Is Yuka Actually Accurate for Skincare? An Honest Breakdown
Yuka is mostly accurate at what it actually measures, but what it measures is narrow. Here's the methodology, where it gets things right, where it gets them wrong, and when to trust the score.
I get this question a lot, usually some version of "I scanned my favorite moisturizer and Yuka gave it a 38. Should I throw it out?" The answer is almost always no, and the reason is that Yuka is accurate at one specific thing and gets stretched into territory it was never built for. This post is the honest version of what Yuka does well, where it gets things wrong, and when to trust the score.
The short answer
Yuka is accurate at flagging the presence of certain ingredients on a public watchlist. It is not accurate at predicting whether a product will work for your skin, irritate you, or cause harm in real-world use.
The score you see is a verdict on the formula's ingredient list against a fixed set of risk categories. It is not a verdict on the product. Two very different things.
If you understand that distinction, Yuka is a useful screening tool. If you treat the score as the final word on whether a product is "good" or "bad," it will mislead you constantly.
How Yuka actually scores skincare
Yuka's methodology for cosmetics, published on their site, works like this:
- Each ingredient in the INCI list is checked against a list of "risk" classifications drawn from public sources. These include the EU's CMR (carcinogenic, mutagenic, reprotoxic) categorizations, the European Chemicals Agency endocrine disruptor watchlist, and allergen lists.
- Each ingredient is assigned a risk level: no risk, low, moderate, or high.
- The product score is dominated by the worst-rated ingredient, with adjustments based on whether other ingredients are organic certified.
- The result is a 0 to 100 score, color-coded green to red.
That is the whole algorithm. It is a deterministic lookup against a watchlist.
This is good engineering for the problem Yuka set out to solve: "tell me at a glance whether this product contains any flagged ingredients." It is bad engineering for a different problem most users assume Yuka is solving: "tell me whether this product is good for my skin."
Where Yuka is right
Let's give credit where it is due, because Yuka does several things well.
The watchlists are real. Yuka isn't making up the risk categories. CMR and endocrine disruptor lists are maintained by the European Chemicals Agency and other regulators. If a product contains a substance flagged on those lists, that is a fact, not Yuka's opinion.
Brand independence. Yuka does not sell skincare. It does not run an affiliate program tied to product scores. Whatever the algorithm's blind spots, the score is not bought.
Catching genuinely outdated formulations. Some products on shelves today still contain old preservatives like methylisothiazolinone (MI), which is a confirmed sensitizer with high allergy rates. Yuka flags these reliably. If you are sensitivity-prone and want to avoid the most common contact allergens, Yuka is a useful first filter.
Catching greenwashing. Brands marketing as "clean" while using a formula that is functionally identical to a drugstore product Yuka rates the same way. The score does not lie about what is in the bottle.
Speed. As a "should I even pick this up off the shelf" check at Target, scanning a barcode and getting a score in under a second is a real workflow. I use it for that.
Where Yuka is wrong (or at least, where it misleads)
This is the longer section, because this is where most of the confusion lives.
1. It scores the worst ingredient, not the formula
A product's overall score in Yuka is dominated by the highest-risk ingredient, regardless of concentration. If a moisturizer has 38 ingredients and 37 of them are well-tolerated humectants, emollients, and a clean preservative system, but the 38th is phenoxyethanol (which Yuka flags as "moderate risk"), the score plummets.
Phenoxyethanol is one of the most extensively studied preservatives in cosmetics. The EU regulates it at a maximum of 1% in finished products, and this limit was reaffirmed by the Scientific Committee on Consumer Safety as safe for use in cosmetics, including for children over 3 years old. At the concentrations used in real products (typically 0.5% to 1%), there is no credible link to harm in normal use.
Yuka does not weight by concentration. A formula with 0.8% phenoxyethanol gets the same flag as a formula with 1% phenoxyethanol gets the same flag as a hypothetical formula with 5% phenoxyethanol (which would not be legal in the EU anyway). The score does not distinguish.
2. It does not consider concentration
This is the big one. In skincare, concentration is everything.
A "1% retinol" serum and a "0.025% retinol" serum will trigger the same retinol flag. The 1% will probably wreck a beginner's barrier. The 0.025% is a gentle starting concentration most dermatologists recommend for first-time users. Yuka cannot tell them apart.
A vitamin C serum at 20% L-ascorbic acid is a high-actives product that should not be paired with other strong actives. A 2% derivative product is gentle and layerable. Same flag in Yuka.
This is why a product marketed for sensitive skin and a product marketed for advanced users can both score in the 70s. The score is not telling you the strength.
3. It does not consider formulation context
Two products can have the same ingredient list and behave very differently based on how they are formulated. Order of addition, emulsion type, particle size, encapsulation, and pH all matter.
A poorly formulated 0.5% salicylic acid product can be more irritating than a well-formulated 2% one. Yuka has no view into this. The INCI list is the only input.
This matters most for actives. The same retinol can be encapsulated for slow release (gentle) or unencapsulated (potent and faster-acting). Same name, very different effect on skin. Yuka treats them identically.
4. It does not know your skin
Yuka's score is the same regardless of who is asking.
Dry skin and oily skin have different needs. A product that is "fine" for one can be a barrier disaster for the other. Sensitive skin and resilient skin tolerate different concentrations of the same active. Acne-prone skin needs to avoid certain comedogenic emollients that work fine for dry skin.
There is no profile in Yuka. No skin type, no concerns, no climate, no current routine. The same green-rated product flashes up regardless of whether you are a 24-year-old with combination skin in Singapore or a 55-year-old with rosacea in Berlin. They probably should be using different products.
5. It does not consider routine context
This is the most consequential gap, because most skincare problems come from how products interact, not from any single product.
Yuka cannot tell you that the retinol you scanned (rated 88) and the AHA toner you scanned (rated 91) should not be used on the same night. It cannot tell you that your benzoyl peroxide cleanser is going to deactivate the L-ascorbic acid serum if you layer them. It cannot tell you that you have three products that all do roughly the same thing, and only one of them is needed.
Most reactions happen because of stacking, not single products. Yuka does not see the stack.
6. It can over-flag well-formulated K-beauty and Asian-market products
Many K-beauty and J-beauty products use ingredient combinations that are less common in Western mass-market skincare but are well-tested in Asian markets. Things like centella asiatica derivatives, traditional fermented extracts, and certain plant-based emulsifiers. Some of these get flagged or ignored by Yuka simply because the database is more thorough on European formulations.
This is not Yuka's fault, exactly. It is a coverage problem with the underlying ingredient classification lists. But if you use K-beauty heavily, your Yuka scores will look worse than the products warrant.
Three real-world cases where the score gets it wrong
I am going to keep these generic to avoid singling out specific brands, but these are patterns I see weekly.
Case 1. The drugstore moisturizer rated 38.
A common drugstore moisturizer with petrolatum, glycerin, ceramides, and a phenoxyethanol-based preservative gets flagged. The actual formula is well within the boundaries of what dermatologists routinely recommend for compromised barriers. The "low score" is from the preservative flag. Verdict: keep using it if it works.
Case 2. The "clean" moisturizer rated 95.
A "clean beauty" moisturizer using essential oils as a preservative system instead of phenoxyethanol scores green. Essential oils are some of the most common contact allergens in cosmetics. For someone with sensitive skin, the "low rated" drugstore moisturizer is safer than the "high rated" clean one. Yuka does not flag this trade-off because essential oils are not on the standard CMR or endocrine disruptor lists.
Case 3. The retinol that should not have been bought.
A 1% retinol marketed at TikTok-level intensity. Yuka rates it green because retinol itself does not appear on the relevant watchlists. A first-time retinol user buys it, applies it nightly, and ends up with a barrier breakdown two weeks later. Yuka had no way to communicate "this concentration is not for beginners" because Yuka does not know concentration.
When to trust the Yuka score, and when not to
Trust Yuka when:
- You want a fast yes or no on whether a formula contains anything on the major regulatory watchlists.
- You are screening unfamiliar brands at the drugstore for obvious red flags.
- You want to confirm that a "clean" brand is actually using cleaner preservation than a comparable mainstream brand. Sometimes they are. Sometimes they are not.
Don't trust Yuka when:
- You are deciding whether a product is right for your skin type.
- You are deciding whether a product is too strong for a beginner.
- You are trying to understand whether two products will work together.
- You are evaluating a K-beauty or J-beauty product. The coverage is weaker.
- You are using the score to decide whether to throw out a product that has been working for you. The score is not stronger evidence than your skin's actual response.
What to use when you need real answers
Yuka is one tool. It does one job. For the things it cannot do, here is what I'd reach for instead.
For ingredient lookup with formulation context: INCIDecoder is the most thorough single-ingredient resource on the open web. It tells you what an ingredient does, common concentrations, and known interactions, without a moralizing score. If you want to understand one ingredient deeply, this is where to go.
For interaction and routine questions: This is the gap I built HadaBuddy to fill. Scan your whole shelf, set up a skin profile, and the app builds a 7-day routine from what you own and warns you about ingredient stacking conflicts before they become a problem. K-beauty coverage is strong because that is what I personally use. Free tier covers scans, ingredient analysis, conflict warnings, and basic routine building.
For dupes and product comparison: SkinSort is genuinely the best at this. If your goal is to find a cheaper version of a product you like, start there.
For regulatory and safety deep-dives: Go straight to the source. The European Chemicals Agency and the Scientific Committee on Consumer Safety publish their actual evaluations. Most of what shows up in app scores is downstream of these documents.
FAQ
Is the Yuka app reliable?
It is reliable at the specific thing it does, which is flagging ingredients on certain regulatory watchlists. It is not reliable as a general "is this product good" verdict, because it does not consider concentration, formulation, your skin, or how the product fits with the rest of your routine.
Why does Yuka rate everything red or orange?
Because the algorithm is dominated by the worst-rated ingredient in the formula. If a single common preservative or fragrance allergen is present, the score drops sharply, even if 90% of the formula is well-tolerated. This is also why "clean beauty" formulas tend to score better even when they use essential-oil preservation systems that have higher allergy rates than the synthetic preservatives Yuka flags.
Does Yuka consider concentration?
No. This is the biggest single gap. A 0.025% retinol and a 1% retinol are scored the same way. A 2% L-ascorbic acid and a 20% L-ascorbic acid are scored the same way. Concentration is the most important variable in real-world skincare, and Yuka cannot see it.
Is Yuka biased against certain brands?
No, in the sense that the algorithm is the same for every product. Yuka does not have a financial relationship with the brands it scores, so the bias people sometimes perceive is the bias of the underlying watchlists, not commercial bias. The watchlists are skewed toward European regulatory categories, which means they catch some things (preservatives, common European allergens) more reliably than others (concentration, formulation strength, K-beauty-specific actives).
Should I throw out products that Yuka rates badly?
Almost never. If a product has been working for your skin and you've tolerated it for weeks, that is stronger evidence than any score. Use Yuka for screening unfamiliar products, not for revoking products you already trust.
What's a better alternative to Yuka for skincare?
There isn't a single replacement, because Yuka is doing several jobs at once. For ingredient lookup, INCIDecoder. For routine and interaction analysis, HadaBuddy. For dupes, SkinSort. For "is this brand greenwashing," Yuka is still useful as a quick check.
The bottom line
Yuka is accurate at flagging certain ingredients against a fixed list. That is real, useful information. It is not the same as accuracy about whether a product is right for you. Use the score as one input among several, not as the verdict.
If your skin is happy, the score does not matter. If your skin is unhappy, the score is rarely the explanation. The explanation is almost always upstream of the algorithm: concentration you didn't know about, two products fighting on the same night, a barrier that needed rebuilding before any active was a good idea. Yuka is not built to see any of that.
That is not Yuka's fault. It is a tool with a job. The mistake is asking it to do a different one.