Type your symptoms into most telehealth platforms today and the first thing that responds usually isn't a person. It's a triage system — sometimes a simple decision tree, increasingly an AI model — deciding how urgent your situation seems, which intake questions to ask next, and which track you get routed down. By the time a clinician sees your file, a lot of the judgment has already happened.
This isn't inherently a problem. Triage has always been a sorting function, and software has done parts of it for years — the "is this an emergency, call 911" branching logic on every symptom checker is a primitive version of the same idea. What's changed in 2026 is how much more the AI layer is now doing, and how much less visible that layer has become to the patient going through it.
What AI triage actually does
In a modern telehealth intake, an AI system typically handles some combination of:
- Symptom extraction from free text. You describe what's going on in your own words; the system parses it into structured clinical data points.
- Urgency scoring. Based on your answers, it assigns a rough priority level — same-day video visit, async review within 24 hours, or a suggestion to seek in-person or emergency care.
- Dynamic follow-up questions. Instead of a fixed form, the questions that appear next depend on what you've already said — closer to how a real intake conversation branches.
- Drafting the clinical summary a provider actually reads. In many systems, the clinician's first exposure to your case is an AI-generated summary of your intake, not your raw answers.
That last point is the one worth sitting with. If the summary a provider reviews is already a compressed, AI-written interpretation of what you said, the quality of your care partly depends on how well that summarization captured what actually matters — and on whether the provider has time to go back to your original answers when something seems off.
Where this genuinely helps
Done well, AI triage can catch things a rushed human intake might miss, because it doesn't get tired and doesn't skip questions on a busy day. It can also route people faster: someone with a straightforward refill request doesn't need the same intake depth as someone describing a new, unfamiliar symptom, and a good triage layer can tell the difference and adjust accordingly.
It's also worth saying plainly: an AI system that flags a red-flag symptom and tells someone to seek in-person care instead of continuing a telehealth flow is doing exactly what it should. Some of the more responsible platforms use AI triage specifically to catch people who shouldn't be treated over telehealth at all — the opposite of the "prescribe everyone" pattern regulators and researchers have been raising concerns about.
Where it gets riskier
The summary can flatten nuance
Clinical judgment often lives in details a structured intake doesn't ask about — tone, hesitation, something a patient mentions in passing that turns out to matter. An AI-generated summary optimized for brevity can lose exactly that kind of signal.
"AI-assisted" can quietly mean "AI-decided"
There's a meaningful difference between an AI system that helps a clinician triage faster and one whose urgency score effectively determines the outcome with only a rubber-stamp human review layered on top. From the patient side, both can look identical — a chat interface, a form, a result. You generally can't tell which one you're dealing with just by using the product.
It can make thin oversight look more thorough than it is
A slick, conversational AI intake feels more attentive than a static PDF form, even if the actual level of clinical review behind it is the same — or lower. That's worth keeping in mind if a platform's intake experience is doing a lot of the trust-building work that used to come from talking to an actual person.
A useful gut-check
If you can't find, anywhere in a platform's process, a point where a licensed clinician reviews something more than an AI-generated summary before a prescription is issued, that's worth asking about directly. A good platform will have a straightforward answer.
What to actually ask before you sign up
- "Does a licensed clinician review my actual answers, or a summary of them?" Both can be legitimate, but you deserve to know which.
- "What happens if the AI triage misjudges my situation?" A platform with a real safety net has an answer beyond "contact support."
- "Can I reach a human if something about the process doesn't fit my situation?" If there's no path around the automated flow, that's a limitation worth knowing about upfront.
The AI layer isn't the problem. Not knowing where it starts and clinical judgment ends is.
Providers that keep a visible human in the loop
The two platforms below both disclose where automation fits into their intake process rather than presenting the whole flow as an undifferentiated "consultation."
Platforms that disclose the human review step
Sesame's intake includes a scheduled consultation with a licensed provider rather than routing purely through automated triage — the AI-assisted intake feeds into that visit rather than replacing it.
See Sesame Care's process → Paid linkCare Bare's "Get Started" flow is structured around a clinician-reviewed intake with a documented path to ask follow-up questions after your initial submission.
See Care Bare Rx's process → Paid linkThe bottom line
AI triage isn't going away, and it shouldn't — used well, it's a genuine improvement over a static form. The thing worth watching for isn't the technology itself, it's whether a platform is transparent about where the software's job ends and a clinician's judgment begins. If a site can't answer that question clearly, that's more informative than anything the chatbot tells you.