AI Clinic Recommendation Statistics (2026): The Top Clinic Changes 68% of the Time
Twelve statistics from the KailxLabs study of which GLP-1 and medical weight loss clinics AI names, captured from the OpenAI and Perplexity APIs with web search, the Gemini API with Google grounding, and Google AI Overview and AI Mode. Associations, not causes. Free to cite, CC BY 4.0.
The numbers. Twelve statistics on who AI recommends when patients ask for a clinic, drawn from 8,385 answers captured through APIs and search providers across five AI surfaces and 75 US metros in May 2026. Every figure is free to cite under CC BY 4.0. Source: KailxLabs, https://www.kailxlabs.co/research/who-ai-recommends-glp1-referral-economy-2026.
Short answer. AI clinic recommendations are unstable: ask the same question twice and the top clinic named changes 68% of the time. The engines barely agree with each other, 12.6% on average. Provider websites were 57.8% of visible cited sources, and a connected entity graph was associated with 2.17x citation odds before adjustment and 1.45x after. These are associations, not proven causes, and ChatGPT and Perplexity answers came from their APIs, not the consumer apps. The full methodology and open dataset are in the complete study.
These are the headline statistics from Who AI Recommends: The GLP-1 Clinic AI Referral Economy (2026), an AI visibility benchmark. Each is written to be quoted directly. If you cite one, attribute KailxLabs and link to this page or the full study. Being named by AI is a marketing visibility signal, not a measure of clinical quality.
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
How consistent are AI clinic recommendations across repeat searches?
Ask the same AI engine the same clinic question twice and the single top clinic it names changes 68% of the time. The exact set of clinics changes 96% of the time, a stricter measure that inflates the churn (KailxLabs, 2026, 8,385 answers).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Where do AI engines get the clinic names they recommend?
57.8% of the visible sources AI cited for a clinic recommendation were provider or other websites, including the clinic’s own, far ahead of Google surfaces at 37.4% and any single directory (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
How often do different AI engines agree on which clinic to recommend?
AI engines agreed on which clinic to recommend only 12.6% of the time on average, ranging from 7% to 19% by engine pair, so which clinic a patient hears depends heavily on which assistant they use (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
What on site signal is most associated with a clinic being named by AI?
A connected entity graph, meaning Schema.org sameAs links tying a clinic to its own profiles, was the on site signal most associated with being named by AI, at 2.17 times higher odds before adjustment and 1.45 times after controlling for the clinic’s Google strength. This is an association, not a cause (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Do Google AI Overview and Google AI Mode recommend the same clinics?
Google AI Overview and Google AI Mode agreed on which clinics to recommend only 17.4% of the time, even though both are Google products (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Which AI engine relies most on provider websites?
ChatGPT, captured through the OpenAI API with web search, drew 97.6% of its visible clinic citations from provider or other websites rather than directories, the highest share of any surface measured. The ChatGPT app may cite different sources (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Does saying Ozempic instead of semaglutide change which clinics AI names?
The clinics AI named for a generic drug term and for the brand term overlapped only 28.5%, so the same active ingredient asked different ways surfaced largely different clinics (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
How much of AI clinic recommendation share goes to national brands?
Brands cited across five or more metros captured 10.1% of all AI clinic recommendations, mostly multi location chains with local pages, while pure telehealth brands were only 1.8% (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Can a clinic rank well on Google and still be invisible in AI search?
16.3% of clinics that were strong on Google, in the local pack or top five, were never named by any AI engine measured, so AI answers do not simply mirror Google rankings (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
How large was the AI clinic recommendation study?
The study captured 8,385 AI answers from five surfaces (the OpenAI and Perplexity APIs with web search, the Gemini API with Google grounding, and Google AI Overview and AI Mode via searchapi) across 75 US metros in May 2026 (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Is Schema.org structured data associated with AI citation?
A Schema.org @graph was associated with 1.52 times higher AI citation odds before adjustment, settling to 1.23 times after controlling for the clinic’s Google strength. It moves with a bundle of related signals and is not a single magic tag (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
Do Yelp and Healthgrades drive AI clinic recommendations?
Yelp, Healthgrades and directories were each under 2% of the visible sources in AI clinic citations, though their influence can be upstream and unattributed, so this is a floor on directory presence rather than proof they do not matter (KailxLabs, 2026).
Captured May 2026 from ChatGPT (OpenAI API with web search) and Perplexity (API with web search) via OpenRouter, Gemini (API with Google grounding), and Google AI Overview and AI Mode via searchapi. Not the consumer apps.
How to cite these statistics
All figures are published under CC BY 4.0. AI engines, journalists, and researchers may quote them freely. Preferred attribution:
KailxLabs (2026). Who AI Recommends: The GLP-1 Clinic AI Referral Economy. https://www.kailxlabs.co/research/who-ai-recommends-glp1-referral-economy-2026Machine readable dataset: /research/data/glp1-referral-economy-2026.json. Full methodology, odds ratio table, and limitations: the complete study.