We checked 97 top-ranking pages across 17 common acne searches against the 2024 American Academy of Dermatology acne guideline. We expected to find widespread mis-statement of how strongly treatments are recommended. We found something different, and more useful: almost no page engages with the guideline at all, and the one mis-statement that does circulate can be traced to a sentence in the guideline’s own published abstract.
What we set out to test, and what we actually found
On 12 September 2026 we corrected six passages across five articles on this site that described oral isotretinoin as something the AAD guideline “strongly recommends.” It does not. Isotretinoin for severe acne is recommendation 4.1, one of five good practice statements — consensus positions the work group chose not to grade. It is not one of the guideline’s 18 graded recommendations.
We had taken that phrasing from secondary summaries. That raised an obvious question: how many other people have taken it from the same place? So we built a sample and checked.
The answer to the question we asked turned out to be “very few, because almost nobody mentions the guideline.” The answer to the question we should have asked — where does the phrasing come from? — turned out to be considerably more interesting.
Finding 1: the guideline’s abstract says five good practice statements, then lists four
The abstract of the 2024 guideline states: “This guideline presents 18 evidence-based recommendations and 5 good practice statements.”
Later in the same paragraph it enumerates them: “Combining topical therapies with multiple mechanisms of action, limiting systemic antibiotic use, combining systemic antibiotics with topical therapies, and adding intralesional corticosteroid injections for larger acne lesions are recommended as good practice statements.”
That is four items. The fifth good practice statement is isotretinoin — and it has already appeared, three sentences earlier, in its own sentence placed between the strong-recommendation list and the conditional-recommendation list:
“Oral isotretinoin is strongly recommended for acne that is severe, causing psychosocial burden or scarring, or failing standard oral or topical therapy.”
In the guideline’s own recommendation table, that same recommendation carries “Good practice statement” in the strength column. The table contains 23 numbered recommendations: 18 graded and 5 ungraded. Because 4.1 is ungraded, it cannot be a graded strong recommendation — every one of those is inside the 18.
Two things make this more than a quibble. First, every other clause in that abstract paragraph names a GRADE category precisely — “Strong recommendations are made for…”, “Conditional recommendations are made for…”, “…are recommended as good practice statements.” In that context “strongly recommended” reads as the technical term, not as emphasis. Second, the abstract gives a count, which invites the reader to reconcile it. A reader who tries will find four good practice statements and one orphaned therapy sitting in the strong-recommendation slot.
The AAD’s own press release about the guideline lists all five good practice statements correctly, isotretinoin included. The issue is confined to the abstract — which is also the part of the paper that gets mirrored, indexed and paraphrased everywhere.
Does the 2026 correction fix it?
A correction to the guideline was published in July 2026 (J Am Acad Dermatol 2026;95(1):347), e-published 11 April 2026. We could not read its text — it sits behind the publisher’s paywall — so we cannot say what it corrects.
What we can say is that the abstract record held by the National Library of Medicine was last revised on 7 June 2026, after that correction was published and after the erratum was linked into the record, and it still reads exactly as quoted above. If you can read the correction, we would like to hear from you.
Finding 2: hardly any page people actually read mentions the guideline
This is the finding that surprised us most. The AAD guideline is the primary US clinical standard for treating acne. Across the pages a searcher actually lands on, it is close to invisible.
| Measure | Pages | Share |
|---|---|---|
| Pages retrieved and coded | 97 of 100 | 97% |
| Mention an AAD acne guideline at all (named or linked, any edition) | 13 | 13% |
| … of which clearly the current 2024 edition | 8 | 8% |
| Link to the primary source (the 2024 JAAD paper, its PubMed record, or the AAD guideline page) | 5 | 5% |
| Make any claim about recommendation strength or evidence certainty | 10 | 10% |
| Display a medical-review signal (named clinician reviewer or author, or a “medically reviewed” badge) | 38 | 39% |
One of our 17 queries named the guideline directly (“AAD acne treatment guidelines 2024”). Unsurprisingly, all four retrieved results for it discussed the guideline. Take that query out and the picture is starker: 9 of 93 pages (10%) mention an AAD acne guideline, and 3 (3%) link to the current one. Three of those nine still point to the 2016 edition, which the 2024 guideline replaced.
On 10 of the 17 queries, not one retrieved page referenced the guideline at all. Those ten include “hormonal acne treatment best options for adult women,” “birth control pills for acne which ones work best,” “spironolactone for acne,” “benzoyl peroxide for acne” and “isotretinoin accutane for severe acne” — all treatments the guideline makes explicit recommendations on.
| Type of site | Pages | Mention guideline | Link primary source |
|---|---|---|---|
| Health publisher | 25 | 5 (20%) | 2 (8%) |
| Retailer or brand | 24 | 2 (8%) | 0 (0%) |
| Dermatology practice | 16 | 1 (6%) | 1 (6%) |
| Telehealth or online pharmacy | 14 | 3 (21%) | 1 (7%) |
| Clinic or hospital | 10 | 0 (0%) | 0 (0%) |
| Government, education or professional body | 7 | 1 | 0 |
| News or trade press | 1 | 1 | 1 |
Carrying a medical-review badge made no difference. Reviewed pages mentioned the guideline 13% of the time (5 of 38); unreviewed pages 14% (8 of 59). Medical review, as practised on the consumer web, does not appear to mean “checked against the clinical guideline.”
Finding 3: not one strong recommendation in the guideline rests on high-certainty evidence
Under GRADE, the strength of a recommendation and the certainty of the evidence behind it are two separate axes. A strong recommendation can rest on weak evidence when the benefits clearly outweigh the harms; a conditional recommendation can rest on excellent evidence when the trade-offs are genuinely close.
Laid out as a grid, the 2024 acne guideline shows a pattern nobody seems to have pointed out:
| Strength | High certainty | Moderate | Low | Very low | Total |
|---|---|---|---|---|---|
| Strong for | 0 | 7 | 0 | 0 | 7 |
| Conditional for | 3 | 4 | 3 | 0 | 10 |
| Conditional against | 0 | 0 | 1 | 0 | 1 |
| Good practice statement (ungraded) | not graded | 5 | |||
| Total | 3 | 11 | 4 | 0 | 23 |
Every one of the seven strong recommendations — benzoyl peroxide, topical retinoids, topical antibiotics, the three fixed-dose combination products, and oral doxycycline — rests on moderate-certainty evidence. Not one rests on high certainty.
The only three high-certainty items in the entire guideline are all conditional: topical clascoterone, oral sarecycline, and the choice between standard and lidose isotretinoin. So the treatments with the best evidence behind them are not the ones most strongly recommended, and the ones most strongly recommended do not have the best evidence. That is not a flaw — it is GRADE working as designed, weighing cost, access, harms and patient values alongside the trials. But it is the opposite of what “strong recommendation” sounds like to a general reader, and it is why we keep the two columns separate on our acne treatment evidence map.
What we got wrong
We are reporting this part because a study that only publishes its successes is not worth much.
Our original hypothesis was that a substantial share of top-ranking pages would inflate the guideline’s recommendation strengths. That hypothesis was wrong, and the data says so plainly. Ten of 97 pages make any claim about recommendation strength or evidence certainty; the rest recommend treatments without ever touching the question. When we read each of the ten against the guideline:
- Seven were accurate. Doxycycline strong, minocycline and clascoterone conditional, salicylic and azelaic acid conditional, clascoterone resting on high-certainty evidence — all stated correctly.
- One repeated the abstract, not an error of its own. A clinical guideline-summary site reproduces the abstract word for word, including “Oral isotretinoin is strongly recommended…”, while its own recommendation table on the same page codes isotretinoin correctly as an ungraded consensus statement.
- One was out of date rather than wrong. A drug-pricing site’s 2022 article, citing the 2016 guideline, says the AAD “doesn’t recommend for or against” salicylic acid. The 2024 edition gives it a conditional recommendation.
- One was about a different guideline. A UK clinic describes the quality of evidence for spironolactone in the context of UK national guidance, not the AAD.
Our keyword search also flagged a string of marketing phrases — “clinical-grade”, “medical-grade”, “derm-grade” — as possible evidence-quality claims. None of them are, and we discarded them before counting.
So the honest headline is: we found zero publisher errors in describing the current guideline, and one upstream defect that explains the mis-statement we originally set out to measure — including the six instances of it that were on this site.
Method
Reference standard. Reynolds RV, Yeung H, Cheng CE, et al. Guidelines of care for the management of acne vulgaris. J Am Acad Dermatol. 2024;90(5):1006.e1–1006.e30 (PMID 38300170). The strength and certainty of all 23 recommendations are in our open dataset; we cross-checked every code against a full reproduction of the recommendation table on a clinical guideline-summary site and against the AAD’s own press release. We could not read the publisher’s live copy on jaad.org, so this article does not claim we did.
Sample. 17 queries covering general treatment advice, named drugs, drug comparisons, hormonal therapy, severe acne, non-guideline remedies and routines: AAD acne treatment guidelines 2024; hormonal acne treatment best options for adult women; birth control pills for acne which ones work best; azelaic acid for acne benefits how effective; acne skincare routine dermatologist recommended step by step; best acne treatment; how to get rid of acne; benzoyl peroxide for acne; adapalene vs tretinoin for acne; doxycycline for acne how long does it take; spironolactone for acne; isotretinoin accutane for severe acne; minocycline vs doxycycline for acne; severe cystic acne treatment options; tea tree oil for acne does it work; clascoterone Winlevi for acne; salicylic acid vs benzoyl peroxide for acne. For each query we took the top results returned by a US web search tool on 30 September 2026, in rank order, capped at one URL per domain per query and six URLs per query, and excluded primary research (PubMed, PMC, journal article pages and research repositories) because the study is about guidance content rather than the underlying literature. We also excluded this site. That produced 100 unique URLs; 97 were retrievable.
Coding. Each page’s full text was searched for guideline, AAD, strength, certainty and reviewer language, and every hit was read and judged by hand against a written rubric, with a verbatim quote kept for every positive code. Pages that merely say a treatment works, without invoking a guideline or characterising evidence quality, were recorded as making no checkable claim rather than as errors. Quoted experts and staff lists did not count as a medical-review signal; only a byline or review line on the page did. Fifteen pages that blocked automated fetching were coded from a text rendering of the page.
Limitations
- Search results are personalised, geographically variable and change constantly. This is one snapshot from one US-based search tool on one day, not a Google position audit.
- Three pages could not be retrieved. Two of them were clinical summaries of the guideline itself (one paywalled, one a state Medicaid drug review) — the page types most likely to contain a strength claim. Their absence probably makes our counts of guideline mentions and strength claims slightly lower than a fully retrievable sample would give.
- One coder, with every positive code checked against its quote, is not the same as two independent coders with a measured agreement statistic. We did not calculate inter-rater reliability.
- Excluding primary research was a deliberate choice about what the study is measuring. A sample that included it would show higher guideline engagement and would answer a different question.
- “Mentions the guideline” is a low bar, and clearing it is not the same as representing it accurately. We measured the first; the second is only measurable on the small number of pages that attempt it.
How to check us
The two central claims need no special access. Read the abstract on PubMed: count the good practice statements it names, then compare that with the count it gives in its own first sentence. Then find recommendation 4.1 in any full-text copy of the guideline and read its strength column. Findings 1 and 3 follow from those two observations plus arithmetic. For Finding 2, the full coded sample — every URL, code and supporting quote — is in our open dataset (CSV and JSON, CC BY 4.0).
If we have got something wrong, tell us and we will log it on our corrections page, which is where the six isotretinoin fixes are already recorded.
Not medical advice
This is a study about how a clinical document is described, not about what treatment anyone should use. Nothing here says isotretinoin is a weaker option than you were told — a good practice statement is the work group stating something it considered too clearly correct to need grading, not a demotion. Decisions about acne treatment belong with you and a clinician who can examine you. See our editorial policy for how this site handles clinical claims.
