AI Scribe for Audiology: Fitting Notes, Counselling and Consent

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for AI Scribe for Audiology: Fitting Notes, Counselling and Consent.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for AI Scribe for Audiology: Fitting Notes, Counselling and Consent.

An audiologist can use a general clinical scribe such as Heidi or Jane, or a hearing-specific one such as HearScribe, which starts at 30 pounds a month for 100 consultations (it prices in pounds only). For consent, tell the patient before the fitting that an AI tool will record or transcribe, get their agreement, and offer a note-by-hand alternative.

One claim deserves a pause: HearScribe's home page says "no consent required", comparing the scribe to a pen. Its own privacy policy then explains that text goes to Google's Gemini service and that browser speech recognition may send audio to the browser maker. Whatever a vendor says, you are the one who has to be open with the patient.

Follow me on Instagram@sagnikteaches

What a fitting appointment asks of a scribe

Hearing-care appointments are different from a GP consultation, and a scribe that suits one does not always suit the other. A fitting mixes spoken counselling with numbers and device details that a language model easily scrambles. Think about the pieces of a single fitting.

Connect on LinkedInSagnik Bhattacharya
  • Numbers that matter. Thresholds in decibels, frequencies in hertz, gain targets and real-ear measurement results. If a scribe turns "four kilohertz" into "for kilohertz" or swaps left and right, the note is wrong in a way that is easy to miss.
  • Device names. Manufacturer, model, receiver length, dome type, ear mould details, programme names. Speech engines stumble on brand and product names.
  • Counselling talk. Expectations, adaptation to amplified sound, wearing schedule, cleaning, realistic limits. This is the part where an ambient scribe is genuinely helpful, because it is long, conversational and hard to type while you are facing the patient.
  • Patients who answer slowly or ask you to repeat. A recording of a fitting can contain a lot of repeats and rephrasing, and the scribe has to boil it down without dropping a safety point.
  • Relatives who speak for the patient. Family members often join a fitting. Their voices are on the recording, and they are part of the consent conversation.

So the useful test is not "does it write a tidy note". It is "does it get the numbers and devices right, and does it capture the counselling points without inventing any".

Subscribe on YouTube@sagnikteaches

Four options compared on what matters in a fitting clinic

Vendor pages were read in November 2026. Prices are list prices and change, so confirm on the vendor's page before you commit.

ToolBuilt forCost as listedNotes for audiology
HearScribeHearing care onlyPriced in pounds only: Starter 30 pounds a month for 100 consultations (up to 2 team members); Professional 50 pounds for 200 (up to 5); Practice 100 pounds for 500 (up to 15). 7-day free trial, card required.Says it interprets uploaded audiograms (type, severity and configuration), takes tympanograms and otoscopy images too, and produces four documents per consultation: clinical notes, a patient summary, admin tasks for reception and a report for the GP. Its policy says patient data is not stored by HearScribe and that text goes to Google's Gemini service.
HeidiAny clinicianFree plan listed with unlimited AI documentation and standard templates; Clinician plan $110 per user a month billed annually ($150 billed monthly), which adds advanced templates.You write or adapt a fitting template yourself. Its pricing FAQ says scribing never counts against the free plan's weekly usage allowance (that allowance is for its agent features), so the free plan's limit is templates, not note numbers.
JaneClinics run on Jane's practice softwareAI Scribe Free gives five notes a month per practitioner; Unlimited is $15 a month per practitioner, added to a Jane plan.Only makes sense if your appointments, charts and billing already live in Jane. Record from the chart, choose a template, review, then sign.
TandemClinicians in EuropeFree plan with unlimited note generation and limited daily "Pro Actions"; Pro listed at 125 euros per user a month billed yearly (the page I read priced in euros only).Its site says patient data is processed and stored in European data centres, audio is deleted immediately after transcription, and the scribe is a CE-marked medical device under EU MDR Class IIa.

If your appointments are mostly fittings and follow-ups and you want something shaped for that, HearScribe is the one to trial first. If you already use a general scribe for other work, test whether a good template inside it does the job at no extra cost. If your practice software is Jane, its built-in scribe saves you an extra login and a paste step. If you want European hosting and a free tier to test first, Tandem is worth a trial.

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

Reading "no consent required" against the privacy policy

Here is a worked example of how to check a vendor claim, using the HearScribe page as it stood when I read it. The home page says "no consent required" and compares the tool to a pen. A pen does not send anything anywhere. So read the privacy policy for what happens to the information.

  1. The policy describes a design where patient-identifiable information is kept in your browser while you work and is not stored on the vendor's servers.
  2. It says that when you generate notes, the consultation text passes through the vendor's systems transiently and is sent to Google's Gemini API.
  3. It says live transcription uses your web browser's built-in speech recognition, which may process audio through the browser provider's speech service, for example Google for Chrome or Microsoft for Edge.
  4. It says the vendor acts as your data processor and offers a data processing agreement, and that you remain responsible as controller for your obligations.

None of that is unusual, and the zero-storage design is a real privacy strength. But it means a third party's speech and language services touch your patient's words. A patient who is asked "may an AI tool listen to and process this appointment?" is entitled to say no. The pen comparison does not hold, so the sensible reading is that consent and transparency are still yours to handle. Do the same exercise for any scribe: find the sentence that says where the audio and text go, name each service, and make your consent wording match.

A fitting note, before and after the scribe

This is a fictional fitting for a patient with mild-to-moderate sensorineural loss, fitted with a receiver-in-canal device on both sides. Here is what a hurried hand-typed note might look like, followed by the structure to ask your scribe or template for.

Before (typed in a hurry):
"Fitted new aids. REM done. Pt happy. Told about adjusting.
F/u 4 wks."

After (template-driven draft, reviewed by the audiologist):
FITTING APPOINTMENT
Reason: Bilateral hearing aid fitting, first fit.
Devices: [manufacturer/model], receiver-in-canal, left and right.
  Receiver length: [__]. Dome: [type/size] left, [type/size] right.
Programming: First-fit prescription formula [name], adjusted
  to [__]% of target at first fit as agreed with patient.
Real-ear measurements: Completed both ears. Match to target
  within [__] dB across [__ to __] Hz. Values recorded in [system].
Patient response: Reports speech clear, own voice acceptable,
  some sharpness with clatter. Reduced high-frequency gain
  by [__] dB in programme 1.
Counselling given: Adaptation over 2-4 weeks, wearing schedule
  building from 4 hours to full day, cleaning and battery/charging,
  realistic expectations in noise, what to do if sore.
Patient/relative understanding: Teach-back completed; relative
  present and agreed to remain for counselling.
Plan: Follow-up in 4 weeks. Return sooner for discomfort or
  feedback. Written summary given.

The after-note has blanks because a scribe should not guess numbers. Your approach: leave the measurement lines as fields that you type or paste from your fitting software, and let the scribe fill the narrative parts, which are the response and counselling. Then you check it against what you remember of the room. When the scribe invents a figure, such as "within 3 dB", that is the error to catch first.

A counselling note the scribe can draft from a conversation

Counselling is where the ambient scribe pays its way. Say a patient fitted last month returns and spends ten minutes describing background noise in a café and difficulty on the phone. A scribe works from the conversation, so the draft carries the patient's own descriptions.

Sample prompt for your template:
"Write a counselling and follow-up note. Use only what was said.
Headings: Concerns raised, Situations that are hard, Advice
given, Changes made, Patient's goals, Plan. Quote the patient's
own phrases sparingly. If a measurement or setting was not
spoken, write [not stated]. Do not suggest new diagnoses."

Illustrative sample output:
Concerns raised: Cafe noise "swallows" speech; phone calls
  unclear on the left side.
Situations that are hard: Group conversation in busy rooms;
  television at low volume.
Advice given: Face the speaker, choose seats against a wall,
  use the phone streaming option, tell people about the hearing loss.
Changes made: Directional programme adjusted for noisy places;
  phone programme enabled. [Values not stated]
Patient's goals: Enjoy weekly lunch with friends.
Plan: Review in 3 weeks; trial the noise programme in a
  real cafe and report back.

Two things to fix in a real draft. First, scribes tend to turn tentative remarks into firm statements: a patient who says "I think it's the left" should not become "left-sided difficulty confirmed". Second, check that advice you gave out of habit, such as cleaning instructions, did not get recorded as something you did not say. Delete anything you cannot recall saying.

Consent wording for a patient who is hard of hearing

This is where audiology differs from other clinics. A spoken request for consent to record is awkward when the patient may not hear it. Prepare more than one way to ask and make the choice easy to refuse.

Large-print line for the intake form or email before the visit:
"We use an AI tool to help write your appointment notes. It
listens to our conversation, makes a draft, and I check it.
It sends your words to [vendor name] and [any named service]
for processing, and we delete the audio [when/after __].
You can say no and I will write the notes myself. Please tick
one:  [ ] Yes   [ ] No   [ ] I would like to talk about it first"

At the start of the appointment (face to face, clearly):
"I would like to use an AI note-taker today. It listens and
types a draft. Would you be happy with that? You can say no.
Could you tell me in your own words what you are agreeing to?"

The teach-back at the end, asking the patient to say in their own words what they are agreeing to, is the best safeguard for patients who may have misheard. Note the method in the record: "written consent on intake form, confirmed face to face with teach-back". If a relative is present, ask them separately, because their voice will be on the recording.

If the patient says no, switch to your normal note routine without comment. Heidi, for example, lists a standard template library on its free plan, and Jane describes a way to use its scribe without recording by dictating a summary afterwards or by typing a note and using its refine tool. Check that your tool has a no-recording route and use it.

Assessments, tinnitus visits and wax removal need their own templates

A fitting is only one appointment type. HearScribe's page lists earwax removal, hearing assessments, tinnitus consultations, fittings and follow-ups, and a general scribe can cover the same ground if you give it separate templates. Do not use one template for all of them: a tinnitus note should record onset, pitch and sleep impact, while a wax visit should record the otoscopy findings and the technique used. Mixed templates produce notes that look complete and miss the point.

Template idea: hearing assessment (fill the bracketed lines yourself)
Reason for visit: [in the patient's words]
History: onset, one ear or both, noise exposure, family history,
  dizziness, discharge, pain. Use only what was said.
Otoscopy: [typed by you]
Tests performed: [typed by you, with thresholds from the audiometer]
Findings in plain words: what was explained to the patient.
Outcome: hearing aids discussed / referral / reassurance.
Safety netting: red-flag symptoms mentioned and what to do.
Follow-up: [date or "none needed"]

A realistic mistake to guard against: in a tinnitus consultation the patient says the ringing is "like a kettle in the left ear, started after the flight". A careless draft writes "bilateral tonal tinnitus, sudden onset after air travel", which sounds more clinical and changes what a reader would do. You catch it by comparing the draft with your own memory of the conversation, and by keeping a rule in the template instructions: "Do not rephrase the patient's description into diagnostic terms."

Another realistic mistake: in a wax removal visit the draft says "irrigation performed" when you used microsuction. The procedure line should be a field you pick from a list, not text the scribe composes. Where your tool allows structured fields, use them for procedures, devices and settings, and let the scribe write only the conversation sections.

Where the data goes, and who gets the question

Data location is a question with three layers: where the audio goes, where the transcript goes, and where the finished note lives. Ask for each separately. The answers differ a lot even among the four tools above.

  • HearScribe: patient content is not stored by the vendor, but passes transiently to Google's Gemini API and, for live transcription, possibly to the browser maker's speech service. Your own records live in your own system.
  • Heidi: its HIPAA page says data is hosted in the US, and its UK compliance page says UK data is hosted in the UK. Pick the regional site that matches your clinic and confirm the hosting statement for that region. The pages I read did not say how long audio is kept, so ask.
  • Jane: its UK guide says UK clinics' recordings and drafts are stored in the UK, and that data is not used to train AI models. Deletion is set clinic-wide: keep indefinitely, delete right after the draft, or delete after a set number of days.
  • Tandem: its site says all patient data is processed and stored in Europe and that audio is deleted after transcription.

Write whatever you choose into the consent wording, and test the setting yourself. For a deeper look at questions to put to a vendor, see the audio, retention and BAA vendor questions.

Free or paid: a break-even sum for a one-room clinic

Say a solo audiologist sees 12 patients a day, four days a week: 48 appointments a week, or roughly 208 a month. If the scribe saves three minutes of typing per appointment, that is 144 minutes a week, 2.4 hours. Now the caps. HearScribe's Starter plan covers 100 consultations a month and Professional covers 200, so 208 a month only fits the Practice plan at 100 pounds for 500, about 23 pounds a week. Set against 2.4 hours of your time, that still pays if the notes need only light edits. The other route is to scribe only the fittings and counselling-heavy follow-ups, say 30 a week or about 130 a month, which Professional at 50 pounds covers, and keep typing the short wax and hearing-check visits. (These figures are illustrative.)

Caps work differently in each tool, so count your appointments before you choose. Jane's free tier covers five notes a month per practitioner, which is a trial, not a routine. Heidi's free plan lists unlimited AI documentation, so the trade-off there is templates and features rather than note numbers. HearScribe counts consultations, so a busy month or a second clinician can push you up a tier, and the Starter plan's two-person limit matters if a colleague covers your list.

Where a scribe slips in hearing-care notes

From the pattern of what speech tools get wrong, expect these slips and look for them in review. They are the issues I would put on your checklist.

  • Left and right swapped. Check every ear-specific statement. A patient who says "my right is worse" may be recorded as left if the scribe mishears.
  • Numbers with units. "Eight kilohertz" and "eighty" are close in speech. Cross-check against the audiogram.
  • Device names. Model names, receiver lengths and dome sizes get mangled. Keep a list of product names in your template instructions.
  • Medical history added from nowhere. A draft that mentions tinnitus, dizziness or earwax that was never discussed. Delete it unless you can confirm it from the conversation.
  • Counselling advice you did not give. Drafts can pad out the advice section with generic tips.
  • Safety referral points lost. If you mentioned sudden hearing loss or one-sided symptoms and your referral plan, make sure they appear in the plan and are not softened.

Teaching the scribe your clinic's vocabulary

Most scribe tools let you save instructions with a template. Use them for a short glossary, because speech engines guess at hearing-care jargon. Here is a filled-in instruction block for a small private clinic (the product names are placeholders, so list the ones you really fit).

Template instructions:
- Terms to spell exactly: real-ear measurement, receiver-in-canal,
  open dome, closed dome, tinnitus masker, [brand A], [brand B].
- Write frequencies as "4 kHz" and levels as "40 dB HL".
- Always state which ear: write "left" or "right", never "the ear".
- If a number was not spoken, write [not stated]. Never estimate.
- Put patient quotes in quotation marks and keep them short.
- Do not add history, diagnoses or advice that was not spoken.

After a week, read ten drafts and add every term it still gets wrong to the list. Clinics usually find a handful of repeat offenders, such as a brand name or an acronym, and fixing those cuts editing time more than any other change. Keep the glossary in one shared document so a second audiologist can paste it into their own template.

Relatives complicate drafts too. If a daughter does most of the talking, the draft may attribute her descriptions to the patient. Add a line asking the scribe to label who said what ("patient reports", "daughter reports"), and check it. A fitting plan built on a relative's view of the loss, written up as the patient's own report, would mislead the next clinician who reads the record.

One more habit pays off: sign nothing until you have read the draft against the audiogram open on screen. It takes about a minute for a fitting. A draft that matches the audiogram on every ear and every frequency, and contains nothing you did not say, is safe to sign. A draft that fails either test goes back to you for editing, and the failure goes on your trial scorecard.

A two-week test that tells you which tool to keep

Choose two tools at most. For ten fittings or follow-ups, write your note the usual way and also generate the scribe draft. Do not copy from one to the other, and score each draft on four things: numbers correct, devices correct, counselling captured, and invented content. Record minutes to review the draft against minutes to write your own. At the end of two weeks, the decision is usually obvious. If both drafts need heavy rewriting, the tool is not ready for your clinic, and that is a result, not a failure. If you want a structure for trials, the buyer's checklist for another allied-health role is a good model, and the small-clinic cost comparison helps with the money side.

Where state recording laws apply, the one-party and all-party map in the consent form tutorial applies to hearing clinics too. A clinician who wants help planning a trial can book an AI implementation consultation on a 1:1 call.

Audiology scribe questions

Can a general scribe read an audiogram?

Not reliably. A general scribe transcribes speech and drafts a note, so audiogram thresholds still need to be typed or pasted in by you. HearScribe says on its site that it interprets uploaded audiograms, so if that matters, test it with ten of your own audiograms and compare every threshold.

Do I need consent if the scribe only listens to me?

If you dictate after the patient leaves, the patient's voice is not captured, but data-protection duties still apply. If it listens during the appointment, ask the patient first. A vendor claiming no consent is needed does not change your own duty to be open with patients.

How should I get consent from a patient who cannot hear well?

Do not rely on a spoken line alone. Give a large-print form before the appointment, repeat the question face to face with your mask off if safe, check they understood by asking them to say it back, and note the method used.

Is a free scribe good enough for a one-person clinic?

It can be, if the free tier covers your volume and the data terms suit you. Heidi lists a free plan with unlimited AI documentation, and Jane gives five free notes a month, so test whether the output needs less editing than your current notes.

How long should I keep recordings?

Keep them as briefly as your policy allows. Many clinics choose deletion as soon as the draft note exists, because the approved note is the record. Check the tool's setting, and write the choice into your consent wording.

Further reads

Sources: HearScribe site, pricing page and privacy policy; Heidi pricing, HIPAA and UK compliance pages; Jane App pricing, AI Scribe pages and help guides; Tandem Health site and pricing page. Prices and claims as read on 4 November 2026.

Want a scribe trial that fits a fitting clinic?

On a 1:1 call we can map your appointment types, choose which one to test first, and set up templates and consent steps for patients who are hard of hearing. No promises about results, just a clear plan.

Book a 1:1 call with me