Field note · 8 min read

How to Spot Fake and Bought Reviews

Five-star bursts, templated text, one-review accounts, and what a "verified purchase" badge actually proves.

The ad looks fine. The store looks fine. Then you scroll to the reviews, find 4,000 of them at 4.8 stars, and relax. That is the moment most people stop checking. It is also the exact moment the manipulation is aimed at.

This is not a fringe problem. Trustpilot's own Trust Report says it removed 4.5 million fake reviews in 2024 — 7.4% of every review submitted to the platform that year, with about 90% caught automatically. That is one platform's count of what it caught. It says nothing about what got through, and nothing about the platforms that publish no numbers at all.

Most of this has been illegal in the US since 21 October 2024, when the FTC's rule on consumer reviews and testimonials took effect. Illegal is not the same as gone. The rule gives regulators a lever; it does not clean your screen before you shop. So you check yourself. Here is how, in the order that actually saves you money.

A rating is a shape, not a number

Stop reading the average. Read the distribution and the dates. A real product collects reviews at a boring, roughly steady rate, and the star spread has a middle to it — some fours, some threes, the occasional two. Manipulated listings look different: a wall of fives, a thin spike of ones, and almost nothing in between. That shape alone is not proof, because genuine buyers do skew toward delighted or furious. Combined with timing, it is a strong signal.

Timing is the more useful half. The FTC's consumer guidance flags a burst of reviews over a short period as a warning sign, especially on a product that has been around a while. Research published in Marketing Science followed sellers who bought reviews through private Facebook groups and found the same arc: average ratings rose sharply while the buying was happening, then fell once it stopped, with the share of one-star reviews rising significantly afterwards. That decay is the tell. It means the rating was never a measurement of the product. So sort by most recent. If the last month reads far worse than the lifetime average, the average is a historical artifact and the recent reviews are the news.

Read the text for what it doesn't say

A real review is specific and slightly lopsided. It names one thing that annoyed the buyer even when they liked the product: the strap was stiff for a week, it arrived eleven days late, the colour is greyer than the photo. Fake and templated reviews are smooth. They restate the full product name, praise "quality" and "fast shipping" in the abstract, and never once describe using the thing. AI-written reviews are the same emptiness at higher fluency — grammatically clean, tonally even, structurally identical to each other, and completely free of friction. The FTC has confirmed AI-generated reviews are covered by its rule, so a shop cannot launder this by generating text instead of paying writers.

Look for repetition across reviews rather than inside one. Same sentence rhythm, same three benefits in the same order, the same slightly odd phrase turning up in five reviews from five different accounts — that is a brief being followed. Watch for reviews that quote the ad. When a review uses the exact marketing phrase from the video that brought you here, it did not come from someone describing their week. Genuine customers do not talk in ad copy. They talk about the parcel.

Check the account, not just the review

Click the reviewer. Profiles with exactly one review are the cheapest thing in the world to manufacture. One review on its own means nothing — plenty of people write one review in their life and never come back. What matters is concentration. If the five-star block is overwhelmingly single-review accounts, and those accounts all appeared within the same week or two, that is a pattern, not a coincidence.

The opposite profile is also worth catching: an account posting five stars at high frequency across wildly unrelated products — a phone case, a supplement, a dress, a driveway sealer. That is a working account, rented or sold. There is a documented supply chain behind it. In 2022 Amazon sued the administrators of more than 10,000 private Facebook groups it said were brokering reviews in exchange for money or free product, and reported that Meta had removed a majority of the groups named. On review platforms, also check the collection method: reviews are usually labelled as invited (the company asked, using an order reference it supplied) or organic (the customer turned up alone). A profile whose reviews are almost entirely invited and clustered into short campaigns is telling you how the score was assembled.

Incentives, gating, and what the law actually bans

The FTC's rule, in force since 21 October 2024, bans a specific list: writing, selling, buying or disseminating reviews from people with no genuine experience of the product; paying anyone for a review conditioned on it being positive, or on it being negative about a competitor; insiders — officers, managers, employees and their relatives — reviewing without disclosing the connection; running a review site you control while presenting it as independent; buying fake followers, likes or engagement; using groundless legal threats or intimidation to get a bad review pulled; and claiming your review section shows all or most reviews when the negative ones have been filtered out.

Two nuances change how you read a page. First, incentivised reviews are not banned outright. A shop can give you something for reviewing — it just cannot make the reward depend on your sentiment, and a material incentive has to be disclosed. So "I received this free in exchange for my honest review" is legal, and you should still discount it. Second, review-gating — the "Were you happy? Yes leads to Google, No leads to a private form" funnel — is targeted through the misrepresentation: presenting the survivors as if they are the whole picture. Enforcement is live. In December 2025 the FTC sent warning letters to ten companies over precisely these practices, demanded written confirmation of fixes within five days, and cited civil penalties of up to $53,088 per violation. The UK moved the same direction: fake reviews and concealed incentivised reviews became directly enforceable by the CMA under the Digital Markets, Competition and Consumers Act 2024 from 6 April 2025.

What "verified purchase" actually guarantees

Less than you think. On Amazon, the badge means the account leaving the review bought that item through Amazon at a price available to ordinary shoppers — deeply discounted purchases do not qualify. That is a fact about a transaction. It is not a fact about the reviewer's honesty, their independence, or whether they ever opened the box. It rules out the laziest fakes and nothing else.

The standard broker workflow is designed specifically to earn that badge. The reviewer buys at full price with their own card, posts the review, and gets reimbursed off-platform afterwards. Verified badge, purchased opinion. Two further limits. Reviews can be pooled across variants of a listing, so praise for a different colour, size, or occasionally a substantially different item can sit under the thing you are looking at; Amazon has been tightening this, but you should still check the review describes what you are buying. And on review platforms, "verified" generally means the company handed over an order reference and invited that customer. It confirms the company asked them. It tells you nothing about who the company did not ask.

Read the one- and two-star reviews first

Flip the sort order and start at the bottom. You are not looking for anger — you are looking for repetition. One person saying the parcel took six weeks is noise. Fifteen people saying it across eight months is the delivery model. For stores that sell through Facebook and Instagram ads, the recurring complaint is usually about what happens after payment rather than the item itself: multi-week shipping from a country the ad never mentioned, sizes running two down from the chart, an item that does not match the photo, returns that require an RMA number nobody sends, return postage to an overseas address that costs more than the order, restocking fees, and refunds that only arrive after a card chargeback. Those complaints repeat because they come from the same business model, which is why the same failure pattern shows up in the documented cases below.

Here is the check. Four minutes, in this order. One: sort by most recent and read the last month — if it is much worse than the lifetime average, trust the recent reviews. Two: read the one- and two-star reviews and write down any complaint that appears more than three times. Three: confirm the star distribution has a middle; all fives and ones with nothing between is either manufactured or genuinely polarised, and both mean you need more information. Four: click three of the five-star reviewers — if they are single-review accounts dated the same week, discount the entire block. Five: check the negative reviews are actually about the product you are buying, not a pooled variant. If the repeating complaint is about refunds and returns rather than the item, you already have your answer. The price on the page is not the price of the mistake.