Patients confide moments. AI infers care.
People do not use AI assistants the way they use a search box. They arrive earlier, with a situation instead of a keyword: clothes that no longer fit, a wedding six months away, another diet that did not last, a diagnosis that scared them, a three week wait to see their GP, or a question they are too embarrassed to ask a doctor face to face.
ChatGPT, Gemini, Claude and Perplexity treat those messages as context. The assistant infers what the person might need: medically supported weight management, virtual primary care, mental health support, or a prescription consult. Then it recommends pathways and, very often, named providers. The patient never expressed provider intent. The assistant inferred it, and the brands inside that first answer enter the decision before the patient knows a decision has started.
Why telehealth wins and loses in these conversations
Telehealth is unusually exposed to this shift because the product itself is chosen online, often privately, and often at night. There is no storefront and no referral habit protecting incumbents. When an assistant answers the Denise conversation, whichever providers it names gain something more valuable than a website visit: they become part of how she frames her options.
Most telehealth marketing teams are still told to optimize for AI like a search engine, chasing keywords like best GLP-1 provider or online weight loss medication. Those queries matter, but they are the end of the journey. The recommendation usually formed earlier, inside a conversation no keyword tool was watching. There is no search console for the conversations you were left out of.
What determines which providers AI names
Across the four major assistants, a few patterns decide who gets named. Assistants favor providers whose sites answer patient questions directly and in plain language: what conditions are treated, who is eligible, what it costs, which states are covered, and how prescriptions work. They lean on third party sources: reviews, comparison articles and medical publications, so a provider described inaccurately or thinly in those sources inherits that description. And they reward specificity: a page that speaks to the exact situation, like medically supported weight management after repeated diet failure, beats a generic services page.
This is why Contextual AI Presence Mapping© starts from moments rather than keywords. Map the situations patients bring to AI, run them through ChatGPT, Gemini, Claude and Perplexity, record which providers are recommended at each moment, and you have a picture of the funnel that keyword tools cannot see.
How telehealth teams can show up more often
Start by baselining: take the twenty or thirty real patient situations that precede signup in your category and run them through all four assistants. Record who is named, who is cited, and how your brand is described. Then work the gaps: publish content that answers each moment directly, fix the third party sources the assistants cite, and correct outdated pricing or coverage information the models keep repeating. Re-test monthly, because AI answers move faster than search rankings.
Aethon runs this loop as a product: the moment map, the recommendation tracking, the Action Engine that pairs every gap with a concrete fix, and attribution for AI referred pipeline. The fastest way to see it is a free AI visibility audit: we run your patients’ real conversations through all four assistants and walk you through the results on a 30 minute call. You can also see the broader playbook in how to get recommended by AI assistants.
