How to Spot Fake Reviews Online (5 Checks That Work)

A few years ago, spotting a fake review was easy. Broken English, wild enthusiasm, five identical reviews in a row. Those tells belonged to humans writing fakes in bulk, and that era is over.

Fake reviews are written by AI now: fluent, specific, plausible, free at any scale. A 2025 study of 714,000 reviews found AI fakes read better than genuine ones. Google removes hundreds of millions of policy-violating reviews a year, and the survivors are the ones built to pass the filters.

So stop reading reviews for tells. Read the structure around them. Five checks, two minutes, works on any platform including Bharosilo.

1. Read the middle ratings first

Bought praise piles at five stars. Paid attacks pile at one. Real customers spread out, and the useful writing sits at three and four stars, where people who liked the place still say what went wrong.

"Great momo, service collapsed after 8pm, would still go back on weekdays" beats ten "best in town" reviews with nothing else. A profile with forty reviews and zero middle ratings is the biggest warning sign on this page. Real crowds disagree. Manufactured ones do not.

2. Demand details only a visit produces

Real reviews leak accidents: the hour the crowd shows up, parking on Saturdays, the waiter who remembered the order, the texture of the achar. Nobody invents those without eating there.

Fakes describe feelings. Amazing. Wonderful. Highly recommended. "Excellent service, will visit again" fits a hotel in Pokhara, a dentist in Kathmandu, and a jacket shop equally well, which means it describes none of them. One question filters everything: could someone write this without visiting?

3. Click the reviewer, not the review

Four questions, ten seconds. Does the history look like a life, or forty unrelated five-star businesses in a week? Does the geography make sense, or did they praise a Lakeside cafe and a Biratnagar hotel on the same day? Is there any criticism at all? Are the photos crooked and real, or stock-clean on every review?

A farm account fails at least two of these every time.

4. Check the shape and the calendar

Open the star breakdown. Healthy businesses slope down from five stars with a scattering below. A U-shape, piles of fives and ones with nothing between, means the fives were bought and the ones arrived after. A perfect 5.0 across hundreds of reviews is curation, not excellence.

Then sort newest-first and look at dates. Real reviews trickle. Campaigns land in bursts: thirty reviews in three days, then silence. Four reviews in one afternoon at a place averaging four a month is the strongest single signal there is.

5. Verify against the world

The most decisive test is the simplest. Does the review name a service the business does not offer, or a branch that does not exist? Then the writer never walked in, and you have proof, not a hunch.

Cross-check somewhere else while you are at it. A 4.8 on the business's own site next to a 3.1 on independent listings is its own verdict. Search the name with "complaint" or "reddit". Forum threads going back years are nearly impossible to fake.

The sixty-second routine

Star breakdown for the U-shape. Newest-first for bursts. Three four-star reviews for the honest complaints. One reviewer profile for the farm check. One outside search. Two red flags and you walk.

And when a place treats you well, return the favor properly. Vague five-star praise ("nice place!") is indistinguishable from the fakes and helps them blend in. Write what you ordered, what it cost, when you went, what was good, what was not. That is the review that starves the fakes. Leave it on Bharosilo, where claimed listings and moderation give it a chance.

Updated October 2026. For the business side of this, see [how to earn your first ten reviews](https://bharosilo.com/blog/claim-business-reviews).