Skip to main content

Responsible Use

  • Educators
  • Students
  • Practice supervisors

AI scribes on placement

More and more NHS services use ambient voice technology (AVT), often called "AI scribes". The tool listens to a consultation and drafts a note, letter or summary for the clinician to check. Students are likely to see these tools on placement before they meet them in university teaching. This page explains the rules and how to prepare students.

What an AI scribe does​

  1. It records the conversation between the clinician and the patient, with the patient's knowledge.
  2. It turns the audio into a transcript.
  3. It uses a large language model to draft a structured output, such as a clinical note, referral letter or suggested clinical codes.
  4. A clinician reviews, edits and approves the output before it goes into the patient's record or anything is done with it.

The fourth step is the one that matters most. NHS England is clear that clinicians must "review and revise the outputs" before they are used (NHS England).

The rules in England​

NHS England guidance (updated 29 July 2026)​

NHS England's guidance on AI-enabled ambient scribing products sets out what organisations must have in place before staff use these tools:

  • Clinical safety: a named Clinical Safety Officer and clinical risk management under the NHS standards DCB0160 (for the organisation deploying the tool) and DCB0129 (for the supplier).
  • Data protection: a Data Protection Impact Assessment, Data Security and Protection Toolkit compliance, and clear decisions on how long audio, transcripts and outputs are kept. The guidance doesn't set a retention period; each organisation must decide one.
  • Transparency: patients are told how their information is used, through privacy notices, signage and other information, and are given the chance to object.
  • Training: staff are trained in how to use the tool, including how microphone position affects accuracy.
  • Review: clinicians review and approve outputs before any further action, and notes are saved to the correct clinical record.

MHRA: when an AI scribe counts as a medical device (29 July 2026)​

The MHRA has clarified the regulatory status of AVT:

Not regulated as a medical deviceRegulated as a medical device
Transcribing or summarising conversations, drafting letters, or suggesting clinical codes for a clinician to reviewSupporting diagnosis, treatment or prevention, or taking automated actions (such as placing orders) without clinician review

Whether or not the tool is a medical device, the clinician remains responsible for checking the output.

Mental health, learning disability and neurodivergent people​

The Royal College of Psychiatrists and NHS England are developing specific guidance on AVT for people with mental health needs, learning disabilities and neurodivergent people, recognising issues of capacity, privacy and trust (RCPsych, May 2026). It was due by September 2026 and had not been published when this page was checked. Check for it before teaching this topic.

Known risks​

NHS England's guidance lists the kinds of errors AI scribes can make:

  • Missing critical information.
  • Incorrect information or context.
  • Misreading complex clinical terms and abbreviations.
  • Misinterpreting what was said, for example recording a patient's hypothetical statement ("what if it's cancer?") as a confirmed diagnosis.
  • Delayed outputs.

These are exactly the errors a tired clinician can miss when approving a note that "looks right". That is why the review step is a professional responsibility, not a formality.

What students should and shouldn't do​

Local policy always comes first. Ask your practice supervisor how AI scribes are used in your placement area.

Students should:

  • Know when an AI scribe is in use, and be ready to explain it to patients in plain language if asked.
  • Watch how their supervisor checks and corrects AI-drafted notes, and ask about the changes they make.
  • Speak up if they notice an error in an AI-drafted note, just as they would for any other record.
  • Respect a patient's choice not to be recorded.

Students shouldn't:

  • Use an AI scribe, or approve its output, unless local policy allows it and they are supervised.
  • Record patients on personal devices or put patient conversations into personal AI tools.
  • Assume an AI-drafted note is accurate because it is well written.

Teaching ideas​

  • Spot the error. Write a short fictional consultation, then an "AI-drafted" note containing three planted errors (an omission, a misheard drug name, a hypothetical recorded as fact). Students find and correct them, then discuss which could cause harm.
  • Explain it to the patient. In pairs, students practise explaining an AI scribe to a patient in under a minute, including the patient's right to say no.
  • Accountability debate. "If the AI scribe gets it wrong and the nurse approves the note, who is responsible?" Use NMC Code standard 10 (keep clear and accurate records) and the National Commission's recommendations on shared responsibility.
  • Reflection for placement. After placement, students reflect on one occasion they saw an AI tool used: what went well, what the risks were, and what they would do differently.

Related: NMC Code review and AI · Cost to the individual · Responsible use checklist