How review manipulation works

Review manipulation is bigger than fake reviews

The obvious version is easy to imagine: a business pays strangers to leave five-star reviews for a product they never used.

That happens. But it is only one way to bend a rating.

A review system is a pipeline. Someone decides who gets asked, someone decides what qualifies as a review, a platform decides what stays visible, and an interface turns all of it into a star average, a count and a ranked list of comments.

Manipulation can enter at any of those points.

A business can add praise that never came from customers. It can offer rewards only for positive sentiment. It can funnel unhappy customers away from the public review page. It can pressure critics to delete posts. A marketplace seller can attach an existing pool of good reviews to a different product. Even feedback from real customers can become misleading if it is collected before they have actually experienced the thing being rated.

The core trick is not necessarily making every review fake. It is making the visible review pool look more representative than it really is.

That distinction matters because a page full of believable sentences can still create a false picture.

The manipulation can happen before a review is written

Suppose a business sends every customer a satisfaction survey.

People who choose “great” are shown a large button asking them to post a public Google review. People who choose “poor” are sent to a private support form instead.

The positive reviews may come from real customers describing real experiences. Nothing in the text itself has to be fabricated. But the collection process has given one group a much easier path into the public score.

This is often called review gating or selective solicitation.

The difference between asking customers for feedback and manipulating the result is the filter. Google allows businesses to encourage genuine reviews, but its current policy prohibits incentives and attempts to influence the rating or content. The US Federal Trade Commission’s review rule also prohibits compensation or incentives that are conditioned, explicitly or implicitly, on a particular positive or negative sentiment.

A review can therefore be authentic at the individual level while the sample is engineered.

That is harder to spot than a paragraph of obvious spam.

Fake reviews add experiences that never happened

The bluntest method is still fabrication.

Fake positive reviews can inflate the apparent reputation of a business or product. Fake negative reviews can damage a competitor. The reviewer may be an employee, friend, paid third party, automated account or invented identity. Generative AI can make the writing stage cheaper, but the deception does not depend on AI; fake reviews existed long before modern language models.

The FTC’s rule, which took effect on 21 October 2024, covers reviews and testimonials that falsely represent a real consumer experience, including reviews attributed to people who do not exist or people who did not actually use the product or service. It also addresses certain undisclosed insider reviews and the buying or selling of deceptive reviews.

Google’s Maps policies take a similar platform-level position: reviews should reflect genuine experiences, and paid reviews, multi-account posting and coordinated attempts to inflate or deflate ratings are prohibited.

The goal is simple: manufacture consensus.

One fake five-star review barely moves a mature listing. Enough of them can shift the average, bury legitimate complaints farther down the page and create the visual impression that “everyone” is satisfied.

Suppression can be more powerful than adding praise

A reputation can also be manipulated by removing information rather than adding it.

The FTC rule prohibits certain forms of review suppression, including using unfounded legal threats, intimidation or false accusations to prevent or remove negative reviews. It also prohibits misrepresenting a review section as representative of submitted reviews when negative feedback has been suppressed because of its rating or sentiment.

There is a reason this matters.

A 2018 study in the Journal of Business Research examined both adding positive reviews and deleting negative ones. The researchers found that manipulation could affect purchase intention, but excessive positive additions could also provoke suspicion. Deletion was less visible to consumers and therefore harder for them to detect.

That asymmetry is important. A suspicious five-star review leaves something to inspect. A review that never appears leaves nothing.

Review manipulation is therefore partly a problem of missing evidence.

A real customer can still produce a misleading “review”

One of the clearest examples came from the FTC’s case against Sitejabber.

According to the FTC, the platform collected ratings and comments from shoppers at the point of purchase, before they had received or had a chance to use what they bought. Those ratings were then used in ways the agency said could make consumers think they reflected experience with the product or service. The FTC approved a final consent order in 2025 prohibiting misrepresentations about the ratings and reviews Sitejabber collected, moderated or displayed.

No imaginary reviewer is required for that problem.

The person can be real. The purchase can be real. The star rating can be genuinely clicked by that person.

The misleading part is what the rating is presented as measuring.

A checkout experience is not the same thing as product quality after two months of use. A delivery rating is not a durability rating. A first impression is not a long-term ownership review.

This is why “verified customer” and “truthful review” are related but different ideas.

Reviews can be attached to the wrong product

Manipulation can also happen at the listing level.

In June 2025, the UK’s Competition and Markets Authority announced undertakings from Amazon that specifically addressed catalogue abuse: moving or combining reviews from a well-rated product so they appear on a different, unrelated product. The CMA’s example was blunt — a shopper might think they have found highly rated headphones, only to discover that many of the reviews actually discuss a phone charger.

This kind of manipulation exploits the interface itself.

The review text may be genuine. The reviewers may have bought something. The rating history may be years old.

It is the connection between review and product that has been corrupted.

For readers, this creates one of the most useful checks on marketplace listings: open several reviews and confirm that people are actually discussing the product currently on the page. References to a different size, model, item type or use case deserve attention.

Why manipulation works at all

Review systems compress a lot of messy evidence into a very small signal.

4.8 stars · 3,214 reviews

That line feels quantitative. It looks like a measurement rather than an argument. Most shoppers will not audit thousands of individual experiences before deciding whether to buy a cable, book a hotel or call a plumber.

That compression is the feature — and the vulnerability.

A manipulated review system does not need to convince a reader that every review is authentic. It only needs to move the aggregate signal enough to change the first impression.

Academic research supports the idea that manipulation has limits. The 2018 study found that excessive positive manipulation could backfire by increasing suspicion. A 2023 study in Information & Management likewise found that rating patterns that deviate from what users expect can increase perceptions of manipulation and reduce review adoption.

In other words, the best-looking manipulation is not necessarily a wall of perfect praise.

It is a distribution that looks ordinary enough to inherit the credibility of the platform around it.

Platforms look for patterns, not just suspicious adjectives

Readers often look for linguistic tells: vague praise, repeated phrases, excessive enthusiasm, strange grammar.

Those clues can be useful, but they are weak on their own. Genuine customers write generic reviews too, and generated text can vary its wording.

Platforms have access to signals readers do not: account histories, timing, device and network patterns, relationships between reviewers and businesses, purchase data, reporting behaviour and clusters of activity across many listings.

Trustpilot says it removed 4.5 million detected fake reviews in 2024, equal to 7.4% of reviews submitted that year. It says 90% of those detected fakes were removed automatically using systems that included machine learning, neural networks, graph-based models and generative-AI tools. Those are Trustpilot’s own platform figures, not an estimate that 7.4% of reviews across the internet are fake.

Google says businesses that violate its fake-engagement rules can face review restrictions, temporary unpublishing of existing reviews and consumer-facing warnings that suspicious reviews were removed.

Detection is therefore an arms race over behavioural patterns, not a spelling test.

What review manipulation looks like from the outside

No single sign proves a review is fake or a rating is manipulated. The useful approach is to look for several mismatches at once.

What you notice	What it might mean	What to check next
A sudden burst of extreme ratings	Promotion, viral attention, coordinated activity or manipulation	Compare dates with sales, launches or public events
Many reviews with little product detail	Low-effort genuine feedback or fabricated volume	Read reviewer histories and older reviews
Reviews mentioning another product	Listing merge, variant issue or catalogue abuse	Check model names, photos and older listing history
A near-perfect score with detailed recent complaints	Old reputation, selection effects or suppressed negatives	Sort by newest and lowest-rated reviews
Repeated reviewer activity across unrelated businesses	Active reviewer, campaign behaviour or a coordinated network	Inspect timing and geographic plausibility
“Verified” status	Some underlying transaction or platform check occurred	Ask what exactly was verified and what it does not prove

The last row matters most.

A badge can narrow uncertainty. It rarely eliminates it.

A verified purchase can help establish that an account bought something through the platform. It does not automatically prove independence, honest sentiment, long-term use or freedom from incentives.

The same evidence discipline applies to reviews as it does to screenshots: identify what the visible artifact actually establishes, then stop before the claim grows larger than the evidence.

The law is increasingly treating the whole review system as the problem

Regulators have moved beyond the narrow idea that only a fictional five-star paragraph counts as deception.

In the United States, the FTC’s Consumer Reviews and Testimonials Rule covers fake or false reviews, sentiment-conditioned incentives, certain undisclosed insider reviews, company-controlled “independent” review sites and specified forms of review suppression.

In the United Kingdom, strengthened consumer rules came into force in April 2025, and the CMA published dedicated fake-review guidance for businesses that publish reviews or information derived from them. The CMA separately secured commitments from Google and Amazon around detection, sanctions and suspicious review activity.

The details differ by jurisdiction, and platform policy is not the same thing as law. But the direction is consistent: a trustworthy review system depends on more than whether individual sentences were typed by real customers.

It depends on how the reviews were obtained, filtered, attached, ranked and represented.

Treat the rating as a lead, not a verdict

A star average is useful. It is also the end product of a system you usually cannot see.

The practical response is not to assume every review platform is corrupt. It is to avoid giving one compressed score more authority than it deserves.

Read across rating levels. Sort by recent. Look for product mismatches. Compare another platform. Pay attention to public warnings. Check whether reviewers describe specific experiences rather than merely repeating the sales pitch.

Most importantly, separate two questions:

Are these individual reviews genuine?

and

Does this collection of reviews give me a fair picture of the thing I am deciding about?

Review manipulation works by making those questions feel identical.

They are not.