As of November 4, 2026, review credibility has become a sales variable rather than a platform policy topic. A survey of 6,000 shoppers across the United Kingdom, the United States and Italy for the Coalition for Trusted Reviews found 74 percent in the UK, 73 percent in the US and 61 percent in Italy have avoided a purchase over concerns about fake or misleading reviews. For merchants and marketplaces, that means a listing can lose the order before the product page finishes loading, and the fix sits in how ratings are produced, not in how they are displayed.
Review doubts now end purchases, not just delay them
The coalition behind the research includes Amazon, Booking.com, Expedia Group, G2, Holidaycheck, Tripadvisor and Trustpilot, which makes the finding harder to dismiss as a single platform's complaint. Platforms that compete on listings have a shared interest in review integrity, because a shopper who stops trusting ratings stops using ratings as a shortcut and starts doing comparison work the platform cannot monetize.
There is an uncomfortable trade-off in the numbers. Identity verification would reduce fake reviews, but it also reduces the supply of reviews: 56 percent of UK respondents said they would be less likely to write one if required to show ID. Fewer reviews weakens the signal that shoppers rely on, so platforms face a choice between a smaller, cleaner sample and a larger, noisier one. Most have chosen neither explicitly, which is why the doubts persist.
Ratings still steer choice, but provenance decides trust
Separate research on 2,220 Canadians in the third quarter of 2026 shows how much weight a rating still carries when the alternative is brand messaging. When information conflicts, 57.4 percent trust recent online reviews from verified customers, against 15.3 percent who trust information supplied directly by the brand and 9.4 percent who trust creators and influencers. Provenance, in other words, is doing the work that star counts used to do.
The brand name no longer settles the comparison
| Signal | Share of respondents | What it implies for listings |
|---|---|---|
| Trust recent verified customer reviews | 57.4% | Verification status belongs next to the rating, not in a policy page |
| Trust information supplied by the brand | 15.3% | Brand copy cannot carry a claim that reviews contradict |
| Trust creators and influencers | 9.4% | Creator volume does not substitute for review depth |
| Would choose an unfamiliar brand with excellent reviews | 48.7% | Review quality can outweigh brand recognition in a direct comparison |
| Would prefer an established brand with average reviews | 20% | A recognized name protects less than sellers assume |
The head-to-head result is the one worth pinning on a wall. Asked to choose between a familiar brand with average recent reviews and an unfamiliar brand with excellent recent reviews, 48.7 percent picked the unfamiliar one and only 20 percent stayed with the known name, with 31.3 percent wanting more information first. Brand equity still matters, but it no longer overrides a weak recent record on a product page.
For marketplace sellers this changes the economics of review generation. Buying volume is now a liability if the resulting ratings look synthetic, because the same shoppers who distrust fakes are also the ones who read the middle of the distribution. A product with a large number of five-star reviews and no three-star detail reads as unverified to a skeptical buyer, while a smaller set with specific complaints reads as honest.
AI written reviews split consumer trust
Research published in the International Journal of Electronic Commerce tested three types of AI-generated reviews and found that trust depends on which type is used. Mechanical reviews, assembled from patterns in other sellers' reviews, are trusted less than genuine consumer reviews because they read as detached from real experience. Thinking reviews, which reason about use cases, perform best as a substitute when a product has too few consumer reviews, while feeling reviews work better when genuine reviews already exist.
That distinction matters because platforms increasingly fill thin listings with generated text to avoid an empty page. Filling the gap with the wrong type can lower trust on exactly the listings that need help, and consumers are already primed to discount anything that looks machine-written. The practical rule is to disclose, keep generated text away from experience claims, and never let it sit where a verified review should be.
Where a rating still wins
Reviews remain the cheapest way for a new product to clear the credibility threshold, which is why the fix is not to remove ratings but to make their origin legible. A listing that shows verified purchase status, the number of reviews behind a score, and a visible sample of critical reviews answers the question a shopper is actually asking. Hiding the middle of the distribution to protect a headline number invites the doubt the research documents.
What a product page audit needs
Start with the review sample rather than the average. Count how many reviews sit behind the score, how many are marked verified, and whether the critical reviews describe the same failure twice. Then check the claims: any performance number in the copy should be traceable to a test the seller can produce, because a claim that reviews contradict costs more than the claim earns. Finally, record the review velocity, since a burst of identical five-star text within a few days is the pattern shoppers have learned to distrust.
Marketplaces have a parallel job. Verification rules that reduce review supply need a replacement signal, or listings go quiet and shoppers default to brand names. Purchase-linked ratings, structured attributes and a visible count of returns by reason all provide evidence without asking buyers to prove who they are, and each of them can be audited by the platform rather than trusted on faith.
FAQ
How large was the survey behind the abandonment figures?
The Coalition for Trusted Reviews commissioned research across 6,000 people in the UK, Italy and the United States. It found 74 percent in the UK, 73 percent in the US and 61 percent in Italy have avoided a purchase over fake or misleading review concerns.
Would ID checks fix the problem?
Only partly. Identity verification would reduce fake reviews but 56 percent in the UK, 46 percent in the US and 43 percent in Italy said they would be less likely to leave a review if ID were required, which shrinks the sample shoppers depend on.
Are AI-generated reviews always less trusted?
It depends on the type. Research in the International Journal of Electronic Commerce found mechanical AI reviews are trusted less than consumer reviews, while thinking-type generated reviews work best as a substitute when genuine reviews are scarce.
