Can you trust app store ratings for AI photo apps?

Less than in most categories. These apps are young, the rating counts are tiny, and a handful of reviews moves the average by a full star — so the number you are reading may rest on fewer than a hundred people. Read the written reviews instead; ratings without text carry almost no information.

Why does the same app have different ratings on Android and iPhone?

Partly real — different builds break differently — but a gap of two whole stars is usually about who is reviewing rather than what they used. Check both stores before you decide, and treat the lower one as the more informative number, because unhappy people write and satisfied people mostly don't.

How do I check an AI photo app if it has no reviews at all?

The same way you should check one that does. Is the price visible before you sign up, is there a real example image for each scene, does the free offer run the real pipeline, does the tool say anywhere what it refuses to change about you. A new app has no rating and that tells you nothing — the four checks work regardless.

1.8 on Android, 3.9 on iPhone: The Same App

A phone showing an app store listing with the star rating obscured

We looked up an AI dating-photo app this week and found two numbers for it.

On Google Play: 1.8 stars from 90 ratings. On the App Store: 3.9 stars from 37.

Same app, same company, same week. If you found it through an iPhone you would conclude it was mediocre but fine. If you found it through an Android you would not install it.

Both numbers are real. Neither is useful. This is worth understanding, because "check the reviews" is the standard advice for this category and in this category it is the weakest tool you have.

Why star averages fail here specifically

The counts are tiny. Ninety ratings. Thirty-seven. These are new apps in a young market, and at those volumes a dozen reviews move the average by a full star. You are not reading the wisdom of a crowd, you are reading a small room.

The two populations are not the same people. Store ratings are a self-selected sample to begin with — the angry write, the satisfied mostly don't — and the sample differs by platform, by country, and by how aggressively an app asks for reviews and when it asks.

And the prompt timing decides the number. An app that asks for a rating the moment a generation finishes gets a different average from one that asks after you have seen the bill. Neither is lying. The number simply measures when you were asked.

The thing that actually told us something

Here is what changed the picture, and it took two minutes.

On the Play listing, only 19 of those 90 ratings had any text. Eighteen of the nineteen were one star. One was two stars. There was not a single four- or five-star review with a written word in it.

That is a much stronger signal than 1.8, because a written review costs effort. Every person who cared enough to type had the same complaint.

And on the App Store, where the average was a respectable 3.9, the written reviews split geographically: the one-star reviews were spread across the US, Australia and Israel, and every five-star review sat in a single country's storefront, clustered — "Dates every week now lol", "Best app to ever be created", "i'm pulling all kinds".

We cannot tell you why from the outside, and we are not going to assert what we cannot prove. But that is not the shape organic reviews have, and the written reviews disagreed with the star average on both stores.

A new app has no rating, and that tells you nothing either

The opposite case deserves the same scepticism, and we should say so plainly because we are in it.

CMeIn's iOS app has zero ratings. It went live in August 2026 and nobody has rated it. That is not evidence of anything — not that it is good, not that it is bad. It is evidence that it is new and that we have not asked.

Treat an empty listing the way you should treat a glowing one: as an absence of information rather than a verdict. A rating you cannot trust and a rating that does not exist leave you in exactly the same position, which is why the checks below do not involve ratings at all.

Five checks that work either way

Each of these takes under two minutes and none depends on anybody else's opinion.

Is the price visible before you sign up? Open the site or the listing and try to find what it costs, without creating an account. If the answer requires you to upload photographs first, that ordering was chosen. The most common complaint across every one of these apps is some version of "it only tells you at the end that you have to pay" — and you can detect it before it happens to you.

Is there a real example image for the specific thing you want? Not a general gallery of beautiful strangers. The actual scene, ideally matched to your gender. A tool that will not show you what a scene produces is asking you to pay to find out.

Is the free offer the real product, or a different one? Free credits that run the identical pipeline, just fewer times, tell you something true about how the tool handles your face. A separate "free tool" running a cheaper model tells you nothing — we wrote about that here, including the company that admitted its free tool "works very differently from our actual model".

Does it say anywhere what it refuses to change about you? This is the check almost nobody performs and it is the most predictive one. Most sites in this category say nothing about preserving your age, your build, your skin tone or your face shape — and silence means the default is in charge.

Is it one product or a bundle? An app that analyses your profile, writes your messages and generates your photographs is charging you for three things to sell you one. Check which of the three the subscription is actually priced around.

And when you do read reviews, read them properly

Skip the average. Sort to the one- and two-star reviews and read only the ones with text. Then count the complaints rather than weighing their tone, because the pattern is the information:

If most of them say likeness — it didn't look like me — that is a technical limit and it will apply to you too.
If most of them say billing — charged after cancelling, paywall at the end, "unlimited" that wasn't — that is a decision somebody made, and decisions repeat.
If most of them say it didn't work at all — no images generated, can't log in — check the dates. Clustered means an outage. Spread over months means the product.

Check both stores while you are there. They frequently disagree, and the lower one is usually the more informative number.

What we do, including where we fail this test

We would rather you ran the five checks on us than take our word for it, so here is how we come out.

Price before signup: yes. Credit packs from $2.99, a photograph costs 300 credits, and the pricing page does not require an account.

A real example image per scene: yes, matched to your gender where we have one, shown before you choose.

Free offer is the real pipeline: yes — same model, same reference selection, same instructions, fewer times. And it is a launch offer capped by number of recipients, not a permanent free tier.

What we refuse to change: published. The instruction sent with every generation reads do not beautify, redesign, age-shift, slim, widen, or change the face, and names the traits it holds — face shape, hairline, skin tone, age, body proportions.

One product, not a bundle: yes. We make photographs. We do not write your messages.

And the check we currently fail: reviews. Zero ratings on iOS, nothing meaningful on Play. If your rule is "never install something nobody has rated", we do not clear it, and we are not going to pretend that is a sophisticated position rather than simply where a new app stands.

The short version

Star averages in this category rest on tiny, self-selected samples and routinely differ by two stars between the two app stores for the same product. The written reviews are worth more than the number, and a new app's empty listing tells you nothing at all.

Check the five things that do not depend on anybody's opinion: price before signup, a real example for your scene, whether the free thing is the real thing, whether the tool says what it won't change, and whether you are buying one product or three.