Tell patients in three layers: a one-line notice in the booking confirmation, a plain question at the start of each consultation, and a short written page with the details. Every version should say what's recorded, why, what happens to the audio, who checks the note, and that saying no changes nothing about their care. Then record their answer yourself.
Two cautions before you copy any wording. First, notice and consent are different things: a poster tells people, a question asks them. Recording laws differ from place to place, and some require every person in a conversation to agree before it's recorded, so the question in the room matters more than the poster. Second, the wording below is a starting point to check with your indemnity provider or a healthcare lawyer, not legal advice. What it does is make sure every patient hears the same true information in words they understand.
The seven things every version has to say
Whether it's a one-line text or a full page, the wording should answer these, in more or less detail depending on space:
- That you use it. Name it plainly: "an AI tool that helps write your notes".
- What it does. Listens to the consultation and drafts a note; it doesn't diagnose or decide anything.
- What happens to the audio and transcript. Kept or not, for how long, deleted when.
- Who checks the note. The clinician reads and corrects it before it goes into the record.
- Who else can see it. Whether the vendor can, and that it isn't used to train AI, if your contract says so.
- That they can say no, or stop it at any point, with no effect on their care.
- Who to ask for more information.
Points 3 and 5 must come from your vendor's documentation and your own settings, not from the vendor's marketing. If you haven't written those facts down yet, rolling out an AI scribe without losing patient trust shows how to build a one-page fact sheet first.
Booking confirmation and reminder wording
This is the notice layer. Its only job is to make sure the question in the room isn't a surprise. Keep it short enough that people read it.
TEXT (fits one message)
Some of our clinicians use an AI tool to help write notes.
They'll ask you first and you can say no. More: [link]
EMAIL (one paragraph)
About your notes: some of our clinicians use an AI
note-taking tool. With your agreement, it listens during
your appointment and drafts a note, which your clinician
checks and corrects before it goes into your record. The
audio isn't kept. It's entirely your choice. Your clinician
will ask at the start, and saying no won't affect your
care in any way. You can read more here: [link]Put the line near the top of the confirmation, not below the parking directions. If your confirmations are sent by an automated system, check that the line appears on every appointment type the scribe will be used for, including video consultations.
Replace "the audio isn't kept" with whatever your fact sheet says. If your tool keeps a transcript for seven days, say so: "The audio isn't kept, and the written transcript is deleted after seven days." Precision here is what makes the rest of the wording believable.
Here is how the fact sheet turns into wording for one illustrative set-up. A three-physiotherapist clinic uses a scribe whose documentation says audio is discarded once the transcript is made, transcripts are kept for 14 days by default, and customer data isn't used for model training under its business terms. (Nabla, for instance, says it discards audio and keeps transcripts 14 days by default; check your own vendor's current documentation rather than relying on anyone's summary.) The clinic's email line becomes: "With your agreement, it listens during your appointment and drafts a note, which your physiotherapist checks before it goes into your record. The recording itself isn't kept; a written transcript is kept for 14 days so the note can be checked, then deleted." Every clause maps to a line in the vendor's documentation or the clinic's settings, and nothing in it is a claim the clinic can't show.
What to say at the start of the consultation
This is the consent layer, and it should happen before the patient starts telling you why they've come. Ask at every consultation, not only the first. It takes ten seconds, patients' feelings change with the subject of the visit, and "they agreed last time" is weak ground if a patient later says they weren't asked.
STANDARD
"Before we start: I use an AI tool that listens and drafts
my notes, so I can concentrate on you rather than typing.
I check everything it writes, and the audio isn't kept.
Are you happy for me to use it today? It's completely
fine to say no."
RETURNING PATIENT
"Are you still happy for me to use the note-taking tool
today? Just say if you'd rather I didn't."
VIDEO OR PHONE CONSULTATION
"Can I check before we start: I use an AI tool that helps
write my notes from our conversation. Is that OK today?
If you'd rather I didn't, I'll type as we go."
A RELATIVE, CARER OR INTERPRETER PRESENT
"I use an AI tool that drafts my notes from what we say.
Because it would hear everyone in the room, I need you
both to be happy with it. Is that all right with you both?"
A PARENT AND CHILD
To the parent: "Are you happy for me to use my note-taking
tool today?" Then to the child, in plain words: "This
helps me remember what we talk about. Is that OK with you?"A few notes on delivery. Say "it's fine to say no" and mean it; the tone matters more than the words. Start the tool only after the answer, never "I've already got it running, is that all right?" For the relative-or-carer version, anyone who objects is enough for a no. And if a patient hesitates, treat hesitation as no and type your notes; you can always offer again next time.
The difference between a good and a bad delivery is easy to hear. A rushed version, from a clinician running 20 minutes late: "I'm just going to pop this on, it helps with notes, that's OK, isn't it?", said while tapping start on the phone. It covers none of the seven points except the first, it's phrased to get a yes, and the tool is already running before the answer. The same clinician, on time, says the standard script in about ten seconds and waits for the reply. Running late is when the short version creeps in, so it's worth agreeing as a team that a late clinic is a reason to skip the tool, not to skip the question.
Consent can also change shape during the appointment. Say a patient agrees at the start of a phone consultation, then halfway through says "hang on, my husband's here, I'll put you on speaker". Someone new is now being recorded who hasn't been asked. The simplest rule: pause the tool, ask the new person the relative-or-carer version, and only restart if both agree. Note it in the record, as in "AI scribe: paused when patient's husband joined the call, both agreed, restarted".
The written information page
This is the detail layer: the page your booking link points to, printed copies at reception, and the answer to questions staff can't. An adaptable template:
HOW WE USE AI TO HELP WRITE YOUR NOTES
What it is
Some of our clinicians use [tool name], an AI tool that
listens to your appointment and writes a draft of the
clinical note. It helps them concentrate on you instead of
typing.
What it doesn't do
It doesn't diagnose, suggest treatment or make decisions.
Your clinician does all of that, as always.
Your choice
We'll ask before every appointment. You can say no, or ask
us to pause or stop it at any time, and it won't affect your
care. You don't have to give a reason.
What happens to the recording
[Fact sheet wording, e.g. "The audio is processed during
your appointment and not stored. A written transcript is
kept for [7] days so your clinician can check the note,
then deleted."]
Who checks the note
Your clinician reads and corrects every note before it's
added to your record.
Who else can see it
[e.g. "Nobody at [vendor] can see your information unless
we ask them to for technical support. Our contract says
your information isn't used to train AI."]
Questions or changing your mind
Ask your clinician or contact [role, e.g. practice manager]
at [contact route]. If you'd like us to stop using it for
you in future, tell us and we'll note it on your record.
Last checked: [date]Keep the "last checked" date. It reminds you the page is only true for the settings on that date, and it tells patients the page is looked after. Check the reading level too: short sentences, no "ambient", no "LLM", no "processing" without saying what that means.
Wording that backfires, and what to say instead
| Instead of | Why it backfires | Say |
|---|---|---|
| "By attending your appointment you consent to AI recording." | Attendance isn't agreement; it removes the choice | "We'll ask you at the start, and you can say no." |
| "Our AI system is 100% secure." | Nobody can promise that, and patients know it | "The audio isn't kept, and your information is protected by our contract with the supplier." |
| "We may use AI tools to improve our services." | Vague enough to mean anything, so patients assume the worst | "Some clinicians use an AI tool to help write notes." |
| "We use ambient clinical documentation." | Jargon hides what's happening | "It listens to the appointment and drafts a note." |
| "The AI writes your notes." | Suggests nobody checks them | "It drafts the note; your clinician checks and corrects it." |
| "Recording helps us, so we'd appreciate it." | Puts pressure on the patient | "It's entirely your choice." |
Recording the answer: never let the template do it
The patient's answer belongs in the notes, entered by the clinician, every time. Many scribes let you build note templates, and it's tempting to add a line such as "Patient consented to AI-assisted documentation" to the template so it appears automatically. Don't. A line that appears whether or not anyone asked isn't a record of consent; it's a claim that consent happened.
This is not a hypothetical risk. A proposed class action filed against a large health system in late 2025 alleged that patients' consultations were recorded by an ambient scribe without them being asked, and that the patient behind the claim later found a line in his record saying he had been advised of the recording and had consented, which he says never happened. Whatever the outcome of that case, the lesson for a small practice is simple: the consent line should be typed or ticked by the clinician after the patient answers.
A monthly count is the easiest way to check the habit is holding. Most scribes show how many sessions each clinician ran; compare that with the number of consultations whose notes carry a consent line. In an illustrative month at a four-clinician practice, the scribe shows 412 sessions and the records show 398 consent lines, so 14 sessions have no recorded answer. Trace those 14: in this example, nine were one locum who hadn't been told about the rule, four were notes finished in a hurry where the line was forgotten, and one was a session started by mistake and stopped within seconds. The fixes are different for each, which is why the count is only the start. The locum gets a briefing, the rushed notes get the line added with a date and an explanation, and the accidental session is deleted through the vendor's process and noted.
Examples of a clinician-entered line:
"AI scribe: explained, patient agreed. [initials]"
"AI scribe: patient declined, notes typed. [initials]"
"AI scribe: patient's daughter present, both agreed."
"AI scribe: paused at patient's request 10:42-10:47."When the answer is no, or changes halfway
The response to a no should be short and warm, with no hint of disappointment: "No problem at all, I'll type as we go." Don't ask why. If a patient says no several times, add a note so colleagues stop asking at every visit; some patients find being asked repeatedly more irritating than the tool itself.
If a patient asks to stop partway through: "Of course, I've stopped it now." Stop the tool, then follow your vendor's deletion steps for that session and finish the consultation by typing. Afterwards, note what happened. If they seem worried about what was captured before they asked, tell them what your fact sheet says happens to it and offer the information page.
For patients who ask detailed questions you can't answer in the room, have a single named person who can, usually the practice manager, and say so: "That's a good question. [Name] looks after how we use the tool and can tell you exactly." Patients accept "I'll find out" far more readily than a guess.
Reception gets the same questions again and again, so give the front desk a short card with agreed answers. A filled-in version for the physiotherapy clinic above:
- "Does the company hear my appointment?" "Their system processes the audio to write the draft, but nobody at the company listens to it, and the recording isn't kept. The written transcript is deleted after 14 days."
- "Will it be used to train AI?" "No. Our contract with the supplier says your information isn't used for that."
- "Can I say no for all my future appointments?" "Yes. I'll note it on your record now, and your physio will see it, so you won't be asked again."
- "What if the note's wrong?" "Your physio checks and corrects every note before it's saved. If you ever see something wrong in your record, tell us and we'll correct it the usual way."
Each answer has to be true for your set-up on the day it's said, so the card carries the same "last checked" date as the information page and is reprinted when that page changes.
Situations to decide before they happen
Scripts cover the ordinary appointment. A handful of situations come up often enough that the practice should agree an answer in advance, rather than leaving each clinician to improvise.
Patients who may not be able to agree. Someone with advanced dementia, or who is very unwell, may not be able to weigh up the question. Most practices simply don't use the scribe in these consultations unless their adviser has agreed a process involving whoever lawfully speaks for the patient. "Not today" is always an acceptable default.
Patients with hearing, sight or language needs. A patient who can't read the booking text or the information page hasn't really had the notice layer. Explain it in the room with more care, use the interpreter if one is present, and offer the information page in large print or another language if you have it. If you're unsure they've understood, don't use the tool.
The patient who asks what the note says. Some patients like the idea and want to see the result. If your system lets patients view their records, tell them the note will appear there once you've checked it; if not, follow your usual process for access requests. Never hand over the unchecked draft.
A sensitive turn mid-consultation. A routine visit can become a disclosure about abuse, self-harm or a safeguarding concern. Agree that clinicians pause the tool as soon as a conversation moves that way, say they've paused it, and write that part of the note by hand. What you'd say is simply: "I'm going to switch the note tool off for this part, so it's just us."
Keeping the words true as the tool changes
The most likely way your wording becomes wrong isn't a mistake when you write it; it's the vendor changing something later. One widely used therapy platform began keeping de-identified transcripts by default for clinicians who switched its note tool on from June 2026, so practices that had copied older template wording about transcripts being deleted were suddenly describing a different set-up. Review your wording whenever any of these happen:
- The vendor updates its privacy terms, retention defaults or data-processing agreement.
- You change settings, such as transcript retention.
- You add a feature, such as sending patients an AI-written summary after the visit.
- You switch vendor, or the vendor is bought by another company.
- Six months pass without any of the above.
Your own settings can make the wording wrong just as easily. Imagine a practice whose notice says "the audio isn't kept". One clinician, wanting to replay a tricky consultation, switches on an option in the scribe that saves audio for playback, not realising it applies to every session from then on. Nobody else knows. It comes to light three months later when a patient asks for a copy of everything the practice holds about them, and the practice finds 11 weeks of saved recordings it had told patients didn't exist. The immediate steps are to switch the option off, delete the saved audio through the vendor's process, and ask the practice's data-protection adviser whether anything more is needed. The lasting fix is that settings affecting retention are changed only by the practice manager, and the fact sheet is rechecked against the live settings each quarter, not just against the vendor's documents.
Therapists face the same questions with extra sensitivity; should therapists tell clients they use AI and the confidentiality checklist for therapists' note tools cover that setting. Lawyers deal with a close parallel in using AI note takers in client meetings, and whether AI meeting note-takers are safe covers the general risks of recording tools in any business.
Further reads
- Is an AI Scribe Worth It for a Small Private Clinic? — Whether a scribe earns its keep before you write any wording.
- What to Check in an AI Vendor's Data Processing Agreement — Where the facts in your wording should come from in the contract.
- What to Check in an AI Tool's Privacy Policy and Terms — Reading a vendor's privacy terms so your promises stay true.
- Do You Need a DPIA Before Using AI Tools? — Whether you need a formal impact assessment before going live.
- How a Small Clinic Can Use AI at the Front Desk Safely — Using AI at the front desk safely, including patient notices.
- How to Write an AI Usage Policy for Your Small Business — The internal policy your patient wording should match.
- What to Ask Before Buying Any AI Tool for a Medical Practice — A grouped checklist of questions for any AI vendor selling to a medical practice, what a good answer looks like, a scored example and the red-flag answers.
- AI Tools for Physiotherapy Clinics: What Each One Actually Does — What each AI tool a physio clinic might buy actually does, what it doesn't, and which two to start with.
- What Is an AI Scribe and How Does It Work in a Consultation? — The consultation stage by stage: capture, transcript, speaker separation, note drafting and sign-off, with the errors each stage produces and how to spot them.
- Dental Practice AI Mistakes: Consent, Data, and Over-Automation — Ten AI mistakes dental practices make with consent, patient data and over-automation, with examples of how each shows up and the fix.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: reporting on a proposed class action over ambient scribe consent filed in late 2025 (Medscape, Becker's Hospital Review); SimplePractice support documentation on Note Taker transcript retention (checked September 2026).