Create a set of 8 copy-paste prompts for a practicing [SPECIALTY] physician working in [CARE SETTING]. The prompts must support these recurring tasks: patient-friendly after-visit explanations, differential-diagnosis organization, pre-visit chart review, medication counseling, test-result messages, referral questions, handoffs, and prior-authorization or appeal drafts. My workflow and constraints: - Jurisdiction/organization: [JURISDICTION OR ORGANIZATION] - Patient population and common conditions: [PATIENT POPULATION] - Documentation style, tools, and policies: [WORKFLOW CONSTRAINTS] - Source materials I can provide: [AVAILABLE CLINICAL MATERIALS] For each prompt, provide: 1. A short task name and when to use it. 2. A ready-to-paste prompt with clearly marked input fields. 3. The exact output format requested, such as a problem list, SBAR handoff, plain-language message, or authorization letter. 4. A brief safety note identifying what the clinician must verify before using the output. Build in clinical safeguards. Prompts must tell the model to distinguish documented facts from inferences, preserve uncertainty, identify red flags and contraindications where relevant, never fabricate examination findings, citations, guideline recommendations, dosages, or patient history, and defer final diagnosis and treatment decisions to the licensed clinician. For patient-facing material, request a sixth- to eighth-grade reading level, teach-back questions, and escalation instructions tailored to the supplied facts. For medication-related work, require verification against the current medication list, renal/hepatic status when supplied, allergies, interactions, pregnancy status when relevant, and local formulary. Before answering, self-check that every prompt has a defined use case, inputs, output format, and verification step. Also flag any task that requires protected-health-information safeguards, local policy, or specialist consultation.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [SPECIALTY] | Your medical specialty, such as family medicine, cardiology, or emergency medicine. | Primary care |
| [CARE SETTING] | Your practice setting, such as an outpatient clinic, hospital ward, or urgent care. | Outpatient community clinic |
| [JURISDICTION OR ORGANIZATION] | Your country, state, health system, or relevant institutional requirements. | Massachusetts clinic network; follow internal messaging policy |
| [PATIENT POPULATION] | The patients and conditions you commonly treat. | Adults with diabetes, hypertension, depression, and musculoskeletal complaints |
| [WORKFLOW CONSTRAINTS] | Your EHR, message-length limits, approval steps, and documentation rules. | Epic inbox; patient messages under 250 words; clinician signs all notes |
| [AVAILABLE CLINICAL MATERIALS] | The kinds of de-identified notes, labs, medication lists, policies, and guidelines you can provide. | De-identified visit notes, active medication list, lab results, and payer criteria |
How to use
- Replace each bracketed field with your specialty and workflow details.
- Paste the prompt, then select the prompts that fit your actual work rather than adopting all eight.
- Check every generated clinical statement against the chart, current guidance, and local policy before use.
- Follow up with: "Adapt the test-result and medication prompts for a patient who uses an interpreter and has limited portal access."
Variations
Patient message pack
Use this when you mainly need safe, readable portal messages.
Create 6 reusable patient-portal message prompts for a [SPECIALTY] clinic. Cover normal results, abnormal results needing follow-up, new medication counseling, missed appointments, home-monitoring instructions, and referral updates. Each prompt must request a message under [WORD LIMIT] words at a [READING LEVEL] reading level, include only facts supplied in [CLINICAL DETAILS], and include clear next steps and urgent-care escalation wording when justified. Do not diagnose, prescribe, or invent results. Add a clinician verification checklist after each prompt.
Handoff prompt pack
Use this for transitions between clinicians or care settings.
Write 5 copy-paste prompts that convert [SOURCE NOTE TYPE] into concise clinical handoffs for [HANDOFF CONTEXT]. Use SBAR or I-PASS when appropriate. Require the model to separate confirmed facts, pending items, uncertainty, contingency plans, and ownership of next actions. Include placeholders for [PATIENT FACTS], [MEDICATIONS], and [TESTS OR CONSULTS]. Do not infer a diagnosis or a plan that is absent from the source. End every prompt with a request to flag missing high-risk information for clinician review.
Authorization drafts
Use this when preparing payer-facing clinical rationale.
Create a prompt that drafts a prior-authorization or appeal letter for [TREATMENT OR TEST] based only on [CLINICAL RECORD], [PAYER CRITERIA], and [PRIOR THERAPIES]. Request a one-page letter with diagnosis, relevant severity markers, prior treatment response or intolerance, requested service, and criterion-by-criterion rationale. Require citations only when I provide them. Flag unsupported claims, missing dates, and criteria not yet documented. State that the clinician must verify medical necessity and payer requirements before submission.
Tips
- Keep identifiable patient information out of consumer AI tools unless your organization has approved the specific system and workflow.
- For chart review, provide the source note and ask for a structured list of unresolved questions; do not treat a model summary as a substitute for reading the record.
- Patient messages should name the action, deadline, and escalation path, but should not imply a diagnosis that has not been discussed.
- For prior authorizations, supply the actual payer criteria and dates of prior therapies; generic medical-necessity language is usually weaker.
FAQ
Can I use these prompts to make a diagnosis?
No. They can organize documented information and surface questions, but diagnosis and treatment require clinician judgment and appropriate evaluation.
Should I paste a patient chart into ChatGPT or Gemini?
Use only a tool approved by your organization for the data involved, and follow its privacy, security, and retention rules.
Can the model provide medication doses?
Treat generated dosing information as unverified. Confirm the indication, patient factors, interactions, current labeling, and local guidance yourself.