Free and printableA clinician's checklist
A clinician's checklist
for using AI safely
I use AI tools to save time on paperwork, never to replace clinical judgment. This is the checklist I follow and teach. Print it, share it with your team, and adapt it to your organization's policy.
1
Before you type
- Use only tools your organization has approved for patient data.
- Remove names, dates of birth, record numbers and addresses unless the tool is approved for protected health information.
- Know whether the tool keeps or trains on what you enter.
2
While you use it
- Ask for reasoning and sources, then open the sources yourself.
- Treat drug doses, interactions and calculations as unverified until you check them in a trusted reference.
- Watch for confident answers to questions the tool cannot know, such as findings it never saw.
3
Before it reaches the patient
- Read every word of an AI draft before you sign it. Your name means you own it.
- Match the reading level and language to the patient in front of you.
- Never let AI make the final call on triage, diagnosis or disposition.
4
After
- Report errors through your organization's safety system, the same way you would for a device.
- Tell patients when AI helped, if your organization's policy calls for it.
- Keep learning: tools change every few months, and so do their failure modes.
This checklist is education, not a policy. Your organization's rules on approved tools and patient data come first.
Written by Harvey Castro, MD, MBA, board-certified emergency physician. Last reviewed November 12, 2026.