Why AI sends GLP-1 patients to telehealth

Run the query yourself first

Open a fresh, signed-out session and ask any of the four engines where to start GLP-1 treatment.

You will get national telehealth brands — the ones that have raised money, bought coverage and been written about continuously for two years. You will not get a clinic in your city, however good it is.

Then ask the same question with your city on the end. The answer changes completely.

That difference is the whole strategy, and most clinics never notice it because they only ever ran the first query, concluded AI visibility was hopeless, and stopped.

Why the category term is not winnable

An assistant answering an unqualified question has no location to work with. With nothing to localise, it falls back on the most-cited entities in the category — and citation volume is where a well-funded national brand is strongest.

This is the same mechanism described in how AI engines choose which business to name: engines resolve from sources, and the brand with the most sources wins the unqualified question.

You cannot out-cite a company with a press budget. Nor do you need to. Nobody books a local appointment off an unqualified category question — they book off “near me”, and that query resolves somewhere else entirely.

What the winnable query looks like

The queries a local clinic can realistically win share three traits:

• They carry a location. “near me”, a city, a neighbourhood.
• They carry a qualifier. Cost, eligibility, insurance, first appointment.
• They imply an appointment. Somewhere to physically go.

where to get semaglutide near me [city] is a completely different question from best GLP-1 provider, and it resolves through local directories — Google Business Profile, Bing Places, Healthgrades — rather than editorial coverage.

That is a surface you can actually influence. Most of it is free, and most clinics in this category have not touched it.

The compounded question decides whether you match

Patients search by drug name, and increasingly by whether it is compounded or branded.

A clinic whose site never states which it prescribes cannot be matched to that question. Not penalised — unmatched. The engine has nothing to compare against the qualifier in the query, so it selects a clinic that was explicit.

The fix is not marketing copy. It is stating plainly which medications you prescribe, in what form, and what a month actually costs. Clinics avoid publishing this because it invites price comparison. The trade is real, but so is the cost of being unmatchable.

Reddit matters more here than anywhere else

GLP-1 decisions get researched in community threads more heavily than any other category we measure. People compare experiences, side effects, providers and prices in public, at length, for months.

Perplexity in particular leans on that discussion — the mechanism is covered in how Perplexity picks local businesses.

The honest implication is uncomfortable: this presence is earned slowly and cannot be bought. Astroturfing a subreddit is detectable, gets removed, and in a health category invites problems worse than the marketing loss. The clinics that show up in these threads are the ones patients actually brought up unprompted.

That is a real constraint. It is also why the position holds once you have it.

Eligibility content outranks marketing content

The searches with genuine volume behind them are unglamorous:

• BMI thresholds and who qualifies
• What the first appointment involves
• What a month costs, all in
• What happens when you stop
• Contraindications

These are the pages almost no clinic writes, and they are exactly the pages an assistant can extract a direct answer from. A page that answers a specific question in plain language gets cited. A page describing your “personalised approach to wellness” does not, because there is nothing in it to quote.

Write them as clinical explanation rather than promotion. That keeps them inside FTC substantiation rules for weight-loss claims and state telehealth requirements, and it happens to be the register engines extract from most reliably.

Where to actually start

1. Run both query forms across all four engines, three times each. Record who gets named and which source was cited. The prompt set builder will generate the set.
2. Claim Bing Places. Two of the four engines resolve local questions through Bing, and it is the most common gap we find — see why Bing Places decides your ChatGPT visibility.
3. Publish eligibility and cost pages before writing anything promotional.
4. State which medications you prescribe, explicitly.
5. Check your review position against the local threshold, which is covered in how many reviews before AI recommends you.

The full breakdown for this category sits on AI visibility for medical weight loss clinics.

The usual caveat

No AI platform publishes how it selects sources or weights them. Everything here comes from measured output across repeated runs, not from confirmed ranking factors.

This category also moves faster than most — the regulatory position on compounded medications has shifted repeatedly, and the answers engines give have shifted with it. Re-run your own measurements quarterly rather than trusting anything written today, this page included.