How to Detect Fake and AI-Generated App Reviews: A Practical Guide

A couple of weeks back I was poking around a competitor's App Store page, curious how they'd got to 4.8 stars so fast.

Fifteen new reviews in nine days. Every one was four sentences long. Every one loved "the interface," and not one said which screen they meant.

I didn't need software to call it. I just needed to stop reading the reviews one at a time.

That's the whole trick, and it's what most guides on fake app reviews get backwards. They give you a checklist of writing quirks to hunt for. That worked until roughly 2023, before generated text stopped having obvious quirks. Even the FTC's own advice on evaluating online reviews says plainly that you can't assume you'll spot a fake just by reading it.

So read the crowd instead. Who posted, when they posted, and how much they sound like each other. The rest of this post is about those three things.

What Actually Counts as a Fake App Review?

Bots, obviously. Bought five-star bundles, obviously. But the category is wider than most people think.

Two indie devs reviewing each other's apps? Fake. A founder's cousin leaving five stars without saying so? Fake, legally speaking. A "free premium month if you leave a review" dressed up as a spontaneous opinion? Misleading enough that regulators treat it the same way.

Then there's the side nobody talks about: coordinated one-star attacks after a pricing change, from a rival or just an angry Reddit thread. Google calls this review bombing. It says it blocked 160 million spam ratings and reviews in 2025, and that catching them saved affected apps an average of half a star.

How to Spot Fake Reviews Without a Tool

Four things I check, in this order.

The reviewer, before the review

A five-star review from an account with no other activity is close to meaningless. Click through to the profile if the store lets you. What I'm looking for is someone who reviewed a meditation app, a budgeting app and three match-three games on the same Tuesday, in the same sentence shape. That person probably doesn't exist.

The timeline

Review volume is lumpy for normal reasons. Launch day, a big update, somebody with 200k followers mentions you. Those spikes have a cause you can point to. The suspicious kind is two reviews a week for a year, then thirty in three days with nothing in the release notes. If nothing happened in that window, you have your answer.

One specific moment

Real people remember a thing. "Sync broke after the March update." "Offline mode is why I kept it, I'm on the metro an hour a day." The detail is incidental, which is why it's convincing.

Fake reviews float. Amazing. Must-have. Highly recommend. Plenty of real users write exactly that, because they're busy. One vague five-star tells you nothing. Fifteen in a week tells you quite a lot.

The star spread

Healthy listings are messy: lots of fives, a decent number of fours with small gripes, some threes, and a stubborn tail of ones. When the middle is nearly empty and it's all fives, or all ones, somebody has been tilting the table.

How to Tell If App Reviews Were Written by AI

I'll be blunt: you can't reliably tell whether a short app review was written by AI from the text alone. Anyone selling you certainty on this is selling you something. What follows are hints, and they only mean much in bulk.

Uniformity is the big one. No typos across fifteen reviews, identical punctuation, the same three-beat structure every time: praise, small caveat, recommendation. People typing with their thumbs on a train don't produce that.

Hedging is another. Generated text likes to soften everything ("generally quite good," "mostly works well for my needs") rather than just saying what happened.

Filler words, too. "Overall" turns up a lot. So does "seamless." One instance is nothing. The same word in a dozen reviews from a dozen "different" people is a tell.

And watch for planted keywords. "Best calorie counter for home workouts" is not how anyone talks about an app they use. It's how someone writes when they're thinking about search rankings.

Why AI review detection is so unreliable

Lots of genuine people write polished sentences. Some have a real experience and ask ChatGPT to tidy the wording, which isn't fraud. And reviews are short, and short text is where detectors fall apart. If a tool hands you "82% AI-written," treat it as a rumour, not a verdict.

How to Check If App Reviews Are Fake in Practice

When I actually sit down to investigate a listing, this is the routine.

Sort by newest first. Bursts jump out immediately when the list is chronological.

Then read the extremes. Five one-stars, five five-stars. I'm looking for a specific complaint or a specific delight. Not adjectives.

Click into a few reviewer profiles. Two minutes, tops.

Compare the stores. An app at 4.8 on iOS and 3.2 on Google Play has some explaining to do. Sometimes it's an Android bug. Sometimes it's where the budget went.

Match review dates against release notes. I once found a review raving about a dark mode that shipped three weeks after the review was posted. That one's hard to argue with.

Finally, search the app name plus "fake reviews" or "review manipulation." Often somebody on a forum has already done the work.

And don't lean on the AI summary both stores now show at the top of the reviews tab. It summarises whatever's underneath, fakes included.

Does a Fake Review Checker Help?

A bit. For a consumer evaluating one app, probably not. For a developer with thousands of reviews across two stores, yes, because the useful signals (posting velocity, account age, near-duplicate text, rating clusters) are the ones a human can't eyeball at scale.

What a fake review checker can't do: prove any single review is fake, or remove one. Only Apple and Google can do that.

The consumer tools have mostly gone anyway. Mozilla shut Fakespot down in July 2025 and ReviewMeta disappeared around the same time, so for ordinary users it's back to the manual routine above.

On the developer side, AI Review Sensor pulls App Store, Google Play and Google Business Profile reviews into one inbox and flags spam as it lands. The feature I actually find useful is the spam and rating recovery view, which shows your rating with and without the flagged stuff. More than once that gap has been 0.05 stars, which is a polite way of saying: stop obsessing and go fix the onboarding bug people keep mentioning.

What to Do When You Find Fake Reviews

If they're on someone else's app, report the ones that break the rules and move on. Don't expect every report to stick.

If they're on your app, document first. Screenshots, timestamps, account names, a chart of reviews per day, side-by-side examples of copied phrasing. Moderation teams respond to patterns, not to "this one review is unfair."

Then report through Apple's or Google's official channels. Anyone promising guaranteed removal for a fee is lying.

If you reply publicly, don't call anyone a fraud, even when you're sure. "We couldn't find an account matching this experience, please reach out to support" does the job and tells the next reader you're paying attention.

And don't flag real negative reviews as spam because they sting. Someone can hate your pricing change and be entirely genuine. The FTC's Consumer Reviews and Testimonials Rule, in force since October 2024, explicitly covers fake and AI-generated reviews, and suppressing honest criticism is its own kind of problem.

Keep answering the real ones. A decent review analytics dashboard will show you which store is dragging your average and which complaint keeps coming back. That's worth more than knowing a handful of reviews look dodgy.

Bottom Line

One review tells you almost nothing. Fifteen reviews, read together, tell you nearly everything.

Look at bursts, not sentences. Click the profiles. Compare the stores. Check the dates against the changelog. Use a tool if you've got the volume, but don't mistake its score for a verdict.

And if someone's targeting your app, write it all down, report the pattern, and keep talking to the real customers. That's the part you control, and it's the part that pays off.