Instant Medical Advice
Sub-second AI triage. Describe what’s happening — get an immediate ER vs urgent care vs home decision, with the reasoning attached. Free, anonymous, no signup.
Red-flag symptoms — go to the ER now, do not consult here first
If any of the following describe what you’re experiencing, call your local emergency number immediately. Don’t wait. Don’t check this site first.
- Cardiac: Chest pain or pressure (especially radiating to arm, jaw, or back), shortness of breath, sudden cold sweats, fainting
- Stroke: Sudden facial drooping, arm weakness, speech difficulty, sudden severe headache ("worst of your life"), sudden vision loss
- Severe respiratory: Inability to speak full sentences, blue lips or fingertips, severe wheezing, choking, anaphylaxis (swelling, hives, throat tightening)
- Severe trauma: Heavy bleeding, suspected fractures with deformity, head injury with confusion or vomiting, deep wounds
- Mental health crisis: Active suicidal ideation, plan, or attempt; severe disorientation; psychotic symptoms requiring acute intervention
- Pregnancy emergencies: Heavy bleeding, severe abdominal pain, severe headache with vision changes, signs of preterm labor
For a broader look at AI medical tools, see our guide to AI doctors in 2026 — what they are, top services compared, and when to use each.
Why speed matters in triage
Triage is the one part of medicine where response latency is itself clinically relevant. If you’re trying to decide whether the chest discomfort you’re feeling warrants an ER visit, waiting 10 seconds for a chatbot to start responding is already too long. Hesitation in triage is harmful — it’s why every emergency system in the world prioritizes immediate, structured first-contact assessment.
Dr.Khan’s instant advice flow runs on Llama 3.3 70B served through Groq’s inference platform, which delivers around 500 tokens per second. In practice that means the response starts under one second after you submit, and a complete triage assessment arrives in under three seconds. By comparison, the same query on OpenAI’s infrastructure typically takes 8–15 seconds to complete.
The four-bucket triage model
Every triage response sorts your symptoms into one of four acuity buckets, with the reasoning visible:
- Emergency (ER now / 911): Symptoms that suggest a time-critical condition. Stroke, MI, severe respiratory distress, anaphylaxis, severe trauma, active mental health crisis.
- Urgent (urgent care or ER within hours): Concerning but not immediately life-threatening. Severe pain, persistent high fever, suspected infection requiring same-day antibiotics, possible fractures without deformity.
- Routine (PCP within days): Symptoms that warrant a clinician’s assessment but don’t require same-day care. Persistent symptoms over a week, suspected chronic conditions, medication review.
- Self-care with monitoring: Symptoms consistent with a self-limiting condition. The AI gives you specific guidance on what to monitor and what would change the decision.
How the triage flow differs from chat or consultation
Chat is open-ended conversation — multi-turn, exploratory, depth over speed. Consultation is the full clinical structure — patient profile, history, differential, plan. Triage is the decision-making slice in the middle: given what’s happening right now, what do I do in the next hour? Same underlying engine, different prompt and optimization target.
Frequently asked questions
- How fast is "instant"?
- Sub-second. The chat runs on Llama 3.3 70B served via Groq, which delivers around 500 tokens per second — about 20× faster than the same query on OpenAI or Anthropic. You see the start of a response before you've finished re-reading what you typed.
- Should I use this in a real emergency?
- No. If you are experiencing any of the red-flag symptoms listed above, call your local emergency number immediately — don't wait to consult an AI. The triage tool is for the ambiguous middle — symptoms that are concerning but unclear in severity, where you genuinely don't know whether to wait, see an urgent care, or go to the ER.
- What does the triage decision look like?
- For each set of symptoms, the AI categorizes urgency into four buckets: emergency (call 911 or go to ER now), urgent (urgent care or ER within hours), routine (see a primary care clinician within days), and self-care (manage at home with monitoring guidance). It explains why it placed you in that bucket and what would change the decision.
- What information will it ask for?
- Age, gender, primary symptom, severity (1–10), how long it's been going on, any associated symptoms, and whether anything is making it dramatically worse. Five quick fields, then an immediate response.
- What if I'm wrong about how serious my symptoms are?
- The AI is calibrated to err toward escalation, not minimization. If your description matches any red-flag pattern, you'll be told to seek emergency care immediately, regardless of how mild you think it feels. False positives are acceptable here; false negatives aren't.
- Is the response actually accurate?
- For triage decisions on common presentations, yes — Llama 3.3 70B handles standard symptom-to-acuity mapping reliably. For unusual presentations or ambiguous cases, the AI defaults to escalation. Treat the recommendation as an informed second opinion, not a substitute for a clinician's physical assessment.
Need a triage decision now?
Sub-second response. Free. Anonymous. Built for the ambiguous middle.
Get triage decision