AIWF All articles
Practical Guides

Who’s Taking Notes in Your GP Appointment? Making Sense of AI Scribes

AIWF
Who’s Taking Notes in Your GP Appointment? Making Sense of AI Scribes

Your GP is looking at you, not the keyboard. On the desk, a phone or a small mic is quietly listening. In some surgeries in 2026, that device is an AI scribe: software that turns the conversation into a draft set of clinical notes. It can save a clinician time and help you feel heard. It also raises a very reasonable question: what exactly is being recorded, where does it go, and who checks it before it lands in your record?

This is not a reason to panic, nor a reason to assume the tech is automatically fine. It is a reason to ask a few clear questions.

What an AI scribe actually does

An ambient AI scribe is not a chatbot giving medical advice. It listens to the consultation — usually with your knowledge — and produces a transcript or summary. The clinician then reviews, edits and signs it off. In theory, that means less typing and more eye contact. In practice, the quality depends on the product, the room, the accents in the room and the clinician’s review. AI can mishear, miss a caveat or tidy up a messy conversation into something that sounds more certain than it was.

Some tools record audio and send it to a cloud service for processing. Others may process on a device or within a local system. Some delete audio quickly; others keep it for a set period. The details matter, and they vary by supplier and practice.

The consent question is not just a tick box

Under UK GDPR and the Data Protection Act 2018, a GP practice must be transparent about how it uses patient data. For direct care, the lawful basis is often public task or provision of health or social care, rather than consent in the everyday sense. But that does not mean AI scribes can be slipped in unnoticed. Patients should be told what is happening, why, and what their options are. If you object, the practice should be able to offer a reasonable alternative — typically the clinician taking notes in the usual way.

Consent can still matter in how a practice approaches the conversation. Being told ‘we use an AI scribe, is that okay?’ is different from discovering a mic after the fact. The Information Commissioner’s Office expects organisations to be clear and to carry out a data protection impact assessment for higher-risk processing. A practice rolling out ambient recording should be able to explain its DPIA, its retention rules and its supplier arrangements without disappearing into jargon.

Where the words go

Ask whether the audio leaves the building. If it does, who processes it, where are their servers, and is there a contract in place? The practice remains the data controller for your health record; the AI supplier is usually a processor acting on the practice’s instructions. That distinction matters if something goes wrong. It also matters for international transfers: if audio or transcripts are processed outside the UK, there should be safeguards and a clear explanation.

Retention is the other practical question. Is the audio kept for 30 days, 30 minutes or not at all? Are transcripts stored separately from your clinical record? Can staff search them? The safest answer is not always ‘deleted instantly’ — sometimes a short retention period helps with safety checks and incident review — but you are entitled to know the policy.

Who checks the notes?

The clinician should. An AI scribe produces a draft, not a final record. The human review step is the safety net. If a summary says ‘patient denies chest pain’ when you said something more nuanced, the clinician is supposed to catch it. That is why the review should be meaningful, not a rushed click at the end of a ten-minute appointment.

If you want to see what was written, you can ask for access to your GP record. In England, the NHS App gives many patients access to parts of their record, though not every entry appears immediately. You can also make a subject access request to the practice. If something is wrong, ask for it to be corrected. You do not need to be a data protection expert to say: that is not what I said.

Questions to ask before the mic goes on

  1. Is an AI scribe being used today, and what does it record?
  2. Is the audio stored, and for how long?
  3. Who reviews the draft notes before they are saved?
  4. Can I see the note afterwards, and how do I ask for a correction?
  5. What happens if I say no?

These are not awkward questions. They are the same kind of questions you might ask about a trainee sitting in, or a chaperone. A good practice will have answers ready. A vague answer is useful information too.

If you would rather not

You can ask for the AI scribe to be switched off. The clinician may need to type notes instead, which could affect the pace of the appointment, but that is a practical consequence, not a reason to ignore your objection. If you are worried about a sensitive consultation — domestic abuse, sexual health, a safeguarding concern — say so at the start. Ask that the mic be off before you begin.

For practice managers, the checklist is familiar: know your supplier, know your DPIA, train staff on review, tell patients clearly, and have a fallback. The technology may be new; the data protection principles are not.

The AI scribe is best understood as a junior assistant with very fast typing and no clinical judgement. It can be useful. It can also be wrong. The human in the room still owns the record — and you still have rights over it.

All Articles

Related Articles

A Familiar Voice, a Strange Request: The Family Password Plan for UK Voice-Cloning Scams

A Familiar Voice, a Strange Request: The Family Password Plan for UK Voice-Cloning Scams

Press 1 for Nowhere: How to Get a Real Human When AI Is Guarding the Gates

Press 1 for Nowhere: How to Get a Real Human When AI Is Guarding the Gates

The Receipt Shoebox Goes Digital: Is AI Bookkeeping Ready for MTD?

The Receipt Shoebox Goes Digital: Is AI Bookkeeping Ready for MTD?