Why Can the Same Ingredient Score Get Different Checker Warnings?
When two skincare ingredient checkers disagree, compare the matched name, numerical score, and warning category separately. In our interface test, the same cetyl alcohol score appeared under different warning categories. Neither screen established how a finished product would behave on someone's skin. A matching number is not necessarily a matching decision rule.
What did we compare?
On November 2, 2026, we entered short, synthetic name lists into ScanSkinAI's embedded comedogenic checker and SkinSort's pore-clogging checker. These were interface tests, not real product formulas or clinical experiments. We used the free text interfaces without logging in. The findings apply to those interfaces on that date; they do not evaluate ScanSkinAI's separate full Product Scan.
Can the same score produce different warnings?
The identical input was Water, Glycerin, Cetyl Alcohol. ScanSkinAI displayed cetyl alcohol with a score of 2/5 in its low-risk group. Its output also listed glycerin, but did not display a separate water entry. We counted two displayed ingredient entries, not three verified ingredient matches. An omitted entry is an unresolved coverage question, not evidence that the substance is harmless or harmful.
SkinSort's result for that same input reported three ingredients found. It listed water, glycerin, and cetyl alcohol, and marked cetyl alcohol as a potential pore-clogging ingredient with a comedogenic rating of 2. Its summary reported one potential trigger. The numerical cetyl alcohol scores therefore agreed in this test, while the category assigned to that score differed. Saved SkinSort result
| What we observed | ScanSkinAI embedded checker | SkinSort pore-clogging checker |
|---|---|---|
| Displayed cetyl alcohol score | 2/5 | 2 |
| Where cetyl alcohol appeared | Low-risk group | Potential pore-clogging group |
| Displayed ingredient entries | 2 | 3 |
This is narrower than saying one checker is right and the other is wrong. We observed an interface disagreement. We did not test either site's underlying research or the effects of cetyl alcohol on a person.
Was the embedded checker simply showing every input as low risk?
We checked a separate control input: Aqua, Glycerin, Isopropyl Myristate. ScanSkinAI displayed isopropyl myristate at 5/5 in its high-risk group and glycerin at 0/5 in its low-risk group. Aqua was not a separate displayed entry. This control shows that the embedded interface could produce a higher-warning result during our session; it does not validate the score as a prediction for a finished product.
Does a non-warning result mean both tools checked the same things?
Not necessarily. With Water, Glycerin, Jojoba Oil, ScanSkinAI displayed jojoba oil at 2/5 in its low-risk group, alongside glycerin. SkinSort reported three ingredients, represented jojoba oil as Simmondsia Chinensis Seed Oil, and reported no potential pore-clogging ingredients. That output demonstrates a name mapping on this test; we did not use it to infer an unshown numerical jojoba score. Saved SkinSort result
The useful comparison is the actual list of matched names plus the returned category. Two non-warning summaries can hide different coverage. Conversely, a shared score can sit beneath different warnings.
Which result should you keep?
Keep a dated record of the exact input, tool interface, displayed matches, scores, and categories. Preserve the unresolved parts. Do not rewrite the original label merely to obtain a preferred result.
Start a fresh comparison if a tool changes. These observations cannot establish future behavior, database completeness, or personal suitability. For the broader boundary between a label lookup and product evidence, see what a skincare ingredient checker can and cannot settle. If the copied text itself looks damaged, use the separate text-boundary preview.
