Can you advertise GLP-1, semaglutide or weight loss on Facebook? What actually gets flagged

Short answer: yes, within Meta's rules. Long answer: the trouble usually isn't the campaign. It's the data your site sends back. And GLP-1 programmes, more than most businesses, have a website that's full of exactly the words that trigger it.

Let me explain how this goes wrong, because it's a sequence, and each step is easy to miss.

Think about how these businesses sell. Someone sees an ad, fills in a short intake form, and has a consultation with a clinician. If it's suitable, a prescription or a programme starts. So the sale, the thing that matters, happens after the conversation. The ad platform only ever sees the form. And the site behind that form is full of words like semaglutide, tirzepatide, dosage and weight loss programme.

Now look at where those words end up. The page link: something like /semaglutide-consultation-thank-you goes out with the event. The event name: SemaglutideLead goes out too. The form: "What is your current weight?" and "Any medical conditions?" are answers that can travel with the submission.

That's three leaks, each one a reason for Meta to hold back or drop the event. When the event is dropped, the campaign learns less, cost per lead climbs, and you're left wondering why the same ad is suddenly worse. Meanwhile the real issue is invisible in Ads Manager, because it lives in the data.

Two separate problems, then, and both need fixing.

Problem one is what leaves. Strip all three leaks before anything is sent. Cut page links to your domain. Turn event names into neutral codes. Send only a hashed email and phone, plus the ad click. The conversion still counts. There's just nothing in it for Meta to flag.

Problem two is what the ads learn. A form fill is a weak signal for a programme like this, because plenty of people fill in an intake and never book, never qualify, never start. The valuable events are the booked consultation, the completed consultation and the started programme. Those tell the campaign what a real patient looks like. And because they're sent as neutral events, they don't recreate problem one.

A made-up example of why it matters. 300 intake forms. 90 book a consultation. 60 complete it. 35 start a programme. If you optimise on intake forms, the campaign sees 300 examples and 265 of them never started. If you optimise on completed consultations, it sees 60 examples, every one a real prospect who followed through. Cost per intake will look worse the moment you switch, and cost per started programme will improve. Judge it by the second number.

A word about the intake form itself, because it's the riskiest part of a telehealth or weight loss funnel. The questionnaire, with its weight, medications and conditions, is clinical information. It belongs in a secure system that your clinicians use, not on a page whose events are sent to an ad platform. A common good pattern is a short lead form on the marketing page (name, email, phone, consent) and a separate secure intake after the person has engaged. The ad platform only ever sees the first step, and it never needs to see the second. That single design decision removes the biggest source of leaks before any tracking setup has to be clever.

If you advertise in this space: what's the first thing Meta flagged for you?