AI Scribe for Occupational Therapists: A Buyer's Checklist

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for AI Scribe for Occupational Therapists: A Buyer's Checklist.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for AI Scribe for Occupational Therapists: A Buyer's Checklist.

Before buying an AI scribe, an occupational therapist should ask six things: can it produce your note and report headings, where do recordings and transcripts live, how is client consent captured, can you export everything, what does each clinician cost once the free tier ends, and can you test it on real sessions first. Anything vague on those six is a no.

The reason OT needs its own list is that the work does not sound like a standard consultation. Much of the evidence is silent or physical: you watch a child cut with scissors, you measure a doorway, you stand beside someone at a kitchen bench. A scribe that only hears words will miss that, so the buying test is how easily you can add what you observed, not only how well it transcribes what was said.

Follow me on Instagram@sagnikteaches

What an OT scribe has to cope with that a GP scribe does not

Most scribes are built around a doctor and patient talking in a consulting room. An occupational therapist might work in a clinic, a school, a hospital ward, a client's home or a workplace, often with a carer or teacher joining in, and often with long stretches where nobody speaks. Four features of OT sessions decide whether a scribe fits.

Connect on LinkedInSagnik Bhattacharya
  • Observation outweighs conversation. A grip assessment or a kitchen-safety walk-through produces findings that appear in no transcript. You will be dictating them afterwards, or typing prompts such as "add: client needed two attempts to transfer from bed to chair, no loss of balance".
  • Several speakers. A paediatric session may include a parent, a sibling and a teaching assistant. Ask whether the scribe separates speakers and how it labels people it cannot identify.
  • Goal-based documentation. OT notes hang on goals and measurable change. If a scribe writes a smooth narrative but drops the baseline score and target date, it is a worse tool than a template with blanks.
  • Reports for other people. Funder reports, school reports, equipment justifications and home-modification recommendations are read by people who are not clinicians. They are longer, formal and often the real documentation burden.

The checklist, grouped by what it protects

Take these questions to a demo or put them in an email. Score each answer 0 (no answer or "we are working on it"), 1 (yes, but with a caveat) or 2 (yes, shown on screen). Twenty-two questions give a top score of 44; a tool scoring under 30 is not ready for client data, and any zero in the consent, storage or exit groups ends the trial.

Subscribe on YouTube@sagnikteaches

Templates and clinical language

  1. Can I build my own template with my own headings, for example "Occupational performance", "Environment", "Goals and baseline", "Plan"?
  2. Can I save separate templates for a paediatric session, an adult home assessment, a hand-therapy review and a progress note?
  3. Does it handle OT vocabulary correctly: activities of daily living, instrumental activities, sensory processing, fine motor, graded exposure, adaptive equipment names?
  4. Can I paste a past note in my own style and have the output imitate it? Freed, for example, says you can edit a note and ask it to learn your style, so this is a reasonable thing to expect.
  5. Can I add observations by typing or dictating after the session without the transcript overwriting them?
The Complete Excel Course with AI Integration thumbnail

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

Reports and outputs

  1. Can it draft a longer report from several sessions, or only from one recording at a time?
  2. Can I paste in assessment scores myself and have them carried into the report exactly, with no rounding or rewording?
  3. Can it produce a plain-language summary for a parent, carer or school alongside the clinical version?
  4. Does it flag anything it inferred rather than heard? A scribe that writes "client lives alone" when the carer only said "she is on her own in the mornings" has created a fact.

Consent, privacy and storage

  1. Is audio kept, and if so for how long? Can I choose to delete it as soon as the note is generated?
  2. Where are recordings and transcripts stored, and can I choose the region?
  3. Is client data used to train the vendor's models, and is that switched off by default or only on request?
  4. Will the vendor sign a data-processing agreement, or a business associate agreement where one applies, before I put a single real session through?
  5. Does it support a consent step, such as a recorded verbal consent or a stored form, so the chart shows consent was given?
  6. What happens to my data and any backups if I cancel?

Fit with your record system

  1. Does the note paste into, or sync with, the system where I keep records, or will I be copying text by hand?
  2. Does it work on a phone for home visits, offline or with a patchy signal, and what happens to an interrupted recording?

Price and exit

  1. What is the price per clinician per month once any free allowance is used?
  2. Is there a cap on notes, minutes or reports, and what happens at the cap: extra charges, throttling or a hard stop?
  3. Is the price per clinician or per clinic, and does a part-time clinician or a student cost the same as a full-time one?
  4. Can I cancel monthly, and is annual billing the only way to get the headline price?
  5. Can I export every note, transcript and template in a format another system can open?

Price per clinician: what the vendors list now

Prices change often, so treat this table as a snapshot taken from the vendors' own pages in November 2026 and confirm each before you sign. All figures are list prices in USD.

ToolFree optionPaid scribe price listedWhat to check
HeidiFree plan with unlimited ambient documentation and standard templatesClinician plan $110 per user a month, billed annually, with custom templates for notes and documentsCustom templates sit on the paid tier, which matters for OT headings; Heidi's patient page says it does not store audio
JaneAI Scribe Free: 5 notes a month, on by default per staff profile$15 a month per opted-in practitioner for unlimited notesNeeds a Jane plan underneath; recording deletion defaults to immediate
Freed7-day free trial, no card neededStarter $39 a month for up to 40 notes; Core $79 a month with unlimited notesWhether the 40-note cap covers your caseload, and whether the price shown is for monthly or annual billing (the page has a toggle); Freed says audio is deleted automatically once the note and its quality checks are done
CarepatronFree plan includes an AI scribe and note taker, with a 1M AI token allowancePlus listed at $31 a month per user, with unlimited storage and AI tokens (a 50 per cent offer for six months was showing when checked)Token allowance, storage (1GB on free) and what happens at the limit

Two things in that table decide the answer for a solo OT. First, "free" often means limited templates or a cap on notes, not a free trial in disguise, so run your real caseload through the free tier before concluding it is enough. Second, a per-clinician price that looks cheap can be paired with a practice-management plan you also need to buy. Jane's scribe, for instance, is an add-on to a Jane plan, so the real monthly cost is the plan plus $15.

A worked comparison: one solo OT, two weeks, four tools

Illustration only. Say a solo OT sees about 24 clients a week: 14 paediatric clinic sessions, 6 adult home visits and 4 sessions that are mostly report writing. She writes a 150-word progress note after each session and 6 funder reports a month of about 900 words each, taking roughly 40 minutes per report.

Her current load is 24 notes a week at about 8 minutes each, which is 192 minutes, plus 6 reports at 40 minutes a month, so about 4 hours a month on reports. That gives a baseline of around 3.2 hours a week on notes and 1 hour a week on reports: 4.2 hours.

She tests the free tiers first. Jane's five free notes a month cover one afternoon, not a caseload, so Jane is only a real option if she pays $15. Heidi's free plan allows unlimited ambient documentation but she wants custom templates for her three session types, which Heidi lists under the paid Clinician plan. Carepatron's free plan includes an AI scribe, but 1M tokens and 1GB of storage will not last long if she records audio and keeps it. Freed's $39 plan caps at 40 notes a month, below her roughly 100.

So the realistic paid costs for her are $15 a month for Jane's scribe plus the Jane plan she already uses, about $79 a month on Freed's Core plan with unlimited notes, or a Heidi Clinician plan at $110 a month billed annually. Against 4.2 hours of weekly documentation, if a scribe really cut note time from 8 minutes to 3 after review (a figure to measure, not assume), she would save about 2 hours a week, or 8 to 9 hours a month. At any plausible hourly value, the licence is not what decides the choice. Template control, consent handling and where the data sits are.

Testing OT templates with a real prompt

Many scribes let you set an instruction that shapes every note. Write one for each session type and test it before you test the vendor's own templates. Here is an example for a paediatric clinic session.

You are drafting an occupational therapy progress note from a recorded session.
Use these headings exactly: Presenting focus, Activities and performance, Observations, Goal progress, Plan.
Rules:
- Quote any score or measurement exactly as spoken. Never calculate or round.
- Under Goal progress, list each goal on its own line with baseline, today's result and target date if stated; write "not discussed" if absent.
- Do not add diagnoses, causes or recommendations that were not said.
- Use "child" or "[first name]", never a surname.
- If something is unclear, write "[check]" instead of guessing.

A realistic output from a good scribe, using an illustrative session, looks like this.

Presenting focus: Handwriting stamina and pencil grip in class.
Activities and performance: Copied a 4-line sentence; grip changed from tripod to fisted after line 2.
Observations: [check] shoulder elevation noted by therapist but side not stated.
Goal progress: Goal 1, write 20 words legibly. Baseline 8 words, today 12 words. Target date not stated.
Plan: Continue weighted pencil trial; parent to practise 5 minutes daily.

What you would fix: the "[check]" line shows the instruction working, because the scribe flagged a missing detail rather than inventing "left". The goal line is useful but incomplete, so add the target date. If the output had written "child has dysgraphia", that is a diagnosis nobody said, and the template needs a stronger rule or a different tool.

Consent and storage questions specific to OT settings

Recording in a client's home, a school or a care facility raises questions a consulting room does not. The person who agrees may not be the person recorded, and other people can walk into the audio.

  • Children. You need agreement from the parent or guardian, and where the child is old enough, their own. Ask how the tool handles a child's voice and whether the recording can be deleted the moment the note is accepted. Jane, for example, lets a Full Access user pick a clinic-wide policy of keep indefinitely, delete after a set number of days, or delete immediately after the draft is generated, and says the default after enabling the scribe is immediate deletion.
  • Third parties. A teacher, support worker or neighbour may be in the room. Tell them too, or pause the recording.
  • Home visits. You may be on a personal phone on mobile data. Ask whether the app is allowed on a work profile, how audio is encrypted in transit, and what happens if the phone is lost.
  • Employer or funder rules. If you work for an organisation or are paid by a funding body, its policy may forbid tools it has not assessed, whatever the vendor says. Get that in writing before the trial.

Jane's help guides also note that a deleted recording is removed from the clinic account at once but a copy stays in its backup servers and is permanently deleted after one week. That is exactly the sort of detail to put in your consent wording, because "deleted" can mean different things. For wording that clients accept, see consent wording for AI note-taking, and for the wider confidentiality picture, a confidentiality checklist for small practices.

A consent script for a home visit

Keep it short enough to say at the door. Illustrative wording:

"Before we start, I use an AI tool that listens to our session and drafts my notes. The recording is used only to write the notes, I check every word, and it is deleted [straight after / within X days]. You can say no and I will write notes by hand, and you can ask me to pause it at any point. Is that all right?"

Then record that consent in the chart in one line: date, who agreed, and whether the agreement was verbal or written. If a client declines, the practical test is whether your workflow copes without a scribe that day. Plan for it, because some clients will say no, and you do not want a policy that makes you choose between consent and finishing notes.

Running a fair trial in two weeks

Most buyers watch a demo, try the tool on one friendly session and sign. Run a structured trial instead, and set the pass marks before you start so enthusiasm does not move them.

  1. Pick the sessions. Ten to fifteen real sessions, covering at least a paediatric clinic session, a home visit and a report-writing day.
  2. Time the baseline. For the previous week, note how long each note and report took without a scribe. Without that number you cannot show a saving.
  3. Score every output on three things: accuracy (count factual errors per note), completeness (did goals, scores and plan appear), and edit time (minutes from draft to signed-off note).
  4. Log the misses. Keep a list: wrong equipment name, merged speakers, missing observation, invented detail.
  5. Test the exit. Before day 14, export a week's notes and open them in another program. If you cannot, treat that as a failure.

Illustrative pass marks: fewer than one factual error per three notes, edit time at least 40 per cent below your baseline, no invented diagnoses, and a working export. A second vendor trial in the following fortnight, with the same sessions, is worth the effort because the differences between tools are bigger than the differences between a tool and no tool.

Four ways a scribe goes wrong in OT notes

  • Merged measurements. A scribe hears "forty-five degrees" and "four five centimetres" in one session and attaches the wrong unit. Defence: tell it to quote numbers verbatim and check every number against your own record.
  • Carer words becoming client facts. "She doesn't eat much" from a daughter turns into "client reports reduced appetite". Defence: a template line "Source of information" and a habit of scanning for who said what.
  • Silent observation lost. A forty-minute kitchen assessment with ten words of speech produces a three-line note. Defence: dictate a 60-second observation summary at the end of every visit; the scribe will fold it in.
  • Template drift. After a vendor update the headings you built change slightly. Defence: re-run one test note monthly and keep a copy of your instruction text outside the tool.

When a scribe is the wrong first purchase

If most of your documentation burden is reports rather than session notes, a scribe may help less than a good report template and a structured drafting prompt in a tool you already trust. If your records system does not let you paste or import text easily, the time saved on writing can be lost in copying. And if your employer has not approved any recording tool, the first job is an internal request, not a purchase. For a view across the wider category, the tutorial on whether an AI scribe is worth it for a small clinic sets out when the maths works. If you would rather map this with someone, that is what my AI implementation consultation covers, at the contact page.

The one-page summary to keep beside the demo

The 44-point score takes a sitting to complete. For the demo itself, a one-page version with a single verdict per question group is easier to fill in live. Print it and complete one per vendor. Illustrative example for one candidate:

Question groupGroup verdict (0-2)Note
Templates: own headings, saved per session type2Shown on screen, three templates built in the demo
Reports: scores carried exactly1Works for one session; multi-session reports not offered
Consent and deletion2Deletion policy selectable; consent note field available
Storage and training use1Region choice only on higher plan
Fit with records0Copy and paste only
Price and exit2Monthly cancellation; export of notes as PDF and text

This candidate fails on the spot: the zero for fit with records means hand-copying every note, and the 1 for storage needs a written answer on region before any real session goes through. Any zero in consent, storage or exit, or a full 44-point score under 30 once you have the vendor's written answers, is a reason to keep looking, however good the demo felt. You can also compare costs for more than one clinician in seat costs for group practices before you decide whether to buy for yourself or for a team.

Questions OTs ask before trialling a scribe

Can an AI scribe write a home visit assessment?

It can draft one from a dictated or recorded visit, but only if the template has your headings (access, bathroom, transfers, equipment). Test it on a real visit with the client's consent. Expect to correct measurements, equipment names and anything you only observed silently.

Do I need client consent to record a session?

Almost always you need to tell the client and get agreement before recording, and for children the parent or guardian. Rules differ by place and employer, so check your regulator, insurer and employer policy, then record the consent in the chart.

Is a free plan enough for a solo OT?

Sometimes. Heidi lists a free plan with unlimited ambient documentation and Carepatron's free plan includes an AI scribe, while Jane gives five free notes a month. Check the free tier's template, storage and export limits against your caseload before relying on it.

How many clients should I test with?

Run ten to fifteen real sessions across at least three client types, such as a paediatric block, an adult home visit and a report-writing day. Fewer than that hides the cases where the scribe struggles.

Further reads

Sources: Heidi, Jane, Freed and Carepatron pricing pages, and Jane's AI Scribe help guides (checked November 2026).

Want a scribe trial designed around your caseload?

On a 1:1 call we can look at your report types and client mix, pick two scribes worth testing and set the pass-or-fail scores before you commit to a plan.

Book a 1:1 call with me