How Physicians and Practice Staff Use AI for Patient Communication
AI in a medical practice context means one thing above all others: the physician remains responsible for everything the patient receives. These prompts cover communication and administrative tasks where AI creates real time savings: patient education handouts, visit summaries in plain language, prior authorization letters, and appointment preparation instructions. None of the prompts below are about clinical decision-making. That stays with the clinician.
Every output produced by these prompts requires review by a qualified healthcare professional before it reaches a patient, payer, or referral partner.
10 Practical AI Prompts for Medical Practices
Prompt 1: Patient Education Handout
I am a physician. Help me draft a patient education handout about [CONDITION OR PROCEDURE]. The handout should: - Use plain language at a 6th to 8th grade reading level - Explain what the condition is and why it matters - Describe the most common symptoms in everyday terms - List 4-5 practical steps the patient can take at home - Explain what symptoms should prompt a call to our office - Explain what symptoms are a medical emergency requiring immediate care - End with a reminder that this handout is not a substitute for individualized medical advice Audience: adult patients with no medical training Format: short paragraphs with clear headers, suitable for printing This draft will be reviewed and approved by the physician before distribution to patients.
What to customize: Add any practice-specific instructions such as follow-up appointment timing, your office phone number placeholder, or local urgent care resources. Adjust reading level down further for pediatric parent handouts or populations with lower health literacy.
Prompt 2: After-Visit Summary in Plain Language
I am a physician. Draft a plain-language after-visit summary based on the following clinical notes (which I will paste below). Do not create medical content; only rewrite the notes in patient-friendly language. Clinical notes: [PASTE NOTES, ANONYMIZED IF NEEDED FOR THIS DRAFT] The summary should: - Describe what was discussed at the visit in plain terms - List each medication prescribed with what it is for and how to take it - List any tests ordered and what they are checking - List follow-up appointments or referrals with reason - List 3-4 specific actions the patient should take before the next visit - Avoid all medical abbreviations and jargon This summary will be reviewed and edited by the physician before being sent to the patient.
What to customize: For HIPAA compliance, anonymize the clinical notes before pasting into any AI tool that does not have a Business Associate Agreement (BAA) with your practice. Use placeholder names during drafting and add real patient information only in your HIPAA-compliant system.
Prompt 3: Prior Authorization Letter (Not a Diagnosis)
I am a physician writing a prior authorization letter to [PAYER NAME PLACEHOLDER] for [MEDICATION OR PROCEDURE] for a patient with [DIAGNOSIS CODE OR DESCRIPTION, de-identified for this draft]. Draft a prior authorization letter that includes: - The medical necessity statement for this treatment - Reference to the clinical criteria typically required by payers for this treatment type - A statement of treatment alternatives that were tried or considered - The urgency of the requested treatment (routine/urgent/emergent) - A request for expedited review if clinically indicated The letter should follow standard prior authorization format and use clinical language appropriate for a payer medical reviewer. This letter will be reviewed, modified with patient-specific details, and signed by the treating physician before submission.
What to customize: Add the specific payer's criteria guidelines if you have them. Different payers use different clinical criteria for the same medication. A letter that addresses the payer's specific criteria gets approved faster than a generic one.
Prompt 4: Appointment Preparation Instructions
Draft patient preparation instructions for a [PROCEDURE OR APPOINTMENT TYPE: e.g., colonoscopy, fasting blood draw, annual physical, new patient psychiatry intake]. Instructions should cover: - What the patient should do the day before (if applicable) - What to avoid eating, drinking, or taking (specify timeframes) - What medications to take or hold the morning of the appointment - What to bring to the appointment - What to wear or avoid wearing - How long the appointment will take and what happens during it - Who the patient can call if they have questions Format: numbered steps, easy to follow for a patient who has never had this procedure before. This draft will be reviewed by clinical staff before patient distribution.
What to customize: Add your practice's specific pre-procedure protocols. Some of these steps are practice-specific; the AI will provide a standard template that your clinical team should verify against your actual protocols and any equipment or anesthesia requirements.
Prompt 5: Referral Letter to Specialist
I am a physician writing a referral letter to a [specialty] specialist. Based on the following summary, draft a professional referral letter: Reason for referral: [one sentence clinical reason] Relevant history: [3-5 key clinical facts] Current medications: [list or placeholder] Recent relevant test results: [description or placeholder] Specific questions I need the specialist to address: [list 2-3 questions] Urgency: [routine / semi-urgent / urgent] Format the letter with standard referral structure: opening, clinical summary, reason for referral, specific questions, and a professional closing. Use clinical language appropriate for physician-to-physician communication. This letter will be reviewed and signed by the referring physician before sending.
What to customize: Include the patient's specific clinical findings and test results (anonymized for the drafting stage). Vague referral letters frustrate specialists; the more specific your input, the more useful the draft.
Prompt 6: Discharge Instructions for Common Conditions
Draft discharge instructions for a patient being released after treatment for [CONDITION, e.g., minor laceration repair, urinary tract infection, ankle sprain]. Instructions should cover: - What the patient was treated for (plain language description) - Medications prescribed and how to take them - Activity restrictions and for how long - Wound care or at-home treatment steps if applicable - Signs of improvement to look for - Warning signs that require immediate return or emergency care - When to follow up with their primary care physician - One sentence clarifying that these instructions do not replace follow-up medical care Reading level: 6th to 8th grade. Avoid all medical abbreviations. This draft will be reviewed by clinical or nursing staff before patient delivery.
What to customize: Add any practice-specific or procedure-specific instructions. Discharge instructions vary significantly by practice and local standard of care; treat this as a starting framework that clinical staff reviews against your protocols.
Prompt 7: Patient FAQ for Common Condition
Draft a Frequently Asked Questions document for patients recently diagnosed with [CONDITION]. Include 8-10 questions patients commonly ask, such as: - What caused this condition? - Will it get better on its own? - What lifestyle changes help? - What medications are commonly used? - When should I call my doctor? - What should I tell my employer or family? - Are there support groups or resources? Each answer should be 2-4 sentences in plain language. End every answer that touches on treatment with: "Your doctor will discuss what is right for your specific situation." This FAQ will be reviewed by a physician before distribution to patients.
What to customize: Add questions that your practice actually hears frequently for this condition. Your front desk or nursing staff likely know the top 5 questions patients ask; those are more valuable than generic FAQ content.
Prompt 8: Prescription Explanation for Patients
A patient has been prescribed [MEDICATION] for [CONDITION]. Draft a plain-language explanation that: - Describes what the medication does in simple terms - Explains how to take it (timing, with food or not, dose) - Lists the most common side effects to expect - Lists the side effects that should prompt a call to the office - Lists the side effects that are a medical emergency - Explains what to do if they miss a dose - Notes any common interactions (alcohol, other OTC medications if relevant) End with: "If you have questions about this medication, contact our office before stopping or changing your dose." Reading level: plain language for adults with no medical training. This will be reviewed by clinical staff before patient distribution.
What to customize: Verify all specific drug information against your pharmacy references or current prescribing information. AI may not reflect the most current prescribing guidance for recently approved medications or recent label changes.
Prompt 9: New Patient Welcome Letter
Draft a new patient welcome letter for [PRACTICE NAME PLACEHOLDER], a [practice type: e.g., family medicine, pediatrics, dermatology] practice. The letter should: - Welcome the patient warmly to the practice - Introduce the physician(s) briefly (I will add specific bios) - Explain what to bring to the first appointment (ID, insurance, medication list, prior records) - Describe how to reach the practice for urgent questions vs. non-urgent questions - Explain the after-hours and on-call policy - Note the patient portal if applicable - Set expectations for appointment scheduling and wait times - Close with genuine warmth Tone: warm, professional, and reassuring. The patient may be anxious about establishing care with a new practice.
What to customize: Add your practice's specific contact information, portal name, and after-hours protocol. Include the physician's name and a brief personal note if your practice style supports it. Generic welcome letters get filed; personal ones get read.
Prompt 10: Wellness Check-In Follow-Up Email
Draft a follow-up email for patients seen for a [visit type: e.g., annual physical, diabetes check-in, post-procedure follow-up] 2-3 days after their visit. The email should: - Reference that we saw them recently (not the specific diagnosis or details in the email) - Ask how they are feeling since the visit - Remind them of any key action items from their visit summary - Provide the office number for questions - Remind them of any upcoming lab work or follow-up appointments - Invite them to use the patient portal for non-urgent questions Tone: friendly and caring, not clinical. This is a touchpoint, not a medical communication. Important: Do not include any protected health information in the email body. All clinical details should be in the patient portal or provided by phone.
What to customize: Verify that your patient communication platform sends emails through a HIPAA-compliant channel. Do not include diagnosis codes, medication names, or specific clinical findings in unencrypted email regardless of how the AI drafts it.
AI-Assisted vs. Manual Patient Communication Time Comparison
| Communication Task | Manual Time (avg) | With AI Prompt | Time Saved | Clinical Review Required |
|---|---|---|---|---|
| Patient education handout (1 condition) | 45 min | 10 min | 78% | Yes |
| After-visit summary rewrite | 15 min | 4 min | 73% | Yes |
| Prior authorization letter draft | 30 min | 8 min | 73% | Yes |
| Discharge instructions | 20 min | 5 min | 75% | Yes |
| Referral letter to specialist | 25 min | 7 min | 72% | Yes |
| New patient welcome letter | 30 min | 6 min | 80% | Yes |
Time estimates reflect typical drafting time reported by practice managers and medical assistants. Physician review time is additional and varies by complexity.
Worked Example: Patient Education Handout in Practice
The prompt used: Prompt 1 above, with condition "type 2 diabetes, newly diagnosed patient, adult audience."
What the AI returned (excerpt): "What is Type 2 Diabetes? Type 2 diabetes means your body has trouble using insulin, a hormone that helps move sugar from your blood into your cells for energy. Over time, your blood sugar levels stay higher than they should. This can cause problems with your kidneys, eyes, nerves, and heart if left unmanaged. The good news is that for many people, lifestyle changes can make a real difference. Steps you can take at home: 1) Check your blood sugar as often as your doctor recommends and write down the numbers to share at your next visit. 2) Choose water, sparkling water, or unsweetened drinks instead of soda, juice, or sweetened coffee drinks. 3) Fill half your plate with vegetables at meals. 4) Take a 10-15 minute walk after dinner most nights. 5) Take your medications exactly as prescribed, even when you feel fine. Call our office if: your blood sugar is consistently above [target your doctor gave you], you feel dizzy or very weak, or you have questions about your medications."
What the physician did next: Added the practice-specific blood sugar target, adjusted the activity recommendation for the patient's mobility limitations, added the practice phone number, and approved the handout for distribution the same day, a task that previously took 30-45 minutes to draft from scratch.
What AI Should Never Do for Doctors
These boundaries exist because the risks in healthcare are not administrative. They are clinical and legal:
- No patient data in non-BAA AI tools: Protected health information (PHI) cannot be entered into any AI platform that does not have a signed HIPAA Business Associate Agreement with your practice. This includes patient names, dates of birth, diagnosis codes, and any detail that could identify a patient. Draft with anonymized placeholders and add real patient information only in your HIPAA-compliant EHR or communication system.
- No clinical diagnosis or treatment recommendations: AI-generated content for patients must not contain diagnostic conclusions or specific treatment recommendations. Patient education handouts explain what conditions are and what patients can do generally; they do not tell a specific patient what their diagnosis is or what treatment they should receive. The physician makes those determinations.
- No prescribing guidance from AI alone: Medication explanations drafted by AI must be verified against current prescribing information, especially for recently approved drugs, drugs with recent label changes, or complex dosing regimens. AI training data has a cutoff and may not reflect current FDA guidance.
- No unsupervised patient communication: Any message, letter, or document that reaches a patient under the practice's name carries the practice's professional and legal responsibility. AI-generated content must be reviewed and approved by a qualified healthcare professional before delivery. The physician is accountable for everything sent in the practice's name.
- No mental health content without heightened review: AI prompts for mental health patient communication require extra caution. Content related to depression, suicidality, eating disorders, or substance use requires clinical review that goes beyond standard communication review. These topics carry distinct clinical and liability risks.
Frequently Asked Questions
Can our practice use ChatGPT or Claude for patient communications?
General AI tools can be used for drafting patient communication templates when no real patient data is included in the prompt. For workflows where patient-specific information is part of the input, you need a HIPAA-compliant AI platform with a signed Business Associate Agreement. Several healthcare-specific AI vendors offer compliant platforms. Check with your practice's compliance officer before choosing a tool.
Does AI-generated patient education content need a disclaimer?
Yes. Every patient education document should include a statement that the information is general in nature and not a substitute for individualized medical advice from the patient's own physician. This applies to AI-generated and human-drafted content equally. A standard disclaimer reviewed by your practice attorney or risk management team should be added to all patient education materials.
How do we make sure AI content reflects our specialty's current standards?
AI training data has a cutoff date and may not reflect recent clinical guideline updates from organizations like the AHA, ADA, or USPSTF. Before approving any AI-drafted clinical content, a physician should verify it against current guidelines from the relevant specialty society. For conditions where guidelines change frequently, this review is especially important.
Can we use AI to help with prior authorization letters?
Yes, with two conditions: the letter must be reviewed and signed by the treating physician, and the clinical details added must be accurate and patient-specific. AI can produce a solid structural template and standard medical necessity language; the physician adds the patient-specific clinical facts and verification that the payer's criteria are addressed. Do not submit any AI-drafted letter without this clinical review.
What is the biggest risk of using AI incorrectly in a medical practice?
The biggest risk is distributing AI-generated clinical content to patients without physician review. A handout with incorrect dosing, an inaccurate symptom description, or outdated guideline information that was not caught in review can harm a patient and expose the practice to liability. The solution is a simple approval workflow: AI drafts, physician or qualified clinical staff reviews, then the document is released. AI saves time; the review step protects patients and the practice.