ChatGPT for Occupational Therapy Goals: Drafting Without Client Data

Write a short profile of the client that contains no name, date of birth, address, school, employer or rare detail, and give ChatGPT that profile plus the goal structure you want. The tool drafts measurable goal wording; you set the baseline, target and timeframe from your assessment, and add identifiers only inside your own records.

The reason to work this way is practical, not fussy. Goals feel harmless because they are short, but the sentence that makes a goal good, such as "to carry her own tray at the school canteen by March", is the sentence that identifies a child. A de-identified profile gives you the structure without that exposure, and a private practice owner can run it in about a quarter of an hour per client.

Follow me on Instagram@sagnikteaches

What a goal needs before AI can help with it

ChatGPT is good at wording and structure, and has no idea what your client can actually do. So the information in the prompt must come from your assessment, and it has to be specific about function. Before you open the tool, gather five things in your own notes.

Connect on LinkedInSagnik Bhattacharya

• The activity or role that matters. Dressing, handwriting, preparing a hot drink, returning to part-time work, using public transport.
• The baseline. What the client does now, with what help, how often, how long. "Needs hand-over-hand help to fasten three of five buttons" beats "difficulty with fine motor skills".
• The target. What done looks like, in observable terms.
• The timeframe. Six weeks, twelve sessions, by the next review.
• The supports. Equipment, prompts, environment changes, carer involvement.

If you cannot fill one of these in, the problem is in the assessment, and no prompt will fix it. A useful side effect of the process is that the template exposes gaps in baseline data before the plan is written.

Subscribe on YouTube@sagnikteaches

De-identify the profile in four moves

You are not removing meaning; you are removing the details that point to a person. Do this on a scratch page, not in the chat window.

1. Replace the person with a role and an age band. "Adult client, 60s" or "primary-school child, 7 to 9". Never the exact age if the client has a rare condition.
2. Generalise places and people. "A local school" becomes "school". "Her sister, who lives next door" becomes "family support available daily".
3. Drop dates. Use "six weeks after surgery" instead of the surgery date, and "week 1, week 6" instead of calendar dates.
4. Remove rare specifics. An unusual hobby, a notable employer or a uniquely identifying history stays in your notes.

Diagnoses sit in a grey area. A common condition described generally is usually fine. A rare syndrome plus an age band and a profession can narrow the field to one family. When in doubt, describe the functional picture and leave the label out; occupational therapy goals are about what a person does, so the function is usually what the goal needs anyway.

Main example: goals for a child in a one-therapist practice

Take an illustrative private paediatric practice run by one therapist. She is working with a primary-school child who struggles with handwriting stamina. From her notes she writes the de-identified profile below and sends it with the prompt.

You are helping an occupational therapist draft client goals.
Use the SMART structure and write in functional, participation-
focused language. Do not add facts that are not in the profile.

Profile (de-identified):
- Child, age band 7 to 9, mainstream school
- Activity: handwriting during class writing tasks
- Baseline: writes legibly for about 4 minutes before letters become
  uneven and the child stops; copies 2 of 6 sentences in a class task
- Supports in place: pencil grip, slope board, movement break
- Review point: 12 weekly sessions plus a home programme
- Priority for the family: finishing class work without fatigue

Draft: one long-term goal and three short-term goals, each with
baseline, target, timeframe and how it will be measured. Then list
any assumption you made so I can check it.

An illustrative response gives a long-term goal that the child will complete a 6-sentence class writing task with legible handwriting and without a rest break, by week 12. Three short-term goals follow: sustain legible writing for 8 minutes by week 4, copy 4 of 6 sentences by week 8, and use the slope board and a movement break independently on 4 of 5 days by week 6. The assumptions list notes that the tool assumed legibility would be judged by the therapist's own scoring and that the 4 of 5 days measure depends on the teacher's cooperation.

The therapist now does the real work. She changes 8 minutes to 7 because her baseline of 4 minutes does not justify a doubling in four weeks. She swaps the teacher's logged measure for a simple home and clinic observation she controls. She adds a line that the goals are to be shared with the family at the first review. The edit takes a few minutes, and it is the clinical reasoning. The tool supplied a draft structure and the layout of the measures.

A prompt skeleton you can reuse for any client

Save the prompt shape and change only the profile. A consistent shape means you review the same things every time.

Role: assistant to an occupational therapist; draft only.
Structure: SMART; functional language; one long-term goal
and up to three short-term goals.
Inputs: [age band], [setting], [activity], [baseline measured],
[supports], [timeframe], [what matters to the client].
Rules: do not invent baselines, scores or diagnoses; do not use
names; flag assumptions; use plain language a client can read.
Output: goals as a numbered list, then "Assumptions to check".

The last line is the one that earns its keep. When the model lists its assumptions, you can see where it filled a gap, and gaps are where errors begin.

Four more goal drafts, each with a different problem

An adult returning to work after a hand injury

Profile: adult, 40s, desk-based role, six weeks after a hand repair, wants to return to typing and signing documents. A draft that reads "improve hand function" is useless. A reviewed goal reads: type for 20 minutes with a 5-minute break and no increase in pain above 3 out of 10, within 6 weeks. The edit that matters: the tool offered 30 minutes, and the therapist chose 20 because the surgeon's guidance allowed gradual loading.

An older adult wanting to stay independent in the kitchen

Profile: adult, 80s, lives alone, reduced standing tolerance, wants to prepare a hot meal. The draft proposes "prepare a simple hot meal safely". The therapist rewrites to a measurable form: prepare a one-pot meal using a perching stool and a trolley, with no rest breaks longer than 2 minutes, observed on two home visits. A safety line is added about the carer check-in routine, which the draft left out.

A teenager whose goal is theirs, not the family's

Profile: young person, 15 to 17, wants to travel to a youth club on the bus alone. The first draft centres on "reduce anxiety". The therapist reframes in the young person's words: travel the three-stop route alone on 3 of 4 trips, by week 10, with a plan for a missed stop. A goal that begins with the client's priority tends to survive review better than one written around the therapist's.

A client with a goal that should not be a goal

Sometimes the draft is wrong at the root. A parent asks for a goal about "perfect handwriting". The model dutifully drafts one. The therapist declines: perfect is unmeasurable and a poor fit for the child's developmental stage, and she replaces it with a stamina and legibility target. The tool will follow a bad brief confidently, so the professional judgement stays with you.

Rewriting a goal for the client to read

Clients and families tend to engage more with goals they can read without a glossary. After the clinical version is approved, ask the tool to rewrite it in plain language, using only the approved text as input. Take the illustrative handwriting goal. The clinical version reads: complete a 6-sentence class writing task with legible handwriting and without a rest break by week 12. The plain version might read: by the end of our twelve weeks, you will finish a short piece of class writing in your neatest handwriting without needing to stop. Read it aloud to check that it sounds like something a person would say to a seven-year-old's parent, and remove anything that sounds like a promise. A goal is a target, not a guarantee, and the plain version must not turn it into one.

Updating goals at a six-week review

A second use is the mid-point review, when goals need to be revised against progress. Give the tool the approved goals plus your new measures, again with no identifiers: "Goal 1 target 7 minutes; measured 6 minutes at week 6. Goal 2 target 4 of 6 sentences by week 8; measured 3 of 6." Ask it for a short progress statement and a proposal on whether to keep, revise or close each goal. A good draft says that goal 1 is on track, that goal 2 is slightly behind and may justify extra home practice, and flags that goal 3 has no new data. The therapist decides. If a goal was met early, the draft should propose a step up rather than closing the story; if the draft proposes a step that is bigger than the evidence supports, shrink it.

Goal attainment scaling as a cross-check

Some therapists use goal attainment scaling, which sets five levels for a goal from much less than expected to much more than expected, with the expected outcome at the middle. ChatGPT can propose the five levels from a de-identified goal, which is a handy way to test whether a goal is concrete enough. For the handwriting example, the middle level is 7 minutes of legible writing, the two levels above might be 8 to 9 minutes and 10 minutes or more, and the two below might be 5 to 6 minutes and the current 4 minutes, depending on where your convention places the baseline. If the draft levels all look similar to each other, the goal is too vague and needs sharper measures. This is a check on wording, not a substitute for your own scaling method, so use whichever convention your practice already follows.

A mistake that shows how drafts go wrong

An illustrative case: a therapist asks for goals for an adult after a stroke and includes "improved from 4 to 7 on the assessment". The model produces a tidy goal referring to "a score of 9 on the functional scale", a figure that appears nowhere in the input. It reads convincingly because it follows the pattern of earlier numbers. The therapist spots it only because her review step compares every number with her notes. The rule that comes out of this: never let a score, scale name or threshold survive review unless you can point to where it came from. If you add the instruction "use only numbers I gave you" and still find an extra one, treat that draft as untrustworthy and rewrite the whole thing.

Telling clients how you use these tools

Even when no client data enters the tool, a short line in your privacy notice and intake pack builds trust. A plain version: "I sometimes use software tools to help draft the wording of goals and templates. I never enter your name or personal details into them. I write, check and approve every goal myself." Adjust it to your professional body's guidance and your data-protection obligations, and have your own adviser read it if you are unsure. If a client asks not to have AI involved at all, it is easy to honour when the process is a drafting aid you can switch off.

The funder and insurer wording check

Many private practices write goals that go into reports for funders, insurers or schools. The wording those readers look for differs, and the safest check is not memory but the funder's own documents. Before sending, put each goal against a short checklist.

One more point: do not paste a funder's private guidance or a client's funding plan into the chat. If you want help matching a goal to a funder's headings, paste only the public headings and the goal text, both already de-identified. The wording check is yours; the tool can suggest alternatives, but only you know what the funder accepted last time.

Where the line sits with client data

If you find yourself wanting to upload a full assessment, a school report or an intake form, the task has moved out of de-identified drafting and into a different category. At that point you need a vendor that will sign an agreement covering the data, with terms that fit the kind of health information you hold. OpenAI's help centre, as checked on 4 November 2026, says it does not offer a business associate agreement for ChatGPT Business, routes Enterprise and Edu through sales, and allows eligible organisations with API history to accept a standard agreement for the API. Its individual route, ChatGPT for Clinicians, is designed for verified physicians, nurse practitioners, physician assistants and pharmacists, so it is not open to an occupational therapist. Other vendors draw the line in a similar place: Anthropic's privacy centre, for instance, says its BAA covers its API and Claude Enterprise organisations set up for health data, a far bigger purchase than a sole practitioner's plan. Terms change, so ask each vendor for the document in writing. The vendor comparison for therapists is in which vendors sign a BAA, and the rules for a whole practice are in the AI usage policy tutorial.

For session notes rather than goals, the safe routes differ from goal drafting; see the small-practice notes workflow and the physiotherapy treatment-notes tutorial.

Setting up the tool so it is hard to misuse

A few set-up choices make slips less likely. Create a ChatGPT Project (or a Claude Project) called "Goal drafting" and put the reusable prompt skeleton in its instructions, so you never retype the rules and never improvise a prompt under time pressure. Keep a short banner at the top of the project reminding you of the four de-identification moves. Use a personal login on a device only you use, switch off model training on a personal plan, and avoid connecting email, cloud storage or your practice software to the account through apps, because a connected app could surface real client files without you pasting anything. Finally, keep the scratch page where you de-identify inside your normal records system, not in a separate untracked document, so it is deleted under the same rules as everything else.

If you later add a second therapist, write the routine down at that point. A routine that lives only in one person's head cannot be checked, and goals are the sort of document a colleague will want to reuse.

A ten-minute review before any goal goes in a record

1. Baselines. Every number matches your own assessment, not the draft's guess.
2. Targets. Realistic for the timeframe and the client's condition; trim anything ambitious.
3. Language. Plain enough for the client to read; no jargon the family would need explained.
4. Safety. Carers, equipment and environment considerations are present where relevant.
5. Invented content. Everything in the "assumptions" list has been confirmed or deleted.
6. Consent and records. The finished goal is typed into your records system, and the draft chat is deleted.

A sole practitioner can combine these into a single checkbox sheet attached to each client file. For a group practice, make the reviewer someone other than the drafter at least once a month, so habits get tested.

Measuring whether the routine is worth it

Time ten goal-writing sessions without AI, and ten with the de-identified routine, then compare. Suppose the figures are 25 minutes by hand and 14 with the routine. For a therapist writing 6 sets of goals a week, that is 66 minutes saved a week, or around 57 hours across a 52-week year, using the same example numbers. Check your own, as the gain depends on how much time the profile-writing takes. Also track how often a funder or colleague queries a goal, because a faster process that produces more queries has not helped.

Finally, keep a small bank of reviewed goals that contain no client information: "prepare a hot drink", "dress the upper body with set-up", "write for 10 minutes". They become your starting points and reduce the number of prompts you send at all. For the wider admin load, the admin-stack tutorial for a therapy private practice shows where else a small clinic can save time safely. If you want to talk through your own set-up, my AI implementation consultation covers it, starting at the contact page.

Questions therapists ask about AI-drafted goals

Can I paste an assessment report into ChatGPT?

Not into a plan without a business associate agreement or equivalent data-processing terms. Assessment reports are full of names, dates, diagnoses and family detail. Write a short generic profile yourself and use that. Treat the report as a source you read, not a file you upload.

Are AI-drafted goals good enough for funders?

Only after your review. Funders and insurers look for measurable goals tied to function and to the service you provide. A draft can give you the structure and wording, but you decide the baseline, target and timeframe, because those come from your assessment.

Should I tell clients that AI helped draft goals?

Check your professional body's guidance and your own privacy notice. If no client data is entered, many practices simply say that they use software tools for drafting templates and that a therapist writes and approves every goal. Being open early avoids awkward questions later.

Which ChatGPT plan should a sole practitioner use?

For de-identified drafting, an individual plan with model training switched off is workable, though a business plan keeps content out of training by default. Business plans need two seats. None of these is a place for identifiable client data.

Further reads

• Therapy Practice Management Software: What the AI Add-Ons Cost — What AI add-ons in practice software cost.
• Best AI Software for Therapists in a Group Practice: Seat Costs — Seat costs when more than one therapist shares a tool.
• AI Tools for Physiotherapy Clinics: What Each One Actually Does — What each AI tool in a therapy clinic actually does.
• Using AI for Childcare Observations and Learning Journals — Drafting child observations without exposing identities.
• 10 Admin Tasks a Small Clinic Can Hand to AI This Month — Admin jobs a small clinic can hand over with no client data.
• Best AI Tools for Occupational Therapists: School, Clinic, Home Visit — One AI pick per job for school-contract, paediatric clinic and home-visit occupational therapists, with privacy checks, cost bands and a worked example each.
• AI for Occupational Therapy Documentation: A Monthly Audit Checklist — A monthly audit workflow for OT practice owners: sample de-identified notes, score them with a rubric, list gaps by clinician, re-audit.
• NDIS Occupational Therapy Report Template: A First Draft With AI — Report headings, an evidence table, a de-identified drafting method and a review checklist for an AI-drafted NDIS OT report you stay accountable for.
• How Can AI Help an Occupational Therapy Practice? Seven Admin Jobs — Seven admin jobs for an occupational therapy practice, from report drafts to billing checks, with clinical judgement left with the therapist.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI help centre articles on business associate agreements, ChatGPT for Clinicians, data controls and projects; OpenAI ChatGPT Business pricing page; Anthropic privacy centre article on BAAs for commercial customers; Claude help centre article on projects (all checked 4 November 2026); SMART goal and goal attainment scaling conventions as used in occupational therapy practice.