Fourteen US health advertisers, scored against the 2026 Meta ruleset.

Every finding below was taken from live creative in the public Meta Ad Library — no account access, no API keys, nothing the advertiser had to hand over. Each ad is cited by its library ID so any claim can be checked at source.

Scoring runs 100-down across a 10-category rule library (PA · IT · MC · RX · SD · CR · RC · LP · TG · PB): critical −40, high −20, medium −10, low −5. A prohibited-category hit forces 0. Tiers: 90–100 clear · 70–89 watch · 40–69 exposed · 0–39 account-threatening.

14advertisers scored

0–62score range

10rule categories

2rules written from findings

The audits

Score	Advertiser	Headline finding
0	By WinonaAccount-threatening	A live ad claims prescription hormones protect against breast cancer, prevent Alzheimer's and reverse osteoporosis. Prohibited category — forces 0.
10	JoyousAccount-threatening	Tells people with depression that SSRIs made them "tolerate life," and promises 91% relief in a month. Disparages a prescribed drug class.
10	MuselyAccount-threatening	Prescription metformin promoted as a longevity treatment — off-label. One creator ad carries no paid-partnership disclosure.
20	LocklabAccount-threatening	Three different efficacy percentages running for the same product at the same time.
20	Male MDAccount-threatening	Every live ad reviewed promotes a prescription peptide on body-transformation claims — and one runs through a creator.
25	Magnali HealthAccount-threatening	"Cellular reversal" and a "complete aging solution" — claims a supplement cannot make.
25	Male ExcelAccount-threatening	One of the best safety disclaimers in the scan, attached to a hook that tells men they are tired, soft and unmotivated.
30	Happy HeadAccount-threatening	Minoxidil advertised for beard growth — an indication it is not approved for.
30	Midi HealthAccount-threatening	The influencer program is disciplined. The ad putting the account at risk is one they wrote themselves — a testimonial quantifying weight loss.
35	Mochi HealthAccount-threatening	Two law firms are already running plaintiff-recruitment ads naming Mochi — while a creator lists the prescription catalogue on camera.
40	Marek HealthExposed	A creator ad running unreviewed for eight months, and a best-performing line telling patients their doctor won't order the analysis.
45	AlloyExposed	Two doctors advertising for the same brand. One cites a 2024 RCT with methodology; the other attaches a promo code to a prescription recommendation.
55	MaximusExposed	"Double your testosterone" — the most quantified outcome claim found in men's health.
62	Tuyo HealthExposed	A prescription drug named in the ad hook, running across 6 states — precisely what Meta tightens on in 2026.

Two rules written from live findings

The rule library is not theoretical. Two CRITICAL rules were added on 2026-07-31 because live creative demanded them — the finding came first, the rule second.

CR-04 · Credentialed-creator claims. A creator with a medical credential makes a safety or efficacy claim, especially alongside a promo code. Found at Alloy — where a second doctor on the same account cites a 2024 RCT with methodology and is the model to copy. Same channel, same brand, opposite outcome.
RX-04 · Off-label indication promotion. A prescription drug promoted for an indication it is not approved for. Found at Musely.

The operating principle: cite, don't vouch. A creator quoting a trial is an asset. The same creator vouching personally is a liability — and the difference is invisible until someone scores it.

What these audits could not tell you

Each audit above reports findings — "here are the problems we found." That is a consulting opinion, and it has two limits worth stating plainly: there is no denominator, and a single recorded finding may be running as many live ads.

A follow-on measurement pass pulled the complete active ad set for five of these advertisers — 528 ads, every one carrying its Meta Ad Library ID so any number is re-checkable at source. What the denominator changed:

The Joyous copy recorded as one finding is running as 6 live ads.
The Marek MC-04 issue recorded as one finding is running as 15 live ads.
Musely FaceRx — a separate page carrying 67 active prescription-skincare ads — was invisible to the original audit entirely.
73 ads run from 55 distinct creator and partner pages, not the brand's own page. This is the whitelisted-UGC surface, and it is where credential and disclosure risk concentrates.

The difference in one line: the audits find problems. The measurement pass counts exposure. A rate with a denominator is a number a deal team can act on; a list of findings is not.

The classifier producing those rates is still in validation and its error rates are not yet published. No violation rate is quoted here until the judge is measured against human-labelled ads in both directions. The corpus, the library IDs and the counts above are complete and verifiable today; the rate is not, so it is not claimed.