Scan the same moisturizer with two different ingredient apps. One says 95/100. The other says 30/100. Same product, same ingredients, completely different answers.
That's not a bug. The apps just disagree on what a score is supposed to mean.
Here's what's happening behind the scenes, including the parts most apps would rather you didn't think about.
Every ingredient app makes four calls
These choices look small. They're not. They're the reason one app tells you your moisturizer is great and another tells you to throw it out.
1. What counts as "harmful"
This is where most apps go wrong, and they go wrong in both directions.
Some apps over-flag. They penalize anything that's ever raised a concern in any study, at any dose. Phenoxyethanol gets dinged because of studies run at concentrations far higher than the under-1% used in skincare, where it's actually one of the gentlest preservatives available. Flagging it at that dose is like warning someone off water because you can drown in it.
But under-flagging is just as much a failure, and the "everything's fine, relax" apps fall into it. Some ingredients are flagged because they cause real reactions at the concentrations products actually use. Not in theory, in dermatology clinics. Fragrance and methylisothiazolinone (MIT) are leading causes of allergic contact dermatitis. Waving those away in the name of "dose and context" is the same mistake as panicking over phenoxyethanol, just pointed the other way.
The skill isn't being strict or being relaxed. It's telling those two situations apart.
2. What sources they trust
Strong sources: CIR reports, SCCS opinions, peer-reviewed dermatology and toxicology research, regulatory filings from the FDA or the EU's CosIng database.
Weaker sources: activist databases that list "concerns" without weighing study quality, and old studies that never held up. Some apps still lean on a 2004 study that found parabens in breast-tumor tissue. That study had no control group and only showed parabens were present. It never showed they caused anything. It's still cited as settled science. It isn't.
But here's the honest part most "myth-busting" content skips: that doesn't mean all parabens are fine. Short-chain ones (methyl-, ethylparaben) are well studied and well tolerated. The EU restricted propyl- and butylparaben and banned several longer-chain ones over endocrine-disruption concerns. So "parabens are dangerous" and "parabens are harmless" are both lazy. You have to read the evidence ingredient by ingredient instead of picking a team.
3. Single product or full routine
This is the big one, and almost every app gets it wrong by ignoring it entirely.
Most apps score one product at a time, in isolation. Useful for a quick check. But it misses what matters most for real skin: how your products behave together. Are you layering retinol with strong AHAs? Doubling up on the same active across two products? Putting on something that's fine alone but gets compromised by what went on five minutes earlier?
Communities call this "routine creep." You add products one at a time, and nobody's watching the whole system. A single-product scanner can't catch it.
One caveat we'll make that the marketing version wouldn't: an ingredient list can't see you. It doesn't know your tolerance, how often you use each product, or the state of your skin barrier. So routine analysis should surface interactions worth watching, not hand down verdicts. Retinol-plus-AHA is a flag to pay attention to, not proof you've done something wrong. Anyone selling you certainty from a label alone is doing the same false-precision trick as the apps that panic over a preservative.
4. Who pays
Some ingredient apps make money through affiliate links. They recommend products, you buy them, the app takes a cut. Others sell premium tiers where the "better alternatives" happen to be products they have deals with.
When revenue depends on recommendations, every score sits downstream of a business interest. The app that gets paid when you buy Product A has a quiet reason to score Product A above Product B. Most users never notice.
So here's our hand, face up. Yarai makes money from a paid Pro subscription. We don't take ads or brand sponsorships, and our scoring algorithm is independent of any commercial relationships. A product's score comes from its chemistry, not from who pays us. We have a business interest, like everyone in this space. The difference is that our interest points at making scoring you can trust, not at selling you a specific bottle.
How most apps fall on these four questions
The patterns are pretty consistent.
The most popular scanners use a strict precautionary framework, lean on activist-database sources, score single products only, and earn from affiliate recommendations. Users like the conservative feel. Cosmetic chemists keep pointing out that it penalizes well-studied, low-risk ingredients right alongside genuinely concerning ones.
Other tools are reference databases first. Detailed, careful entries with cautious language about evidence. Great if you want to do your own research. Most people don't want a 400-word essay on dimethicone, though. They want an answer.
Others flag ingredients on theoretical concerns at non-cosmetic doses. Fine if your goal is to minimize any conceivable risk regardless of probability. Less useful if you want to know what's happening in the bottle you actually own.
Yarai is built around the four problems above: dose-and-context scoring, peer-reviewed and regulatory sources behind the algorithm, routine-level flags, and a scoring algorithm that's independent of any commercial relationships. And to be clear about the limits: those routine flags are starting points, the AI behind them is a tool and not a diagnosis, and none of it replaces knowing your own skin or talking to a dermatologist. An app that won't tell you what it can't do isn't being rigorous. It's selling.
What this means for you
A bad score doesn't always mean a bad product. Plenty of flagged ingredients are flagged for theoretical reasons at doses you'll never meet in a face cream.
A good score doesn't always mean a good fit, and it definitely doesn't mean an ingredient is safe for you if you're sensitized to it. Safe in isolation and right for your skin are different things.
The same product can be right for one person and wrong for another. Skin type, sensitivity, pregnancy, what else you're using. No single number captures that, and no app should pretend otherwise.
Bottom line
If you're choosing an ingredient app, four questions tell you most of what you need to know:
- Does it score by dose and context, or panic at any concentration?
- Does it weigh real research, or aggregate activist claims?
- Does it look at one product, or your whole routine?
- And does it make money from what it recommends?
That's the test we built Yarai to pass, limits and all.