Cocoa Butter

Cocoa butter is one of the very few ingredients on this site with direct human evidence — and the human evidence disagrees with itself. Two rabbit assays put it at or near the top of their scales. Two human studies, both tiny, both on occluded backs, neither on a face, point in opposite directions. We previously wrote that the rabbit score had "for once been echoed on human skin," having read only the half of the human literature that said so. This page is the correction.

What it is

The fat pressed from the roasted, fermented seed of Theobroma cacao — the same fat that gives chocolate its snap. On a label it is Theobroma Cacao Seed Butter, or Theobroma Cacao (Cocoa) Seed Butter. It is solid at room temperature and softens at skin temperature, which is the whole reason formulators reach for it: it gives a balm or a body butter structure in the jar and slip on application, and it lays down an occlusive film that slows water loss. Its fatty acids, by gas chromatography (Padilla and colleagues, 2000), are palmitic 22.9–26.1%, stearic 34.7–37.3% and oleic 36.2–39.9% — three acids accounting for nearly all of it. A 2019 agronomy study in Frontiers in Plant Science found that both the fat content and the ratio between those acids shift with the tree's genotype, its pollination and its harvest season, so "cocoa butter" names a range rather than a fixed material. What that study does not show — and we previously implied it did — is that the variation changes how the fat behaves in a follicle. It is an agronomy paper. It was financially supported by Mars, and several of its authors were employed by Mars-related entities or by Guittard Chocolate; we disclose that here for the same reason we disclose it anywhere else. One thing worth saying about the family name, because we got it wrong before. "Butter" is a texture, not a chemistry — but that cuts less deeply than we made it cut. Shea butter, on the Cosmetic Ingredient Review's 2024 figures, is stearic 25.6–50.2%, oleic 37.1–62.1% and palmitic only 2.6–9%: compositionally it is closer to cocoa butter than the two ratings suggest, and the clearest difference between them is palmitic acid. Coconut butter is a third thing again, and we do not know exactly what. Cosmetic nomenclature treats Cocos Nucifera (Coconut) Seed Butter and Hydrogenated Coconut Oil as two separate ingredient identities, and Fulton — who scored "coconut butter" a 4 in the same table and the same section as cocoa butter — never says which one he had. So the label word tells you about the texture, and about very little else.

Why we rate it Caution

We rate it Caution at Moderate confidence. Start with the correction, because it is short. This page was live carrying a claim that the rabbit-ear score "has, for once, been echoed on human skin" — while omitting the human study that says the opposite. Omitting the contrary study turned inconsistent evidence into apparently unanimous evidence, which is the selective reading we accuse the lists of. The rating survives. The High confidence does not. The animal evidence, which is the part that agrees with itself. Fulton's 1989 paper in the Journal of the Society of Cosmetic Chemists scores cocoa butter 4 for comedogenicity and 0 for irritancy on a 0–5 scale — where his own key reads 0–1 "not considered significant", 2–3 "borderline", 4–5 "considered positive". The method was the rabbit ear: the ingredient at roughly 10% in propylene glycol, applied to the inner ear of New Zealand albino rabbits, five days a week for two weeks, three rabbits per assay, follicles scored for keratosis. His prose adds the clinical impression the score was meant to confirm: "Clinically, natural oils such as cocoa butter and coconut butter have long been known to cause problems with pomade acne. This is confirmed in the rabbit ear assay." That 4 is not an unqualified 4, and we lost the qualifier. The row sits under Fulton's section heading "VI. Oils*", and he defines that asterisk as "Results depend on source of raw material." The asterisk is on the section, so every oil and butter in it inherits the caveat — cocoa butter included. We had transcribed his rows without their headings, which is how a conditional 4 became a flat one in our own reference file. It is fixed there now. Ten years before Fulton, Kligman and Kwong tested cocoa butter directly, and this page omitted them entirely. Their 1979 paper in the British Journal of Dermatology compared a histological method against a whole-mount method on sixteen materials, two rabbits each, on a 0–3 scale. Cocoa butter scored 3 in every one of the four readings — both rabbits, both methods. On that scale 3 is the maximum, not a middling result, and writing it as a bare "3" next to Fulton's 4 would make the worse result look like the milder one. It is 3 out of 3. Note also what else that table says: isopropyl myristate, the reference positive of this entire literature and Fulton's only 5, scored 1 out of 3 in Kligman and Kwong's hands. The two rabbit assays agree that cocoa butter is near the top of the scale. They do not agree with each other about the substance the scale is calibrated on. Now the human evidence, both halves of it. In 2022, Baek and colleagues published a study in the Journal of Cosmetic Dermatology in which six participants received patches on the upper back containing cocoa butter and were imaged with reflectance confocal microscopy. The ratio of microcomedones to follicles at the treated site rose significantly above the untreated site by two weeks (p = 0.0419), and the microcomedones were significantly larger at two and four weeks (p = 0.0026 and p = 0.0038). That is what the accessible record supports and it is all of what it supports: the abstract and the institutional record do not disclose the concentration, the amount applied, the grade, the supplier or the vehicle, and the full Wiley text was not available to us. We previously wrote that the cocoa butter was tested neat and that it had been chosen because the authors expected it to work. We cannot verify either, and we have removed both. The paper's own stated conclusion is chiefly that reflectance confocal microscopy is a good way to detect microcomedones. And in 2015, Lee, Kim, Jung and Ha published a paper whose conclusion is in its title: "Isopropyl Myristate and Cocoa Butter are not Appropriate Positive Controls for Comedogenicity Assay in Asian Subjects." Eight Asian women, aged 18 to 45, with prominent follicular orifices and acne-prone upper-back skin. Cocoa butter (Biochemica) as supplied, 0.3 mL over 16 cm², three times a week for four weeks, under an occlusive patch, on the Mills and Kligman human model — with petrolatum and a patched site carrying no test material at all as controls. Microcomedone activity rose 47% at the cocoa butter site. It rose 55% under petrolatum. It rose 54% under nothing whatsoever. Isopropyl myristate, the positive control, rose 63%. On a Friedman rank sum test there were no significant differences between any of them. The honest reading of Lee is narrower than either side would like. It does not show that cocoa butter is safe. Every site in that study rose steeply, including the blank patch, which is a model that discriminated nothing — not the butter, and not the reference comedogen either. Eight people is very few, and a non-significant result in eight people is not evidence of no effect. It is one ethnic population, and the authors' own explanation is sebum: they measured 7.63 µg/cm² on their subjects' backs against the roughly 40 µg/cm² reported for Caucasian backs, and they call for larger studies in other populations. What Lee shows is that the human evidence on cocoa butter does not agree with itself. Two studies, six people and eight people, both on occluded backs, neither on a face, pointing opposite ways. That is why the rating is Caution and the confidence is Moderate rather than High. Our confidence tier is derived from the kind of evidence underneath a rating, and a High would mean human evidence that converges. Human evidence that exists and conflicts is a different sentence, and a weaker one. It is still more than almost anything else on this site has.

What the evidence doesn’t tell you

The dose story this page used to tell is gone, and the reason is worth setting out, because it was the page's central argument. We wrote that cocoa butter is "comedogenic neat and not comedogenic at under 1% in a powder," and that every real product sits between those two poles. Neither pole survives. Baek's accessible record does not say the material was applied neat, so the top pole is unverified. And the bottom pole is not a cocoa butter result at all. What Draelos and DiNardo actually tested, in their 2006 paper in the Journal of the American Academy of Dermatology, was a finished, multi-ingredient face powder: zinc oxide over 50%, mineral oil 25–30%, beeswax 1–5%, talc under 1%, cocoa butter under 1%. Six people per product group, patched on the upper back three times weekly for four weeks, microcomedones counted by cyanoacrylate biopsy. The powder raised microcomedone activity by 34%. The untreated site rose 28%. The positive control, octyl palmitate, rose 91%. The powder was judged non-comedogenic because it landed close to the negative control and well below the paper's 50% threshold — not because nothing happened. Our previous sentence, that the powder "did not raise microcomedone counts," was simply wrong. And no causal contribution from an ingredient present at under 1% can be separated from a zinc oxide that is more than half the formula. Draelos supports one proposition: a particular finished face powder containing less than 1% cocoa butter was non-comedogenic on that study's product-level criterion. It does not support the proposition that cocoa butter is non-comedogenic below 1%, and we converted the first into the second. Fulton, whose 4 we are arguing about, considered that move in the same paper the 4 comes from and refused it: "We are not convinced of the statement that lower concentrations of these compounds can be safely used with no comedogenic consequences." There is no dose-response curve for cocoa butter. Nobody we can find has measured one. Both human studies are on occluded upper backs, and neither is a face. Fulton argued that back skin is relatively insensitive compared with facial skin; Draelos and DiNardo describe upper-back occlusion as an exaggerated exposure that overstates ordinary use. We previously wrote that they cannot both be right. They can. Back skin can be intrinsically less responsive than facial skin at the same time as an occlusive patch increases penetration relative to a cream rubbed into a face. The two propositions are not incompatible; what is unsettled is how they net out, and how much a patched back predicts about a face. That is the real gap, and stating it as a contradiction was a cheap way of not stating it. Gala and colleagues (2021) are product-level evidence too, and we cite them as such. Twenty-two evaluable women, patched on the upper back for four weeks: a finished cream scored a mean comedone grade of 0.41 on a 0–3 scale, against 0.82 for glycerin and 2.09 ± 0.68 for coconut oil as the positive control. But the cream contained glycerin, dimethicone, shea butter, aloe butter, cocoa butter and mango butter; the cocoa butter concentration is not reported; and while the paper reports significance tests for coconut oil against glycerin and against the test product, it reports no direct test between the test product and glycerin. Every author is an employee of Dr. Reddy's Laboratories and the product is Dr. Reddy's own. It is a reasonable data point about a formula. It is not evidence about cocoa butter. The fatty-acid explanation everyone reaches for does not survive the paper it is quoted from. Fulton scores free palmitic acid 2 and free stearic acid 2–3 — both below cocoa butter's 4. If the palmitic acid were the mechanism, the palmitic acid should have outscored the butter made of it. It did not. His Table III shows why the reasoning is unsafe in either direction: the same free fatty acids move with their vehicle, palmitic scoring 0 in ether or acetone and 2 in sunflower oil, stearic 0 and 2. The fats in cocoa butter are esterified into triglycerides rather than free, and a mixture need not behave like its isolated components. That is an argument against the simple fatty-acid story, not proof that composition is irrelevant — which is a distinction we should hold to more carefully than the lists do. The family comparison also has to be walked back. We wrote that shea, cocoa and coconut butter "behave the same in a jar and differently on a follicle." We cannot say that, and our own shea butter page says why: we have found no comedogenicity assay of shea butter at all, in either direction. Nobody we can find has put it on a follicle. And then, correcting that, we wrote something else that was false. We said that of the three fats, "only cocoa has ever been tested." The same page contradicts that twice: Fulton scored COCONUT BUTTER a 4, in the same table and the same asterisked section as cocoa's 4. Two of the three were tested. What is true of coconut butter is that its IDENTITY is underspecified — Fulton never says which material he had — and an ambiguous tested substance is not an untested one. A correction is not a safe place. It is another set of claims, and it has to be checked like any other. What is actually true is narrower and duller — the three fats have different compositions, cocoa and shea less different than their ratings imply; cocoa and coconut butter each carry a Fulton 4, one of them for a material nobody can now identify; and shea has no assay we can find, in either direction. What has never been done at all. No study has applied cocoa butter to facial skin, or run for the months an acne cycle actually takes. Lee did enrol acne-prone participants — so "nobody has tested it on acne-prone skin," which this page previously said, is false — but the site was still the back, under a patch, for four weeks. Nobody we can find has tested whether the compositional variation the Frontiers study documents changes comedogenicity from batch to batch; Fulton's section asterisk asserts that the source of the raw material matters, but an asterisk is a caveat, not a measurement. And a 2025 review in JAAD Reviews, surveying the field from 1972 onward, concludes that no standardised comedogenicity test exists at all and that "non-comedogenic" remains an unregulated claim. That is still true of everything above.

Where you’ll see it

Body butters, hand and foot balms, lip balms and lipsticks, stretch-mark and belly creams, cleansing balms, solid moisturiser bars, and a great deal of hand-made and "natural" skincare. It appears on labels as Theobroma Cacao Seed Butter or Theobroma Cacao (Cocoa) Seed Butter. We are not going to tell you how much of it is in your product, because we do not know and neither does the label. This page previously read concentrations off ingredient-list position, saying cocoa butter is "often used near neat" in natural skincare and "far less common" in mainstream facial moisturisers. None of that was cited, and position on an ingredient list is not a dose measurement in any case. The only published concentration figure we have for a cocoa-butter product is Draelos's: under 1%, in one powder, tested once. The categories above are illustrative, not a measured prevalence. We have not audited the market and we are not going to imply a count we did not make. Two neighbours are separate entries here and should stay separate in your head. Coconut Butter carries the same Fulton 4, from the same asterisked section, and we have found no human study of it — though "it" is doing a lot of work in that sentence, because Fulton never says which coconut material he tested. Shea Butter has no comedogenicity assay we can find, by anyone, in either direction. Note the difference between those two and cocoa's own position, because it is the whole point of this site: one is tested but unidentifiable, one is unfound, and neither is a zero.

Sources

Each source says what it actually is. Several widely-cited “sources” for comedogenic ratings are republishing the same 1989 assay, and counting them as independent agreement is how a thin evidence base gets made to look thick.

1. Fulton JE. "Comedogenicity and irritancy of commonly used ingredients in skin care products." J Soc Cosmet Chem 40:321–333 (1989) The origin of the number every consumer list republishes, read off the paper rather than off a list quoting it. Table I scores cocoa butter 4 for comedogenicity and 0 for irritancy on a 0–5 scale (his key: 0–1 "not considered significant", 2–3 "borderline", 4–5 "considered positive"). Method — roughly 10% in propylene glycol, applied to the inner ear of New Zealand albino rabbits, five days a week for two weeks, three rabbits per assay. THE 4 IS QUALIFIED AND WE HAD LOST THE QUALIFIER: the row sits under the section heading "VI. Oils*", and Fulton defines that asterisk as "Results depend on source of raw material." The asterisk is on the section, so every oil and butter in it inherits it. The same table scores free palmitic acid 2 and free stearic acid 2–3, both below the butter made of them, and his Table III shows those acids moving with their vehicle. His conclusion also refuses the dose argument this page used to make: "We are not convinced of the statement that lower concentrations of these compounds can be safely used with no comedogenic consequences."
2. Kligman AM, Kwong T. "An improved rabbit ear model for assessing comedogenic substances." Br J Dermatol 100:699–702 (1979) A second rabbit assay that tested cocoa butter by name, on a DIFFERENT SCALE, and which this page omitted entirely until 12 July 2026. Sixteen materials, two rabbits each, histological sections compared against whole mounts, scored 0–3. Cocoa butter scored 3 in all four readings — both rabbits, by both methods. On a 0–3 scale that is THE MAXIMUM, not a moderate result, and it must never be copied into a score field as a bare "3" beside Fulton's 4. Read the same table for the thing that complicates it: isopropyl myristate, the reference comedogen of this whole literature and Fulton's only 5, scores 1 out of 3 here. The assays agree about cocoa butter and disagree about their own standard.
3. Baek JH, Ahn HJ, Koh JS, Kwon H, Shin MK. "Early detection of microcomedones induced by cocoa butter using reflectance confocal microscopy." J Cosmet Dermatol 21(7):3016–3021 (2022) Direct human evidence, and half of the reason this ingredient is unusual. Six participants received upper-back patches containing cocoa butter; microcomedones were counted by reflectance confocal microscopy. The microcomedone-to-follicle ratio rose significantly above the untreated site by two weeks (p = 0.0419) and the microcomedones were significantly larger at two and four weeks (p = 0.0026, p = 0.0038). Characterised strictly to what we could read: the full Wiley text was not accessible, and the abstract and institutional record disclose no concentration, amount, grade, supplier or vehicle. We previously asserted the material was tested neat and that it was chosen because the authors expected it to work; neither is verifiable and both are gone. The paper's stated conclusion is chiefly that RCM detects microcomedones well.
4. Lee E, Kim N, Jung A, Ha J. "Isopropyl Myristate and Cocoa Butter are not Appropriate Positive Controls for Comedogenicity Assay in Asian Subjects." J Cosmo Trichol 2(1):1000102 (2015) The study this page left out, which is the worst thing that was wrong with it. Direct human evidence pointing the other way, read by us in full. Eight Asian women, 18–45, with prominent follicular orifices and acne-prone upper-back skin; cocoa butter (Biochemica) as supplied, 0.3 mL over 16 cm², three times weekly for four weeks under occlusion, on the Mills and Kligman model; petrolatum and a patched site with no test material as controls; microcomedones read by cyanoacrylate biopsy. Microcomedone activity rose 47% under cocoa butter, 55% under petrolatum, 63% under isopropyl myristate — and 54% under nothing at all. No significant differences (Friedman rank sum test). Weigh it for what it is: every site rose steeply including the blank patch, so the model discriminated nothing; eight people cannot support an accepted null; it is one ethnic population, and the authors attribute the result to low sebum output (7.63 µg/cm² on their subjects' backs against the roughly 40 they cite for Caucasians). It does not show cocoa butter is safe. It shows the human evidence does not converge. Published in an open-access title under the DOI prefix 10.4172 and hosted by Hilaris; we read the full paper rather than the abstract, and every figure above is from its own tables.
5. Draelos ZD, DiNardo JC. "A re-evaluation of the comedogenicity concept." J Am Acad Dermatol 54(3):507–512 (2006) A human study of FINISHED PRODUCTS, and the source of the error at the centre of this page's previous version. Six subjects per product group, patched on the upper back three times weekly for four weeks, cyanoacrylate biopsies. The relevant product is a face powder: zinc oxide over 50%, mineral oil 25–30%, beeswax 1–5%, talc under 1%, cocoa butter under 1%. It raised microcomedone activity 34%, against 28% for the untreated site and 91% for the octyl palmitate positive control, and was judged non-comedogenic because it landed near the negative control and below the paper's 50% threshold. We previously wrote that it "did not raise microcomedone counts." It rose 34%. And nothing in it isolates a sub-1% ingredient from a zinc oxide that is more than half the formula. It supports the paper's actual conclusion — "finished products using comedogenic ingredients are not necessarily comedogenic" — and it does not supply a concentration threshold for cocoa butter.
6. Gala MYN et al. "Evaluation of comedogenic potential of a paraben-free plant-based butter moisturizing cream: A double-blind, comparative study." CosmoDerma 1:52 (2021) Industry-funded and disclosed as such: every author is an employee of Dr. Reddy's Laboratories and the product tested is Dr. Reddy's own. Twenty-two evaluable women, upper-back patches, four weeks. The cream scored a mean comedone grade of 0.41 on a 0–3 scale, against 0.82 for glycerin and 2.09 ± 0.68 for coconut oil as the positive control. Cited as product-level evidence and nothing more — the cream contained glycerin, dimethicone, shea, aloe, cocoa AND mango butter, the cocoa butter concentration is not reported, and no direct significance test between the test product and glycerin is reported. It also disproves a claim this page used to make: it applied coconut oil, an ingredient we rate, directly to human skin, so cocoa butter and petrolatum were never "the only two."
7. Mills OH, Kligman AM. "A human model for assessing comedogenic substances." Arch Dermatol 118:903–905 (1982) The human model both Baek and Lee are running — material under occlusion on the upper back of subjects with large follicles for a month, read by cyanoacrylate follicular biopsy — and an unresolved gap in the evidence behind this page. The abstract confirms the model and the general finding that materials moderately or strongly positive in the rabbit could induce comedones in humans, but it does not identify the substances or their individual scores. Lee's paper states that cocoa butter had previously been used as a human positive control and cites this paper for it. We could not obtain the full text and so cannot confirm whether cocoa butter itself was tested here, at what concentration, in how many people, or with what result. Listed as a gap, not as evidence.
8. Padilla FC, Liendo R, Quintana A. "Characterization of cocoa butter extracted from hybrid cultivars of Theobroma cacao L." Arch Latinoam Nutr 50(2):200–205 (2000) Gas-chromatography fatty-acid profile of cocoa butter — palmitic 22.9–26.1%, stearic 34.7–37.3%, oleic 36.2–39.9%. Food science, not dermatology. Cited for composition only, because a composition claim needs a composition source; it says nothing about pores.
9. "Identification of Climate and Genetic Factors That Control Fat Content and Fatty Acid Composition of Theobroma cacao L. Beans." Frontiers in Plant Science 10:1159 (2019) Agronomy. Cited for one thing and one thing only: genotype, pollination type and harvest season each shift cocoa butter's fat content and the proportions of its major fatty acids, so the material is a range rather than a constant. It does NOT show that this variation changes comedogenicity, and we previously wrote as though it did. Disclosure, applied here as we apply it elsewhere: the study was financially supported by Mars, and several authors were employees of Mars-related entities or of Guittard Chocolate. That does not invalidate an agronomy paper, and we are not going to disclose one funder's interest and stay quiet about another's.
10. Cosmetic Ingredient Review. "Safety Assessment of Butyrospermum parkii (Shea)-Derived Ingredients as Used in Cosmetics" — FINAL REPORT, released 5 October 2017 The industry safety dossier on shea, cited here purely for the family comparison: its fatty-acid table gives shea butter as stearic 25.6–50.2%, oleic 37.1–62.1%, palmitic 2.6–9%, which is closer to cocoa butter than the two ingredients' ratings suggest. Searched in full for "comedogen", "acne" and "rabbit ear": zero occurrences of any of them. That is the reason this page can no longer claim the butters "behave differently on a follicle" — the strongest negative we have on shea is that the document where an assay would live does not contain one. Two corrections to this citation, and they are the same correction twice. We were linking `shea042017tent.pdf` — the AMENDED TENTATIVE REPORT FOR PUBLIC COMMENT, released 28 April 2017, a document explicitly still open for comment. The final report came out on 5 October 2017 and is what we link now. This is the SECOND time we have cited a comment-stage CIR document as settled; the first was the brown-algae draft stamped "Do Not Cite or Quote", and on that one the provisional number was wrong. We were also citing this under a 2024 Int J Toxicol reference we had never opened. We cite the 2017 final report because that is the document we actually read.
11. Cosmetic Ingredient Review — Hydrogenated Coconut Oil Cited for a nomenclature point we previously got wrong. Cosmetic ingredient nomenclature treats Hydrogenated Coconut Oil and Cocos Nucifera (Coconut) Seed Butter as separate identities, so "coconut butter is hydrogenated coconut fat" — which this page asserted — is not a safe equivalence. Fulton, who scored a "coconut butter" 4 in the same asterisked section as cocoa butter, gives no specification of what his material was.
12. "Comedogenicity in cosmeceuticals: A review of clinical relevance, regulatory gaps, and future directions." JAAD Reviews (2025) A recent peer-reviewed review of the whole field. Finds that rabbit-to-human translation is inconsistent, that no standardised comedogenicity test exists, and that "non-comedogenic" is an unregulated label. Independent of Fulton in authorship — it is a literature review, not an independent replication of anything. Cited for the field-wide caveat.
13. acne.org comedogenic ingredient list Where most readers will have met the number 4. Characterised accurately, which we previously did not do: it presents itself as a synthesis of eleven published comedogenicity studies, assigning categories from multiple rabbit assays or at least one human test — not a straight republication of Fulton's table. It is still a synthesis of the same small legacy corpus rather than an independent measurement, and it is not a separate agreeing source. We have not traced the provenance of CosDNA, Face Reality, Sofie Pavitt or Banish ingredient by ingredient, and matching numbers are not proof of a citation genealogy, so we no longer assert that all five are one assay quoted five times.

Others in the same family

Butters behave similarly enough that the evidence for one is often wrongly read across to the others. These are its structural relatives, not a guess. What the evidence says about butters as a family ›

• Coconut Butter Clogging Low
• Shea Butter Clear Low

Last reviewed 2026-07-12 · How we decide

‹ Check a full ingredient list