A directory of clinical pearls for AI prompting

Teaching points for prompting AI, written by clinicians, for clinicians.

Every specialty works differently. PromptPearls collects short, specific teaching points — how to ask AI for help with a given clinical task, what risk level it carries, and what still needs a clinician's judgment.

Explore Family Medicine — 66 pearls Browse all specialties
66 pearls live
19 workflow domains
10 specialties mapped

A pearl looks like this

5

Symptom-based differential organization

Part II · Acute Care Encounters

ORANGE · verify closely
I am a family physician reviewing this de-identified presentation: [history, exam findings, vitals]. Organize a differential diagnosis by likelihood. For each, list supporting and opposing features and the single dangerous diagnosis I must not miss. Use only what I've provided.

Copilot, not autopilot — you decide.

Copy prompt

What makes a teaching point a pearl here

Specific to a workflow

Not "how to use AI in medicine" in the abstract — a prompt built for one real task, like organizing a differential or drafting a referral letter.

Tagged by risk

Every pearl carries a color: how much independent verification it needs before you act on the output. Nothing is presented as a final decision.

Built on one rule

AI organizes, drafts, and structures reasoning. It never makes the call. Every prompt in this library is written to keep it that way.

Specialty directory

View all

Directory

All specialties

Family Medicine is fully built out as the showcase set. Other specialties are mapped and in development — the same structure will apply to each: workflow domains, risk-tagged prompts, one source of teaching.

Specialty · Live

Family Medicine Prompt Pearls

66 prompts mapped to 19 clinical workflow domains — pre-visit prep through documentation, patient communication, and practice operations.

Download as PDF
GREEN — light review YELLOW — review before finalizing ORANGE — verify closely RED — no prompt, never delegate

Every prompt organizes, drafts, or structures reasoning — none produces a final clinical decision. Verify all AI output before clinical or administrative use.

Coming soon

AI in Medicine: Prompt Pearls — the online course

A live and self-paced course built on this library's structure: one workflow, one risk-tagged prompt, one verification habit at a time. Family Medicine is the pilot module; each specialty in the directory becomes its own unit as it's built out.

MODULE 1

Foundations — how to prompt AI safely in clinical work

The CTCOV formula, the risk color code, and the copilot-not-autopilot rule that runs through every pearl in this library.

MODULE 2

Family Medicine in practice

Working through all 19 workflow domains — pre-visit prep through documentation, patient communication, and practice operations — with live prompting exercises.

MODULE 3

Building your own pearls

Adapting the format to your own specialty and workflows, with peer review of prompts you write yourself.

Get notified when it opens

Leave your details and we'll reach out as sessions are scheduled.

About

How this library is built

PromptPearls exists because "use AI more" isn't a teaching point — it's a slogan. A pearl has to be specific enough to use in the next ten minutes: which task, what to ask for, and how much to trust what comes back.

The prompt formula

Every pearl follows the same underlying structure, whether or not it's spelled out in the prompt itself:

C

Context

T

Task

C

Constraints

O

Output

V

Verify

The risk color code

Every pearl is tagged by how much independent verification its output needs before you act on it — not by how complex the prompt is.

GreenLight review — drafting, organizing, nonclinical
YellowReview before finalizing — patient-specific documentation
OrangeVerify closely — clinical reasoning, treatment-adjacent
RedDo not delegate — no prompt is provided at all

The rule behind every pearl

Copilot, not autopilot. AI organizes information, drafts language, and structures reasoning. It does not make the final diagnosis, choose the treatment, or send anything unreviewed. Where a workflow is inherently high-stakes — dosing, acute psychiatric crisis, final diagnostic calls — no prompt is offered at all; that judgment stays entirely with the clinician.

Family Medicine set

The Family Medicine library was developed by Izhar Hasan, MD, PhD, as a practical, primary-care-focused starting point for AI prompting in clinical practice. It's the showcase set for this format — each additional specialty will follow the same structure.

This library provides templates, not clinical decisions. All AI output should be independently verified before clinical or administrative use, in line with your organization's approved tools and PHI policies.