading reviews for provenance. Sort by oldest rather than by most recent, since the earliest reviews reveal what the listing originally sold.
4. Incentivised and solicited reviews. Reviews obtained in exchange for a free or discounted item are lawful in many jurisdictions when disclosed, and disclosure rules differ by country.
5. Patterns that suggest manipulation. Clusters of reviews arriving in a short window after a long quiet period suggest coordination rather than a sudden surge of buyers.
6. Using reviews for what they are good at. Reviews are strongest as a source of failure descriptions, since people describe specific breakages in specific words that are hard to fabricate convincingly.

The takeaway

Read the oldest reviews first, because that is where a listing's history is still visible.

Buy for the failure you can live with, not the feature you will use twice.