AI in the Exam Room Clinician
CLINICIAN · POCKET CARD

Pocket card

The wallet version: scripts, frameworks, and the harm protocol.

All resources

The wallet version. Front: what to say and the frameworks. Back: the liability and humility logic, teaching scripts, and what to do when AI has caused harm.

Front

The opening question

"Before we start, did you look this up online or ask any AI about it?"

Follow-up: "What about that worried you?" / "What didn't AI address?" / "What do you think is going on?"

Response scripts

✓ Right: "ChatGPT got this right. Here is what my exam confirms..."   ~ Partial: "AI was on track. Let me add what I am finding..."   ✗ Wrong: "AI missed something. Here is what I am detecting..."   ⚠ Anxiety: "Each question made AI suggest worse things. Let me tell you what I see..."

Frameworks

Ten billion sensors: you have ~1012 sensory neurons; AI has zero. You detect color, breathing, sweat, tenderness, odor, affect. Velociraptor test: "AI gave probabilities. Your body gave signals. Trust your signals." Use for a patient dismissing instinct, a worried parent, or a vague "something's wrong."

Smart phrases & tracking

.AIOK confirmed · .AICORRECT corrected · .AIDELAY caused delay · .AINONE no AI use. Track monthly: percent using AI, accuracy rate, AI-influenced delays.

Back

The liability reality

AI companies: disclaimers, no license, no malpractice, no liability. You: license at risk, premiums, personal liability, board oversight.

Intelligent humility

General AI (ChatGPT) trains on journals plus forums plus blogs plus SEO garbage. Content-controlled AI (TheDude) trains on validated, curated medical content only. Hallucination is fabrication with full confidence; watch for invented studies, fake citations, confident wrong answers.

Patient teaching scripts

Validate effort: "You used AI exactly right, research then verification."
Explain the gap: "AI for education, physician for diagnosis."
For parents: "Your instinct matters. No AI has watched your child grow."
Positioning: "I'm not anti-technology. I'm pro-appropriate-technology."

When AI has caused harm

  1. Do not lecture. The patient is already sick and scared.
  2. Treat first. Medical care before teaching.
  3. Document factually. No judgment in the chart.
  4. Teach when stable. Explain what AI missed and why.
  5. Consider systems. Why did the patient rely on AI instead of calling?