The Bottom Line
- AI scribes listen to consultations and generate clinical notes automatically — potentially saving <strong>30–60 minutes per day</strong>.
- The market has matured: key differentiators are now <strong>EHR integration, governance compliance (DTAC/MHRA), and note quality</strong> — not just transcription accuracy.
- No scribe is 'set and forget' — <strong>every AI-generated note must be reviewed and approved</strong> by the clinician.
AI medical scribes are the fastest-adopted AI tool in UK primary care. The promise is compelling: the scribe listens to your consultation, generates a structured clinical note, and inserts it into your EHR — eliminating most of the documentation burden that consumes 30–50% of a GP's day. The reality in 2026 is that several products have reached usable maturity, but they differ significantly in governance posture, integration depth, and clinical workflow fit.
1
Heidi Health (+ Heidi Evidence)
Heidi is one of the most widely used AI scribes in UK primary care. It transcribes consultations, generates SOAP-style or free-text notes, and integrates with EMIS and SystmOne. The 'Evidence' add-on provides clinical decision support by surfacing relevant guidelines during or after consultations. Strengths: mature UK market presence, EMIS/SystmOne integration, active NHS adoption. Considerations: subscription cost, note quality varies by consultation style — best for structured primary care consultations.
2
Tortus AI
Tortus is a UK-developed AI scribe with strong NHS governance credentials, including DTAC compliance. It has been piloted in several NHS trusts and generates notes that integrate with existing EHR systems. Strengths: NHS-focused governance, growing evidence base (including HSJ-reported efficiency studies), designed for UK clinical workflows. Considerations: newer product — check current integration depth for your specific EHR.
3
Accurx (AI Scribe feature)
Accurx is primarily known as a patient communication platform widely used in UK general practice. It has added an AI scribe feature that integrates with its existing messaging and workflow tools. Strengths: if you already use Accurx for patient comms, the scribe is a natural extension — no new platform to learn. Considerations: the scribe is a feature within a broader platform, not a standalone product.
4
Nabla
Nabla is an international AI scribe with a strong European presence. It supports multiple EHR systems and consultation styles. Strengths: multilingual support, clean interface, supports both primary care and hospital consultations. Considerations: UK-specific governance and EHR integration depth should be verified for your setting.
5
Abridge
Abridge is a US-based AI scribe with significant traction in the American market, including partnerships with major health systems. Strengths: deep US EHR integration (Epic), strong AI models, patient-facing features (patients can receive visit summaries). Considerations: UK adoption is limited — primarily relevant for US-based clinicians or those considering international tools.
6
How to choose
Decision criteria: (1) Does it integrate with your EHR (EMIS, SystmOne, Epic)? (2) Does it meet NHS governance requirements (DTAC, DCB0129/0160, DSPT) if you are in an NHS setting? (3) What is the cost per clinician per month? (4) What note formats does it support (SOAP, free-text, structured templates)? (5) How good is the note quality for your consultation style (short GP consultations vs longer specialist encounters)? Trial before you commit — most products offer free trials or pilot periods.
Governance is not optional
If you work in an NHS setting, any AI scribe must meet your organisation's information governance requirements. This typically includes: DTAC compliance, DSPT registration, data processing agreement, and clinical safety case (DCB0129). Do not deploy a scribe in an NHS setting without checking these requirements with your IG team. Using a non-compliant tool puts you and your organisation at risk.
The 'review every note' rule
No AI scribe produces perfect notes 100% of the time. Hallucinated details, misattributed symptoms, and incorrect medication names can occur. Your professional responsibility is to review and approve every AI-generated note before it becomes part of the clinical record. Treat the scribe as a highly efficient first draft, not a final product.