How ChatGPT decides which practices to recommend

What happens when a patient asks ChatGPT for a recommendation?

ChatGPT does not answer a "who should I see" question from memory. It runs a live web search, reads the top results, and synthesizes a short answer that names a handful of practices it can stand behind. The patient sees three or four named options, not a page of links.

That synthesis step is the whole game. ChatGPT names practices it can verify and quietly drops the ones it cannot. A practice that is hard to find, or hard to corroborate, never makes it into the short answer the patient actually reads.

What signals get a practice named?

ChatGPT favors content that answers the patient's exact question. A page that directly addresses what a treatment costs in a specific city, or how to choose between two procedures, is far more quotable than a generic homepage that lists services.

ChatGPT looks for corroboration beyond the practice's own site. When directories, review platforms, news mentions, and professional listings agree about a practice, the engine treats the practice as real and safe to recommend. Self-published claims alone rarely clear that bar.

ChatGPT rewards consistency. The practice name, address, phone number, and specialty should match everywhere they appear. When the details line up across sources, ChatGPT can verify the practice quickly. When they conflict, the engine hesitates and moves on.

What gets a practice ignored?

Thin sites get skipped. A practice site that is mostly photos and a contact form gives ChatGPT nothing to extract, so the engine has no sentence to quote and no question to answer with that practice.

No third-party footprint is the most common reason a good practice stays invisible. If nothing outside the practice's own website mentions it, ChatGPT cannot corroborate the practice and will not risk naming it. Unverifiable claims — superlatives with no source — are discounted for the same reason.

Do other AI engines decide the same way?

The mechanism is similar across engines, but the sources differ. ChatGPT searches the web through Bing's index, so what Bing can find shapes what ChatGPT can recommend. Perplexity cites a source for nearly every claim, which makes its reasoning easy to inspect. Gemini leans on Google's index and Google Business Profiles.

Because each engine draws on a different index, a practice can be recommended by one engine and absent from another. That is why a real visibility check runs the same patient questions across several engines instead of trusting a single assistant.

What should a practice do about it?

Measure first. Ask each engine the questions your patients ask, and record whether your practice is named and which sources are cited. The citations show you exactly which content ChatGPT trusts and which competitors already own the answers. ChatGPT drops practices it cannot verify before the patient ever sees the answer. That is why Tenva ran the same kind of check on itself and appeared in 0 of 95 AI answers across two June 2026 multi-engine probes before it built any pages.

Then build what is missing: pages that answer one patient question each, evidence ChatGPT can quote, consistent details everywhere you appear, and third-party sources that corroborate you. This work is AEO, answer engine optimization — earning the recommendation, not ranking in a list.