Gains:
- Ability to produce a structured exercise program draft with AI according to the patient's diagnosis, goals and limitations
- Ability to test AI's recommended set, repetition, intensity and progression logic with clinical wisdom and contraindications
- Ability to adapt the program draft according to patient-specific pain, stage and comorbidity and prepare it for physiotherapist approval
Exercise prescription is the heart of physiotherapy. The right movement, the right load, the right frequency and the right progression; All of these are designed according to the patient's diagnosis, recovery stage, goal and daily life. This construction is a serious thought and takes time. Artificial intelligence (AI) can sketch out this work in minutes, freeing you up for thinking and clinical reasoning. But remember: the program that AI produces is a starting point, not a prescription. In this unit you will learn how to safely use AI to produce an exercise program outline, filter the output with clinical intelligence and adapt it to the patient.
Components of a good program outline
An exercise program is not just a list of "do this exercise this many times." A clinically sound program has the following components: goal (for what: pain reduction, range of motion gain, strengthening, functional return), exercise selection (stage-appropriate movements), dosage (set, rep, load, rest; FITT principle: Frequency-Intensity-Time-Type), progression rule (when to make it harder), regression (how to back off if pain increases), safety margins (whatever stops, stops), and narrative (what the patient will understand). with tongue). AI can produce all of these components in draft form; your job is to filter each one by examination.
Tip: Don't ask the AI for the program directly; Clarify the goal and phase first. "Strengthening goal in the chronic phase" and "pain management in the acute exacerbation" are completely different programs. If the thread is wrong, even the best prompt will produce the wrong program.
Step by step: Program draft with AI
- Complete the examination. He should have the diagnosis, stage, red flag screening, and basic measurements in hand before going to AI.
- An anonymous profile appears. Age, gender, diagnosis, stage, target, comorbidity, limitations; No name or ID.
- Give configured prompt. Write down the goal, phase, equipment, duration, and safety rule.
- Take the draft and filter it. Check each exercise for contraindications, pain and stage.
- Adjust dosage with clinical judgment. The set/rep the AI gives is a general assumption; adjust according to the patient.
- Add progression and regression rule. Clear rules such as "If the pain exceeds 3/10, stop and talk in the next session."
- Adapt to patient and confirm. Simplify the language, confirm as a physiotherapist, inform the patient.
Three mini cases (in numbers)
Case 1 — Chronic low back pain. 42 year old office worker, mechanical lower back pain for 4 months, no red flags, pain at rest 2/10, with activity 5/10. AI was asked to do 2 sets × 10 repetitions of core stabilization + hip mobility outline, 4 days a week. Since the patient's sitting tolerance was low, the physiotherapist reduced the repetitions to 2 sets × 8 in the first week and added the rule to stop if the pain exceeded 4/10. In 6 weeks, activity pain decreased from 5/10 to 2/10.
Case 2 — After anterior cruciate ligament surgery (4th week post-op). 24-year-old amateur football player. AI produced a general “knee strengthening” outline but did not know the surgical protocol and recommended open kinetic chain exercise at week 4. The physical therapist removed it because the patient's surgeon's protocol restricted this movement at this stage. Lesson: AI can improvise post-op protocol; The surgeon's protocol is essential.
Case 3 — Frozen shoulder (adhesive capsulitis), freezing phase. 55 years old, diabetic. AI recommended aggressive stretching. Knowing that aggressive stretching during the freezing phase could increase pain and inflammation, the physical therapist reduced the dosage to a light, pain-free range of motion and planned frequent but short sessions spread throughout the day. Passive external rotation increased from 10° to 35° in 8 weeks.
Weak prompt / Strong prompt
Weak: "Prescribe an exercise program for knee pain."
The diagnosis is unclear, there is no stage, there is no equipment and safety rules; the output becomes generic and risky.
Strong: "Produce a DRAFT of a 6-week home exercise program for a 30-year-old female runner diagnosed with patellofemoral pain syndrome. No acute exacerbations, pain 4/10 on stair descent. Equipment: resistance band and body weight. Focused on hips and quadriceps, with pain-free load principle. Suggest sets/reps for each exercise but note 'these are initial assumptions, clinical adjustment required'. Add progression and regression rule. Explain in plain language."
The powerful prompt gives clear diagnosis, stage, equipment, safety policy and role of AI.
Four copyable templates
Role: Physiotherapy exercise assistant (clinical decision is mine). Task: Produce a DRAFT of exercise program. Anonymous profile: [age, gender, diagnosis, stage, goal, equipment, restriction/comorbidity]. Required: 5-8 exercises, goal + sets/repetitions for each (as initial assumption). Mandatory: (1) contraindication warnings in separate title, (2) progression rule, (3) regression rule, (4) "stop whatever" safety limit. Write "CONFIRMATION REQUIRED" for each dosage you are unsure of. Plain language.
Task: Scan the following program outline for CONTRAINDICATIONS.Program: [draft]Profile: [anonymous patient profile]I want: Possible contraindication/caution point for each exercise +in which case it should be avoided. Don't decide, mark the risk.
Task: Adapt the following exercise to the patient's stage. Exercise: [exercise + current dosage]Target stage: [acute / subacute / chronic]I want: Regression and progression variant suitable for this stage, taking into account the pain-load relationship. Write the criteria clearly (e.g. "pain remains <3/10").
Task: Translate program outline into patient language.Draft: [program in clinical terms]Rule: 8th grade reading level, short sentence, "why" explanation, 1 safety tip for each action. Simplify medical jargon but don't create misunderstandings.
Dosage: AI's weakest point
AI often fills in the number of sets and reps with “typical” values (like 3×10, 3×12). These values do not know the patient's true tolerance, pain, recovery phase and goal. The table below summarizes how to filter AI output with clinical wisdom.
Variable
Typical output of AI
Physiotherapist's filter
Set/rep
Standard 3×10
Adjust by tolerance, stage, goal
Load/density
General "medium"
Determine based on pain response and strength measurement
frequency
3 days a week
According to recovery capacity and life load
progress
Uncertain or absent
Add rule with clear criteria
Contraindication
Often skipped
Be sure to check according to the examination
Caution: In post-operative patients, the general protocol produced by the AI may conflict with the surgeon's specific protocol. The protocol given by the surgeon/physician always takes precedence; AI output cannot override it.
Measuring progress
The weakest point of a program is often the question "when to make it harder?" AI leaves this vague (like “increase when patient is ready”). However, a good progression rule is criterion: for example, "if the patient completes the current load for 2 consecutive sessions without pain (pain <3/10) and with the correct technique, increase the load by 10%". Likewise, regression (withdrawal) should be criterion: "If the pain lasts longer than 48 hours or exceeds 4/10, return to the previous step." If you print these criteria into the AI and adjust it clinically, the patient can make the right decision on his own and the program will proceed safely. A program without measure either holds the patient back or over-accelerates and inflames him; Both delay treatment.
Tip: Define a three-step "ladder" (easy/medium/target) for each exercise and write clearly the passing criteria. When the patient knows which step he is on and how to move up, compliance and confidence increase significantly.
Common mistakes
- Not specifying the stage. The acute and chronic program are completely different; If there is no thread, the output will be risky.
- Giving the AI dosage as is. Set/rep is the general assumption; It should not be given without adaptation to the patient.
- Skipping contraindication. AI often does not prescribe contraindications on its own; you should ask.
- Not setting regression rules. If the pain increases, what to do is not written, the patient will make the wrong decision on his own.
- Leaving the post-op protocol to AI. Surgical protocol is essential; AI doesn't know it or makes it up.
- Ignoring the patient's goal. Pain reduction and return to sports require different programs.
In summary
AI quickly outlines your exercise program, freeing you up clinical reasoning time. But dosage, phase compliance, contraindications and progression rules must pass through your clinical filter. Give structured prompt (diagnosis, stage, target, equipment, safety rule), draft the output, adapt each exercise to the patient and approve as a physiotherapist. The surgeon's plan is essential in post-op protocols.
Application task
Select an anonymous patient profile and generate a draft program with the first template. Then scan the second template and the same draft for contraindications. Change at least three dosage values using clinical wisdom and justify why you changed in one sentence. Finally, with the fourth template, translate the program into the patient language and check its accuracy.
checklist
- [ ] I added the diagnosis, stage, target, equipment and safety rule in the prompt.
- [ ] I took the output as a draft, I did not use it directly as a recipe.
- [ ] I screened each exercise for contraindications and pain.
- [ ] I adjusted the dosage according to the patient's tolerance and stage.
- [ ] I added a progression and regression rule and a "stop whatever" limit.
- [ ] As for post-op, I prioritized the surgeon's protocol and approved the program as a physiotherapist.