Guide
Getting informed client consent for AI note-taking: a practical script
How to explain AI-assisted clinical notes to clients, get genuine informed consent, and handle common questions. Includes a ready-to-use script.
Informed consent for AI note-taking means telling clients, in plain language, what the technology does, who processes their words, how data is stored, and what their options are — then documenting that they understood and agreed. You need this before using any AI scribe tool, and it should sit alongside your existing privacy notice and therapy contract, not replace them.
Why this matters beyond a tick-box
Consent under UK GDPR must be freely given, specific, informed, and unambiguous. For therapeutic work the bar is higher still: a client who felt pressured to agree, or who didn't really understand what they were agreeing to, hasn't given valid consent. Discovering mid-treatment that their words were processed by a third-party system — without their knowledge — can cause real harm to the alliance.
The ICO is clear that organisations must be transparent about automated processing. What counts as adequate transparency depends on the specific tool, the data involved, and your lawful basis — so check current ICO guidance and your professional body's position rather than relying on a fixed checklist. BACP, UKCP, HCPC, BABCP, and NCS have each issued or are developing guidance on AI use; search your body's website for the latest statement before introducing any AI tool.
What clients actually need to know
Strip out the jargon and clients need answers to four questions.
What happens to my voice? Does the audio get recorded and sent somewhere, or does it stay on your device? These are very different things, and clients often assume the worst.
Who reads the transcript? Is it a human transcriptionist, an AI model, or a combination? Where are those servers?
What gets stored, and for how long? Is the transcript kept permanently, or discarded after the note is drafted?
Can I say no, and what happens if I do? Consent must be genuinely optional. If a client declines, you need a clear answer ready — typically that you'll take handwritten or typed notes as you always did.
The answers vary by tool, so read your vendor's data-processing documentation carefully before you can answer honestly. Look for an ICO registration number, a data-processing agreement (DPA) you can sign, and a clear statement about whether client data is used to train AI models.
A practical consent script
The following is a starting point. Adapt it to your own words — it should sound like you.
"Before we start, I want to explain how I write my session notes. I use a tool that helps me draft clinical notes after our sessions.
When we're talking, the audio is transcribed on my device — it doesn't leave my computer or get sent anywhere. After the session, the transcript is processed briefly to generate a draft note, then discarded. The draft note stays in my secure records system, the same as any other note I'd write.
The company that makes the tool is registered with the UK Information Commissioner's Office, stores data on servers in the EU, and doesn't use your information to train AI systems or share it with third parties.
Your name and identifying details don't need to appear in the transcript — I can refer to you by initials if you'd prefer.
You don't have to agree to this. If you'd rather I take notes by hand or type them myself, I'm happy to do that — it won't affect your therapy in any way. Do you have any questions before you decide?"
Then document the conversation. Note the date, what you explained, any questions they asked, and whether they consented or declined. A signature on a written form is cleaner than a verbal note, but at minimum record it in your case notes.
Handling the questions clients actually ask
"Is it like Alexa listening all the time?" No — transcription only runs when you actively start it for that session. Nothing is recorded passively.
"Could someone read what I said?" Explain exactly what your tool does: whether a human ever sees the transcript, or whether it's processed automatically and then deleted. Be specific rather than vaguely reassuring.
"What if I change my mind?" They can withdraw consent at any time. Explain what that means practically — you'd revert to manual notes going forward, and any previously generated notes remain part of the clinical record as they would with any other note.
"Is this even legal?" Yes, when done properly. UK GDPR permits processing of special-category data (which health information is) under specific conditions, including explicit consent and legitimate interests with appropriate safeguards. The exact lawful basis you rely on should be stated in your privacy notice. If you're unsure, that's a question for a data-protection solicitor or your professional body's ethics line.
One honest limitation to name
No consent process removes all risk. Transcripts, however briefly they exist, contain sensitive material. If a client discloses something highly sensitive — a safeguarding concern, a trauma detail they've never told anyone — that content passes through a third-party system, even if deleted afterwards. Some clients, once they understand this, will decline. That's a legitimate choice, and your consent process should make space for it without pressure. For clients who work in sensitive professions (intelligence, legal, healthcare) or who have particular reasons to be cautious about data, a more detailed conversation is warranted.
Where Sorca fits
Sorca's scribe transcribes audio entirely within your browser — the audio never leaves your device and is never stored. The transcript is processed in memory to generate a draft note in your chosen format (SOAP, DAP, BIRP, and others) and then discarded; nothing enters a client record until you explicitly save it. Client data is never used to train AI models. If you'd like to see how the AI clinical scribe works in practice, or how the client companion app collects outcomes between sessions, the free trial runs for three days with no card required.
Frequently asked questions
Do I legally need written consent before using an AI note-taking tool with therapy clients?
UK GDPR requires consent for processing special-category data to be explicit, which strongly points toward a written record. Your professional body may have additional requirements — check their current guidance, as positions on AI are evolving quickly. At minimum, document the conversation and the client's decision in your case notes.
What should I include in my privacy notice about AI-assisted notes?
Your privacy notice should name the tool or category of tool, explain what data is processed and where, state how long it's retained, and confirm whether it's used to train AI models. It should also explain the client's right to withdraw consent and what happens if they do. Review the ICO's guidance on transparency for the current standard.
Can a client refuse AI note-taking and still receive therapy?
Yes — consent must be freely given, which means refusal can't carry a penalty. If a client declines, revert to your previous note-taking method. Make this clear upfront so the client doesn't feel the choice affects their care.
Does the AI tool need to be named in the consent form, or is a general description enough?
The ICO's transparency principle favours specificity: clients should have enough information to understand what they're agreeing to. Naming the tool and linking to its privacy documentation is better practice than a vague reference to 'AI software'. If the tool changes, update your consent documentation accordingly.
Take the admin off your week
Sorca drafts the note while you stay present — audio never stored, nothing saved without your say-so. Three-day free trial, no card needed.
Start free — no card needed