Compounded GLP-1 Ad Claims: What FDA Flagged

FDA flagged five recurring claim types in compounded GLP-1 advertising: equivalence to an approved drug, unsubstantiated outcome figures, implied FDA endorsement, unbalanced risk presentation, and atypical results presented as ordinary. Those categories appear across the 30 warning letters sent to telehealth companies on 3 March 2026 and the 25 sent during the week of 15 June 2026. Curve is the HIPAA-compliant tracking and analytics layer that lets you measure which of these claims your funnel actually depends on before you rewrite them, with a signed Business Associate Agreement on every plan.

Why a taxonomy beats a checklist

Most compliance reviews of GLP-1 funnels fail the same way. Someone opens the landing page, deletes a brand name, softens one adjective, and calls the page clean. Two weeks later the same claim reappears in an email subject line, a quiz result screen, or a creator's script, because nobody understood what category of statement was the problem.

A claim type is portable. Once your team can recognise an equivalence claim, they catch it in ad copy, in an FAQ answer, in a comparison graphic, and in a sales call script. A list of banned words catches none of those, because the next writer simply picks different words.

What follows is the taxonomy, the language patterns that trigger each category, and a rewrite that preserves the marketing job the original sentence was doing. The last part matters. A rewrite that guts conversion gets reverted by whoever owns the number.

Claim type 1: equivalence to an approved drug

This is the largest category and the one the letters return to most often. Compounded semaglutide and tirzepatide are not FDA-approved drugs. They have not been through the approval process that establishes safety and efficacy for an indication. Any statement that positions the compounded product as the same thing as the approved brand asserts something the agency has not granted.

Language patterns

• Direct substitution: "generic Ozempic", "the affordable version of Wegovy", "Mounjaro without the price tag".
• Ingredient equivalence: "same active ingredient as", "identical molecule to", "pharmaceutically equivalent".
• Borrowed evidence: citing the approved drug's clinical trial results as what your compounded product delivers.
• Structural equivalence: a two-column comparison graphic where the only differing row is price.
• Search-driven equivalence: bidding on a brand term and running an ad whose headline implies the searcher will receive that brand.

Rewrites

The marketing job here is usually recognition. The prospect knows a brand name and does not know yours. You can do that job by naming the drug class rather than the branded product, and by describing your own service honestly.

• Instead of "generic Ozempic", write "compounded semaglutide prescribed through licensed clinicians".
• Instead of borrowed trial figures, describe the program: clinician consultation, dose titration, ongoing support.
• Instead of a comparison table implying sameness, describe access. Cost, timeline, and delivery are real differentiators that make no equivalence claim.
• If you use a brand name at all, use it to say what the product is not: state plainly that compounded formulations are not FDA-approved and are not the branded product.

That last one feels counterintuitive to growth teams. In practice, a plain disclosure high on the page removes the strongest single objection a regulator raises, and prospects who read it are better qualified.

Claim type 2: unsubstantiated outcome claims

Any specific figure about what the product does needs substantiation for the product you actually dispense. Not for the approved drug, and not for the drug class in general.

Language patterns

• Numeric weight-loss figures in headlines, ad creative, or quiz result screens.
• Timeline promises: results by a stated week or month.
• Comparative superiority: faster, more effective, better tolerated, without a head-to-head basis.
• Mechanism overreach: describing appetite or metabolic effects in stronger terms than evidence supports.
• Implied certainty: "you will", "patients lose", "expect to".

Rewrites

The marketing job is credibility. Substitute process credibility for outcome credibility. Describe who prescribes, how dosing is supervised, what happens if the patient has side effects, and what support exists between refills. Those statements are verifiable, they differentiate, and they carry no substantiation burden about the drug.

Where you keep an outcome statement, make it conditional and sourced, and be explicit that individual results vary and depend on clinical factors. Vague hedging in a footnote does not repair a specific promise in a headline. The qualification has to travel with the claim.

Claim type 3: implied FDA endorsement

This category catches teams who believe they are being reassuring rather than promotional.

Language patterns

• "FDA-approved ingredients" where the approval attaches to a different finished product.
• "FDA-registered facility" or "FDA-registered pharmacy" placed as a quality signal in a benefits section.
• Trust badges combining an FDA reference with a checkmark or shield graphic.
• "FDA-regulated" used interchangeably with "FDA-approved".
• Statements that a state-licensed pharmacy's status confers federal approval on the preparation.

Rewrites

Registration is not approval, and the agency treats blurring the two as misleading. The fix is precision. Say the pharmacy is licensed in the states where it dispenses, say whether it is a 503A or 503B facility if that is accurate, and say who reviews the prescription. Move any factual regulatory statement out of the persuasion block and into a plain credentials section where it reads as disclosure rather than endorsement.

Claim type 4: unbalanced risk presentation

Promotional communications about prescription drugs are expected to present benefit and risk with reasonable balance. The letters treat page rendering as the thing being judged, not page source.

Language patterns

• Benefit in display type, risk in a small footer.
• Risk information behind a collapsed accordion, a modal, or a "learn more" link.
• Risk on a separate page the funnel does not require the visitor to reach.
• Ad creative naming a benefit with no risk reference and no accessible path to one.
• Video and short-form creative where risk text appears for a fraction of a second.

Rewrites

Review the page as a phone renders it, scrolling at normal speed, and ask what a visitor has actually seen by the time the primary call to action appears. Put material risk information above that point. In short-form video, state risk in audio rather than relying on burned-in text nobody pauses to read. In paid social, the ad copy itself needs a risk reference, because you cannot assume the click.

Claim type 5: testimonials and atypical results

A testimonial is a claim. It carries the same substantiation and typicality expectations as anything your copywriter writes.

Language patterns

• Selected best-case stories presented as representative.
• Before-and-after imagery, which functions as an outcome claim in pictures.
• Creator content where the loosest claims in your entire funnel usually live.
• Review widgets that surface only five-star outcome-focused reviews on the conversion page.
• Case studies with numbers and no context about who the patient was or what else changed.

Rewrites

Prefer testimonials about experience over testimonials about outcome. A patient describing how quickly they reached a clinician, how dose adjustment was handled, or how support responded is persuasive and makes no claim requiring substantiation. Where an outcome testimonial stays, present typicality honestly and adjacent to the story, not in a global footer.

Affiliates and creators need supplied language and a real audit. Promotion made on your behalf is attributed to you. Our related coverage of compounding pharmacy GLP-1 advertising restrictions goes further into where that responsibility sits.

What the ad platforms add on top

Platform policy is a separate regime that bites earlier. Meta requires prior authorization for prescription drug advertising and admits only pharmaceutical manufacturers, online pharmacies, and telehealth providers. It rejects branded pharmaceutical weight-loss terms and most before-and-after weight-loss imagery.

The practical consequence is that Meta will reject an equivalence claim or a before-and-after image long before the FDA reads your page. Treat rejections as a claim-quality signal rather than a routing problem to be solved by rewording until something passes. If your team's habit is to iterate creative until approval lands, you are optimizing toward the most aggressive claim the classifier happens to miss. The current GLP-1 advertising policy picture across Google and Meta is worth reading alongside this.

The measurement problem this creates

Rewriting claims requires knowing which ones your funnel depends on. If a headline changes and conversion falls eleven percent, you need to know that within days, and you need to know whether the drop happened at the ad, the landing page, the quiz, or the intake handoff.

That is where GLP-1 advertisers create a second exposure entirely separate from the FDA question. A standard Meta Pixel or Google tag on these funnels sends the page URL, and on a GLP-1 funnel the URL frequently names the medication. It sends quiz completions with answers attached when forms are instrumented naively. It sends a durable browser identifier that ties the sequence to one person.

Meta and Google do not sign BAAs for their advertising products. Each of those calls discloses an identifiable person's interest in a specific treatment to a vendor with no BAA. Healthcare pixel litigation on exactly this mechanism has produced settlements cumulatively past $100 million. A company can fix every claim in this article and still be carrying that risk.

How Curve lets you test claims without disclosing patients

Curve is HIPAA-compliant ad tracking, attribution, and analytics for healthcare. It replaces client-side pixels with a server-side path, which creates a decision point before anything leaves.

• Server-side collection. The Curve script installs in place of the Meta Pixel and Google tag. Events go to Curve's US-hosted infrastructure rather than straight to ad platforms.
• Per-destination field mapping. Only explicitly mapped fields forward. Medication-naming URLs, quiz answers, and free-text fields stay behind by default, so a claim test does not leak what the claim was about.
• Neutral event aliases. The ad platform sees a neutral conversion name, not the service line, so campaign structure does not become a diagnosis list in an ad account.
• Identifier hashing. Email, phone, and name are SHA-256 hashed per each platform's conversion API requirements.
• Bridge tokens. Attribution survives the jump from a marketing site into a separate intake or booking tool, which is exactly where GLP-1 funnels lose the chain and where variant tests go dark.
• PHI-pattern detection. Payloads carrying PHI-shaped values such as SSNs, MRN-style identifiers, or long numeric sequences get flagged, which is how you find out a quiz field changed shape three weeks ago.

Clean conversions forward server-side to Meta CAPI, Google Ads Enhanced Conversions, TikTok Events API, Microsoft, and LinkedIn. Offline conversion uploads with click-ID matching let you feed genuine downstream outcomes back, so a compliant claim variant is judged on patients who start treatment rather than on quiz completions. A signed BAA comes with every plan. For the underlying architecture, see why client-side pixels break HIPAA.

Frequently asked questions

Is naming the drug class instead of the brand always safe?

Safer, not automatically safe. Naming the class avoids the strongest form of equivalence claim, but an unsubstantiated outcome figure or missing risk balance elsewhere on the page is a separate violation. Review the whole communication.

Can we say our compounded product contains semaglutide?

Stating the active ingredient factually is different from asserting equivalence with an approved finished drug. The trouble starts when the ingredient statement is used to import the approved product's evidence, indication, or safety profile.

Does a disclaimer at the bottom of the page fix a headline claim?

Generally no. The letters treat the communication as it is rendered and encountered. A qualification separated from the claim by two screens of scrolling does not balance it.

Are quiz questions promotional communications?

They can be. Quiz framing frequently contains claims, especially in result screens that tell a respondent what they qualify for and what to expect. Review the quiz with the same taxonomy you apply to the landing page.

What about content a creator posted without our approval?

If the creator is compensated or affiliated, promotion made on your behalf is generally attributed to you. Supply approved language, require it, and audit what actually shipped rather than what was briefed.

How do we A/B test claim variants without sending PHI to ad platforms?

Keep clinical and quiz content inside systems whose vendors have signed BAAs, and send ad platforms only a neutral, hashed, matched conversion signal. Curve's server-side path with per-destination field mapping and neutral aliases exists to make that separation the default rather than an ongoing discipline.

Where to start

Run the taxonomy across every surface, not just the landing page: ad copy, quiz framing and results, email sequences, SMS, creator scripts, and sales call guides. Categorise each finding rather than editing it in place, because the category is what stops the claim coming back.

Then fix the measurement layer so the rewrite is testable. Curve gives you server-side collection, per-destination field mapping, hashed identifiers, neutral event aliases, and bridge-token attribution, so you can see which claim variants actually earn patients without telling Meta or Google who is interested in a GLP-1. Run the free compliance scanner against your funnel, or visit curvecompliance.com to walk through it with us.

Reviewed August 2026. This is general information, not legal or regulatory advice. FDA positions and platform policies change; consult qualified counsel about your specific promotional materials.