Pick up the most expensive serum on your shelf and turn it over. Find the hero ingredient on the front of the box, the one in the product name and printed in gold foil. Now find it in the ingredient list. If it comes after phenoxyethanol, you’ve just learned something the brand would prefer you didn’t know.
Phenoxyethanol is a preservative, and formulators typically use it at 0.5 to 1%. It’s a useful ruler because it sits right at the edge of the most consequential line in cosmetic labeling. Everything listed after it is very likely present at under 1%, sometimes far under. That’s where the snail mucin, the stem cell extract from a rare alpine flower, the copper peptide, and the “exosome-inspired complex” usually live. A formulation academy explainer published this month put it bluntly: at the very end of a list, these ingredients may be present at 0.01% or less, where they’re “decor, not active performance.”
The industry has a name for this. Cosmetic chemists call it fairy dusting or angel dusting, and they talk about it openly, because it isn’t a secret inside the trade. It’s only a secret on the shelf.
The rule that makes it invisible
U.S. cosmetic labeling runs on one federal regulation, 21 CFR 701.3. It requires ingredients to be declared “in descending order of predominance.” That sounds like transparency. Then you get to paragraph (f)(2), which says ingredients “present at a concentration of not more than 1 percent” may be listed “without respect to order of predominance.”
So the list is ordered, but only until the 1% mark. After that, a brand can arrange the sub-1% ingredients however it likes, and most will move the marketable ones up. An extract at 0.001% can legally sit ahead of a preservative at 0.9%. Nothing in the regulation requires a brand to tell you a single percentage. The FDA’s own labeling summary says the same: ingredients at or below 1% “may be declared” in any order after the ones above it.
The 2022 Modernization of Cosmetics Regulation Act, the biggest overhaul of U.S. cosmetics law in more than 80 years, didn’t change this. It added facility registration, product listing, adverse-event reporting, and safety substantiation. It didn’t require brands to prove a product works or to disclose how much of anything is in it. A serum named after copper peptides only has to contain some copper peptide.
What the labs actually find
Angel dusting is the deliberate version of the problem. The accidental version is just as common: products that claimed a real dose and lost it somewhere between the factory and your face. The difference matters less than you’d think, because the result on your skin is the same.
In August 2026, researchers at the University of Ljubljana’s Faculty of Pharmacy published one of the most careful independent audits of vitamin C skincare to date in the journal Cosmetics. They bought 41 leave-on facial products and measured four forms of vitamin C: ascorbic acid, ascorbyl glucoside, ascorbyl phosphate, and ethyl ascorbic acid. Thirty-one products printed a concentration. Only 57% of those measurements landed within 20% of the label. Forty-three percent came in under 80% of the claim, and one in five came in under half. Across all 41 products, 22% were mislabeled, mostly because the vitamin C form on the label was below the detection limit. In two products, the lab found a different, cheaper derivative from the one named on the box.
Pure ascorbic acid, the form most often marketed as the “real” one, held up worst. Measured content ranged from 0% to 116% of the label claim. The authors blamed degradation and poor formulation, and concluded that “stricter control and regulatory measures in the field of cosmetics are” needed.
Hong Kong’s Consumer Council ran a similar test in October 2024 on 17 vitamin C serums and essences. Twelve stated a concentration, from 6% to 15.2%. Ten of those 12 measured between 0.2% and 6% below their labeled value. One model had no detectable vitamin C at all. Retail prices ran from HK$101 to HK$1,580, and unit prices differed more than 14-fold. The most and least expensive models tied for the highest overall score.
Retinoids show the same pattern. A 2021 Journal of Cosmetic Dermatology study from the same Ljubljana group found that in more than half the products tested, the retinoid inside didn’t match the one on the label. Measured contents deviated from declared amounts by 0% to 130%. Two products contained retinoic acid, a prescription-only drug in many markets, and 20% exceeded EU safety committee recommendations. The label was wrong in both directions.
The twist: some things work at a trace
Here’s where most “read your labels” content gets it wrong. A tiny number isn’t automatically a scam.
The most-cited peptide trial in skincare, a 2005 Procter & Gamble study in the International Journal of Cosmetic Science, gave 93 women aged 35 to 55 a moisturizer on one side of the face and the same moisturizer with palmitoyl pentapeptide-4 on the other, twice a day for 12 weeks. The peptide was at 3 parts per million, or 0.0003%. The peptide side still showed significantly more improvement in fine lines and wrinkles on image analysis and expert grading. The effect was small, and the trial was industry-run. But it shows that signal peptides are meant to work at minute doses.
Growth factors are similar. One Korean brand publicly defends using 10 ppm of EGF as a high level for that ingredient class, and the argument holds up: you can’t compare a signaling protein to a botanical water at 700,000 ppm. Fairy dusting is relative. The right question isn’t “is this number small?” It’s “is this number anywhere near the dose that was actually tested?”
That question cuts both ways, because many ingredients sold as miracles were never tested at any dose that would let you answer it.
Where dusting hides best
Dusting works best on ingredients with famous names and thin evidence, because there’s no clear effective dose for anyone to fall short of.
Copper peptide is the clearest example right now. GHK-Cu has had a social media revival, and it’s showing up at the bottom of ingredient lists everywhere. An August 2026 review in Pharmaceutics describes the human evidence for topical GHK-Cu as “sparse, heterogeneous, and often chemically undercharacterized.” It notes that skin studies found retention of only a few percent of applied peptide. None of the human studies it reviewed confirmed how much intact copper complex actually reached the tissue. The one recent controlled human trial it highlights used a nominal 2% serum, in 18 people, for eyebrows. If your $180 serum lists copper tripeptide-1 after the preservative, it’s nowhere near 2%, and there isn’t much evidence for it at 2% anyway.
Tranexamic acid is the opposite problem: an ingredient with a real evidence base, which makes it worth dusting. A January 2026 meta-analysis in Dermatologic Therapy notes that some trials of topical TXA at 3% and 5% for 12 weeks showed no clinical efficacy for melasma, while a 20-person split-face trial found 3% TXA as effective as 4% hydroquinone. In other words, even 3% is contested. A “brightening serum” that lists TXA after the fragrance isn’t in that conversation.
Then there’s the long-list trick. A product with fifteen named actives has less room for any one of them to be meaningful. The same formulation explainer estimates each ingredient in a long active list is often present at 0.01% to 0.1%, below clinically significant levels. Long lists read as abundance. To a formulator, they usually mean dilution.
The country that asks for receipts
It doesn’t have to work this way. China’s National Medical Products Administration issued its Standard for the Evaluation of Efficacy Claims of Cosmetics on April 8, 2021, and it took effect May 1, 2021. Brands selling in China now have to back efficacy claims with evidence, and for many claims that means human testing. A product built around one buzzword ingredient has to show the claim holds up for the finished formula.
The U.S. has no equivalent. Here, the claim and the evidence don’t have to meet. A brand can name a product after an ingredient, cite a study that used fifty times the dose, and never say what dose the bottle contains.
How to read a label like a formulator
Find the 1% line. Look for phenoxyethanol, ethylhexylglycerin, sodium benzoate, potassium sorbate, xanthan gum, or fragrance. These are almost always used at 1% or less. Anything listed after them is probably under that too.
Check where the hero sits. If the ingredient the product is named after appears in the first five or six positions, that’s a good sign but not a guarantee. If it appears after the preservative, assume it’s there for the label.
Match the dose to the evidence. Niacinamide has studies at 2% to 5%. Much of the vitamin C research uses roughly 10% to 20% ascorbic acid. Signal peptides can work in parts per million. If a brand won’t say how much it uses, ask. If it still won’t answer, treat that as the answer.
Watch for trade-name percentages. “3% Matrixyl” or “5% peptide complex” usually means the supplier’s blend, which is mostly solvent and carrier, not 3% or 5% of the peptide itself. That isn’t dishonest on its own, but it isn’t the number it looks like.
Distrust the long list. Fifteen actives in a $40 serum isn’t generosity. It’s arithmetic.
Pay for stability, not story. For fragile actives like vitamin C, the bigger risk isn’t that the brand dusted it. It’s that what went in at the factory isn’t what’s left in month three. Opaque, airless packaging and stable derivatives are worth more than a higher number on the front.
None of this means expensive skincare is a con. Some of the best formulas on the market are expensive because of what’s in them. But the law lets a brand charge you for an ingredient’s name and give you almost none of the ingredient. The only defense is knowing where the line is.