How Physical Therapists Can Use ChatGPT to Write Home Exercise Programs and Patient Education Handouts (2025–2026 Clinical Guide)

Free AI Prompts for Physical Therapists

Document faster. Download 3 copy-paste AI templates to speed up your evaluations, progress notes, and home exercise program setups.

Send My Free Prompts

We respect your privacy. Unsubscribe at any time.

Built with Kit

The Problem: HEPs Take Too Long and Patients Don't Follow Them

Home exercise program adherence is one of the most persistent failure points in physical therapy outcomes. Research consistently shows that patient non-adherence to HEPs ranges from 50–70%, and one of the primary drivers is that handouts are too complex, too generic, or too time-consuming for therapists to personalize properly under productivity pressure.

In a typical outpatient day, a PT seeing 10–14 patients may need to produce 4–6 individualized HEPs—each requiring condition-specific exercise selection, rep schemes, written instructions, precautions, and, increasingly, plain-language explanations that meet a 6th–8th grade reading standard for health literacy. For multilingual patient populations, this workload multiplies further.

Regulatory context adds another layer of pressure. Under Medicare's Benefit Policy Manual (Chapter 15, Section 220), documentation must reflect individualized, skilled care—meaning cookie-cutter printed handouts that don't reflect the patient's actual impairments can be flagged during audits. The FSBPT Model Practice Act (updated guidance, 2024) further establishes that the physical therapist is "fully responsible for managing all aspects of physical therapy care"—including any AI-generated content delivered to patients. The PT's name on the handout means the PT owns the content, unconditionally.

The result: PTs either rush HEPs (risking poor patient outcomes and audit exposure) or skip personalization entirely (risking both outcomes and engagement). ChatGPT offers a third path.

HEP Creation Workflow: AI-Assisted vs. Traditional

*PT clinical review and approval is non-negotiable regardless of method.

Step-by-Step Protocol: ChatGPT HEP and Patient Education Workflow

Step 1: Strip All PHI Before Prompting

Before opening ChatGPT, remove all Protected Health Information from your working notes. Do not enter patient name, date of birth, MRN, insurance ID, or any direct identifiers. Convert clinical facts to de-identified descriptors: "47F post-op rotator cuff repair, 8 weeks status post SLAP repair" is clinically rich and fully de-identified.

Why it matters: Standard ChatGPT does not hold a Business Associate Agreement (BAA) with your practice. Inputting PHI into a non-BAA platform creates a potential HIPAA violation under 45 CFR §164.502, regardless of intent.

Step 2: Assemble Your Prompt Variables

Before writing your prompt, gather the following clinical data points you'll fill into your template:

• Diagnosis or body region
• Specific exercises (by name)
• Sets, reps, frequency, and duration
• Weight-bearing or movement precautions
• Patient's approximate reading level or language
• Desired format (numbered list, table, or paragraph)

Step 3: Run the Prompt — First Draft

Paste your prepared prompt into ChatGPT. A complete HEP prompt takes 20–30 seconds to fill out if you have a template. The AI will return a full draft with exercise descriptions, technique cues, rep schemes, and precautions in the format you specified.

Step 4: Clinical Review — This Step Is Mandatory

Read the output against your clinical reasoning. Verify that:

• Exercise technique descriptions are biomechanically accurate
• Precautions match the patient's presentation and post-op/medical status
• Language complexity matches the patient's actual comprehension level
• No AI hallucinations have inserted exercises you did not prescribe

Step 5: Personalize and Finalize

Add the patient's first name to the handout header (outside the AI platform, in your EMR or Word document). Add your clinic logo, contact information, and the date. Sign or initial per your clinic's documentation policy.

Step 6: Document Your Workflow (Optional but Recommended)

Several state PT licensing boards are developing documentation policies around AI-assisted clinical content. As a best practice, note in your workflow that AI drafting tools were used and that the PT reviewed and approved all content prior to patient delivery. This positions you defensibly if audited or if AI-in-healthcare regulations tighten further—a trajectory clearly underway per the College of Physiotherapists of Manitoba AI Guidelines (updated July 2025).

Common Mistakes That Create Clinical and Compliance Risk

1. Entering identifiable patient information into standard ChatGPT.
This is the most critical error. Even partial identifiers (name + diagnosis + date) can constitute PHI under HIPAA. Always de-identify before prompting, or use a BAA-covered platform.

2. Distributing AI output without clinical review.
ChatGPT can and does produce biomechanical errors, incorrect rep schemes, or contraindicated exercises—particularly for post-surgical or complex neurological presentations. One unreviewed output reaching a patient is a liability event, not just a quality problem.

3. Using generic prompts that produce generic handouts.
A prompt that says "write a shoulder HEP" will produce a generic shoulder HEP. The clinical value—and the Medicare audit defensibility—comes from specificity: diagnosis, impairment level, precautions, and functional goals should all be in the prompt.

4. Failing to adjust reading level or language.
The National Institutes of Health recommends patient education materials be written at a 6th–7th grade reading level. AI output defaults to approximately 10th–12th grade without explicit instruction. Specify reading level in every prompt.

5. Assuming AI-generated content is automatically unbiased or evidence-based.
A 2025 PubMed study evaluating ChatGPT against clinical practice guidelines found the AI was consistent with CPG recommendations only 80% of the time overall—dropping to 60% for spinal conditions. HEPs for complex or surgical spinal cases deserve especially careful review before distribution.

Why This Matters Beyond Time Savings

The administrative burden on physical therapists is not merely an inconvenience—it is a documented driver of professional burnout, early career exit, and reduced patient outcomes. When a PT spends 15 minutes producing a HEP that a patient won't read because it's written at a college level, both the clinician's time and the patient's recovery suffer.

AI-assisted HEP and patient education workflows don't replace clinical judgment—they protect it. By compressing rote drafting tasks into minutes, therapists reclaim cognitive bandwidth for what EMRs and AI cannot replicate: hands-on assessment, therapeutic alliance, and clinical reasoning under uncertainty. In a profession where the PT's license and reputation are the product, working faster without working sloppier is not a luxury. It's a professional obligation.