Short answer: a fake review is usually given away by pattern, not by wording. One suspicious review means nothing. What gives a campaign away is a cluster of reviews posted in a narrow time window, from accounts with no other history, using the same sentence structures with only the brand name changed, and describing the product in terms that do not fit what the business actually sells. Below are the twelve signals that hold up, ranked by how reliable each one is on its own.
The twelve signals, ranked by reliability
| # | Signal | What it looks like | Reliability on its own |
|---|---|---|---|
| 1 | Category mismatch | A review of a bank or a flight-booking site that complains about “delivery” and “the wrong item arriving” | Very high |
| 2 | Repeated body text | The same sentences appear on several listings with only the brand name swapped | Very high |
| 3 | Timestamp clustering | Dozens of reviews within minutes or hours, often at the same second offset | Very high |
| 4 | No reviewer identity | Blank names, default avatars, or accounts with exactly one review | High |
| 5 | Bimodal score with no middle | A wall of 5★ and a wall of 1★, almost no 3★ | Medium–high |
| 6 | Generic praise, zero specifics | “Great service, highly recommended” with no product, date, price or staff name | Medium |
| 7 | Marketing vocabulary | Reviewers using the brand’s own tagline or full legal name repeatedly | Medium |
| 8 | Review velocity spike | A listing that averaged two reviews a month suddenly gets ninety | Medium |
| 9 | Perfect grammar in every review | Real review sets are messy; uniform polish suggests one author | Low–medium |
| 10 | Reviews that argue with other reviews | “Ignore the negative reviews, they are competitors” | Low–medium |
| 11 | Incentive language | “Got this free for an honest review” without a disclosure label | Low on its own, but a compliance breach |
| 12 | Five stars with a complaint in the text | The body describes a bad experience but the rating is 5★ | Low, but a strong tell in bulk |
Why category mismatch is the single best test
Bulk-written reviews are produced from a small set of templates that were written for one industry — almost always retail — and then applied everywhere. The template mentions delivery, packaging, orders and returns because that is what a retail review sounds like. When the same template lands on an insurance company, a coaching platform or a hotel chain, the vocabulary stops making sense.
So the fastest check on any listing is: does the review describe something this business could actually do to you? A hostel cannot deliver your order late. A stockbroking app does not send the wrong item. If you see that, you are not reading a customer.
How to check a listing in ninety seconds
- Sort by newest, not by relevance. Manipulation is almost always recent and bunched. Sorting by relevance hides exactly the pattern you are looking for.
- Read three 1★ and three 5★ reviews. Real complaints and real praise both contain specifics: a date, an amount, a model number, a city. Fabricated ones stay abstract.
- Copy one sentence and search for it in quotes. If the identical sentence appears on other listings, you have found a template.
- Look at the reviewer, not the review. An account with one review, no photo and no name carries almost no evidential weight, whatever it says.
- Check whether the middle exists. Genuine review sets have plenty of 3★ and 4★ entries. A distribution with nothing in the middle has usually been pushed from both ends.
What the rules in India actually say
India has a published standard for this. IS 19000:2022, Online Consumer Reviews — Principles and Requirements for their Collection, Moderation and Publication, was released by the Bureau of Indian Standards with the Department of Consumer Affairs and took effect on 25 November 2022. It is voluntary, not mandatory, and it is adapted from the international standard ISO 20488:2018.
Separately, the Consumer Protection Act, 2019 and the Consumer Protection (E-Commerce) Rules, 2020 treat fabricated reviews as an unfair trade practice and a misleading representation, and the Central Consumer Protection Authority (CCPA) has powers to act against misleading claims. Paid or incentivised reviews are not automatically illegal — but publishing them without disclosing the incentive is what creates the problem.
This section describes the instruments by name so you can look them up. It is general information, not legal advice.
Signals that people trust but should not
- Bad spelling. Plenty of genuine Indian reviewers write in a second or third language. Poor grammar is evidence of nothing.
- A very low rating. Angry reviews are usually the most real ones on a page. People rarely fabricate a detailed complaint with an order number.
- A very high overall average. An average tells you almost nothing without the distribution behind it and the number of reviews it was computed from.
- “Verified” badges. These mean different things on different platforms — sometimes a verified purchase, sometimes only a verified email. Check what the badge is actually claiming.
Frequently asked questions
How can you tell if a review is fake?
Look for pattern rather than wording. The strongest single signal is category mismatch — a review describing delivery or packaging on a business that sells neither, such as a bank or a booking site. After that, the most reliable tells are the same sentences repeating across different listings with only the brand name changed, and large numbers of reviews posted inside a very narrow time window. Any one review is impossible to judge; a cluster is not.
Are fake reviews illegal in India?
Fabricating reviews can be treated as an unfair trade practice and a misleading representation under the Consumer Protection Act, 2019 and the Consumer Protection (E-Commerce) Rules, 2020, and the Central Consumer Protection Authority has powers to act. India also has a published standard, IS 19000:2022, covering how reviews should be collected, moderated and published — but that standard is voluntary rather than mandatory. Incentivised reviews are not banned outright; failing to disclose the incentive is the compliance problem.
Is a 5-star average a good sign?
Usually the opposite. Businesses with genuine volume almost never average five stars, because real customers include people who had an ordinary day. A perfect or near-perfect average on a listing with a large number of reviews is a reason to look at the distribution and the posting dates rather than a reason to relax.
How many reviews should a listing have before I trust the score?
As a rough working rule, under about thirty reviews the average moves too much to mean anything — a single new rating can shift it noticeably. Between thirty and a hundred you can start trusting the shape of the distribution. Above a few hundred the average is stable, but you should still read the recent ones, because a score built over three years can hide a business that got worse last quarter.
Do negative reviews mean a business is bad?
Not by themselves. What matters is the proportion and the theme. Every business with real customers collects one-star reviews. The useful question is whether the complaints repeat the same specific failure — refunds not processed, a charge that reappears, support that never replies — because a repeated specific complaint is a real operational problem, whereas scattered unrelated complaints are ordinary noise.
Can a business pay to delete bad reviews?
On a platform operating properly, no. Reviews should only be removed when they breach a published policy — for example if they are defamatory, contain personal data, or can be shown to be fabricated — and the same rule has to apply to positive reviews too. If a platform removes negative reviews on request while keeping positive ones, its scores carry no information at all.
Comments (0)