'Clean', 'Vegan', 'Non-Comedogenic': Three Labels That Mean Three Different Amounts of Nothing

Somewhere in the 1970s a dermatologist shaved the inside of a rabbit's ear, painted a cosmetic ingredient on it for two weeks, cut the ear off, dropped it in 60°C water for two minutes and peeled the epidermis away in a sheet to count the plugged follicles underneath.

That is where "non-comedogenic" comes from. Not a clinical trial, not a regulator, not a standards body. A rabbit ear assay, refined and published by Kligman and Kwong in 1979, and every ingredient blacklist that circulated through skincare forums for the next forty years traces back to it.

We went looking for whatever sits behind the other two words. Vegan has one non-binding EU sentence. Clean has nothing at all.

Our own claim chips, counted

Fifty-four products carry 206 free-of strings between them, in 44 distinct wordings. Paraben-free appears 46 times, fragrance-free 28, alcohol-free 19, silicone-free 18, sulfate-free 17.

"Clean" appears zero times. So does "natural", so does "organic", and so does "non-comedogenic". That is not restraint on our part so much as an accident of which Korean brands we stock, and the word clean does turn up in three of our product titles, where it describes what the thing does to your face rather than what the formula is made of.

The 44 distinct strings are also a mess we should own. Two of our chips on the GDS hair dye shampoo read "Ammonia" and "Parabens" with no "-free" on the end, which on a claim chip reads as the opposite of what was meant. "Weakly acidic" exists in three non-identical variants. "Single-ingredient" and "Single-use" are sitting in a free-of field while describing something that is not an absence at all. We flagged eight fragrance-free claims that were not true two days ago; this is the same field failing in a duller way.

Non-comedogenic began on a rabbit's ear

The assay is not a scandal. It was a reasonable idea in 1979: rabbit follicles plug easily and visibly, so you get a fast readout on whether a substance drives comedone formation. Fulton and colleagues used it to score a long list of cosmetic raw materials in 1984, and that paper is the ancestor of every "avoid isopropyl myristate" post you have ever read. Isopropyl myristate, isopropyl palmitate, butyl stearate, myristyl myristate, decyl oleate, acetylated lanolins and a run of D&C Red dyes all came off that list.

Read the paper and Fulton says the quiet part himself. The data, he writes, "were gleaned from the rabbit ear assay, which is not an ideal animal model but is the best we have."

The same paper contains a result that should have ended the ingredient-list approach on the spot. Several acne treatments scored comedogenic in the assay, including a tretinoin cream and benzoyl peroxide products, and the authors' response was to suggest those products be reformulated. A test that flags benzoyl peroxide as acne-causing is telling you something about the test.

The people who built it said it was too sensitive

Mills and Kligman published a human version in 1982, applying substances under occlusion to the upper back for a month in young men with large follicles. Their conclusion, in their own words: "The rabbit model is more sensitive than the human." And then the practical corollary, which is the sentence the ingredient blacklists dropped: "Substances that are weakly comedogenic in the rabbit are probably safe for human use with the possible exception of acne-prone persons."

So the assay was already understood by its own authors to over-call, and they said so in print in 1982. The ingredient blacklists that spread afterwards carried the scores across and left that sentence behind.

Then Draelos and DiNardo went at the underlying assumption in 2006. They noted that animal models had been used on raw materials "with the assumption that finished formulations containing these ingredients would also be comedogenic," took six subjects who reliably form comedones on the upper back, patched finished products three times a week for four weeks and counted microcomedones by cyanoacrylate biopsy. Their conclusion is one line long: "Finished products using comedogenic ingredients are not necessarily comedogenic."

Six people is not many. We could not get past the paywall to see which products were tested or what the counts were, so we have the design and the conclusion and no numbers, and we are not going to pretend otherwise.

A raw material that scores clean on the rabbit is very likely fine on you. A raw material that scores dirty may be fine at 0.5 percent in a lotion and was never tested that way. And a finished product labelled non-comedogenic has been tested by whatever method its manufacturer chose, or by none, because no regulator anywhere defines the word.

Vegan is an absence claim with no auditor

The EU's own guidance on cosmetic claims tells you what the vegan claim legally is. Its permitted form is "free from animal-derived ingredients, e.g. in products intended for vegans." An absence claim. The same document opens by stating, in capitals, that it "is not a European Commission document" and that its views "are not legally binding."

Korea is blunter. The Cosmetics Act as it stands in 2026 defines thirteen terms, and vegan is not one of them. What makes this newly interesting is that two of those definitions were deleted in January 2025 — the ones for natural cosmetics and organic cosmetics — and the transitional provision that came with the amendment is headed, in plain language, "transitional measures on the abolition of the certification system for natural and organic cosmetics." Korea built a real quantitative standard, 95 percent natural content by weight, 10 percent organic, and then took it apart and handed the space back to private certifiers.

Seven of our products carry a vegan chip, six saying "Vegan" and one saying "Vegan-friendly."

We pulled all seven ingredient lists. Six came back clean of anything plainly animal-derived. The seventh is the DEEPEER Premium Perfume Repair Treatment Hair Pack, and its 74-ingredient list contains hydrolysed collagen at position 36, hydrolysed silk at 43, propolis extract at 45, hydrolysed keratin at 51 and hydrolysed elastin at 53. Our own product copy advertises them by name in the same breath as the chip.

That is not a grey area. Silk is spun by insects. Propolis is made by bees. Keratin, elastin and unqualified collagen are animal proteins, so the claim on that product is wrong. It is ours, we wrote it, and the fix is a copy change rather than a debate.

What our screening found

The practical takeaway

Treat non-comedogenic as a statement about intent, not a result. It usually means a formulator avoided the Fulton list, which is a defensible thing to have done and tells you nothing about the finished product on your skin. If you break out in closed comedones, the ingredient blacklists will send you chasing isopropyl myristate through 30 labels and the 2006 work says that hunt is aimed at the wrong level.

Treat vegan as a fact you can check yourself, because it is the only one of the three that is checkable. It is a claim about what is absent, so read the list. Collagen, elastin, keratin, silk, carmine, beeswax, honey, propolis, lanolin, snail mucin. Ten words, thirty seconds. We found a false claim in our own catalog that way, and the same method works on ours as well as anyone's.

Treat clean as noise.

The one thing worth internalising from the Korean substantiation notice is the shape of it. Claims that make a specific promise about your skin get a specific evidence burden. Claims that describe a worldview get none, which is exactly why the second kind multiplied. We laid out the tiers for the rest of the vocabulary in the label-claims piece, took apart a plant-derived ingredient claim that borrows a clinical word in the clinic-to-shelf piece, and every product that got through our screen is listed on the awards page.