The five essential questions to ask AI about your health

Good questions turn an AI from dangerous to useful. Ask these five every time you consult an AI about your health, and watch how it answers as closely as what it answers.

1. "What are you basing this on?"

Reveals source quality.

✓Good sign: specific, verifiable sources you can open, for example named peer-reviewed content, a clinical guideline, or FDA prescribing information.

✗Red flag: vague sourcing, "studies show," "research indicates," "the literature suggests."

If it cannot give you a source you can verify, do not trust the information.

2. "What can you NOT detect remotely?"

Reveals the sensing gap. This is the most important question, and it has saved lives.

✓Good sign: a specific list of exam findings it is missing, for example "I cannot detect fever, assess neurological function, or palpate for tenderness; vital signs require measurement."

Compare that list to your own symptoms. If you have anything it admits it cannot detect, get evaluated.

3. "What would require emergency evaluation?"

Reveals the red flags and escalation criteria.

✓Good sign: specific, objective triggers, "seek immediate care if you develop [clear symptoms]."

✗Red flag: vague escalation, "see a doctor if it worsens," "seek care if concerned." Worsen and concerned are subjective.

Write the red flags down. If any appear, escalate.

4. "What are you uncertain about?"

Reveals whether it has humility or overconfidence.

✓Good sign: honest uncertainty, "I am uncertain about this because I lack X," "the evidence is mixed," "I cannot distinguish these without examination."

✗Red flag: perfect confidence, no acknowledgment of any limits.

If it expresses zero uncertainty about a complex situation, be very skeptical.

5. "What should I ask my actual doctor?"

Reveals whether it understands the need for human judgment.

✓Good sign: specific questions and tests for your physician to consider.

✗Red flag: implies a doctor is unnecessary, "this is simple enough to manage on your own."

Use the questions it suggests to prepare for your appointment.

Follow-up questions

If it gives a diagnosis: "How certain are you?" / "What else could this be?" / "What would change your assessment?"

If it recommends treatment: "What are the risks?" / "How do I know if it is working?" / "When should I see a doctor instead?"

If it says do not worry: "What red flags should I watch for?" / "What would make this an emergency?" / "What are you missing by not examining me?"

Score the response

Green flags

Specific, verifiable sources
Explicit limits acknowledged
Clear red-flag criteria
Recommends human evaluation when appropriate
Expresses uncertainty on complex cases

Red flags

Vague sourcing
No mention of limits
Overconfident diagnosis
Suggests no doctor needed
Never says "I do not know"

Decision rule

4 to 5 green and 0 to 1 red: may be useful, still verify. 2 to 3 green with 2+ red: be very skeptical. 0 to 1 green with 3+ red: do not trust this answer.

When to stop questioning and escalate

Stop and call 911 (or your local emergency number)

Chest pain with other symptoms; stroke signs (FAST: face, arm, speech, time); trouble breathing; severe bleeding; anything it says needs emergency evaluation.

Stop and see a doctor

Symptoms match red flags it named; you have findings it says it cannot assess remotely; it expresses real uncertainty; your gut says something is wrong; symptoms persist or worsen despite its advice.

Ask them every time. Your life might depend on it.