NHS AI Scribes Race Ahead of Public Trust as Healthwatch Polling Reveals Consent Gap

The NHS is rolling out AI scribes quickly. Too quickly, some would say, for the consent infrastructure to keep up.

Published on 16 July 2026, Healthwatch England's nationally representative survey of 4,039 adults found public support for AI transcription tools in clinical settings split almost evenly: 38% in support, 37% opposed. But the intensity gap is the telling figure. Almost twice as many people are strongly opposed to their use (21%) as are strongly supportive (11%), and around 81% say they want clinicians to tell them and seek consent before a scribe is activated.

Healthwatch launched the research after patients raised concerns with local Healthwatch branches that AI scribes were generating errors in medical records. The YouGov fieldwork ran from 16 to 27 April 2026, giving the findings a credible, large-sample foundation.

A Rollout Running Faster Than Trust

The timing could not be more pointed. On 4 July 2026, NHS England confirmed it is accelerating a major AI deployment across the health service, backed by £10 billion in technology funding allocated by the government last year. Among the headline commitments: a national rollout of AI tools that record patient-clinician conversations to generate real-time transcriptions and clinical summaries. A new AI triage tool inside the NHS App is also being prioritised, targeting more than 200,000 patients within the next 12 months, with full NHS App availability set for April 2028.

Real efficiency gains exist. West Midlands trusts deploying the Heidi scribe system have reported saving around six minutes per consultation, with one clinic cutting its letter backlog from six months down to 14 days. Earlier this year, the University Hospitals of Leicester and University Hospitals of Northamptonshire provided more than 10,000 clinicians with access to ambient voice technology, supporting around 2.5 million outpatient appointments a year.

But as the Nuffield Trust noted in the Healthwatch findings: there is "some understanding of the outputs (minutes saved), but very little understanding of the outcomes (what those minutes enable)."

The efficiency case is getting stronger. The trust case is not.

Where Comfort Collapses

The Healthwatch data exposes specific fault lines that commissioners and developers should read carefully. Around 48% of respondents said they would be comfortable with an AI scribe during a routine check-up. That figure falls to 28% for mental health consultations, and to 23% for sessions involving domestic abuse. Disabled people were more likely to be sceptical and uncomfortable than those without disabilities. Younger people were the most uncertain.

A public dialogue commissioned by AI healthcare firm Newton's Tree in March 2026 reached similar conclusions: participants supported clinicians having the discretion to switch scribes off during sensitive consultations, wanted consistent national standards from regulators, and were frustrated by what the report described as "current low awareness" of AI scribes in NHS settings.

The Regulatory Machinery Is Moving

The MHRA's regulatory process is accelerating in parallel, though the sequencing is tight. The National Commission into the Regulation of AI in Healthcare, chaired by Professor Alastair Denniston and hosted by the MHRA, is tasked with publishing its recommendations by summer 2026. The Healthwatch findings feed directly into that process, and the Commission's final recommendations may prompt the MHRA to issue new rules on AI scribes as medical devices.

The MHRA's draft Medical Devices (Amendment) Regulations 2026, published in May, introduce Predetermined Change Control Plans (PCCPs) for software and AI-enabled devices. These allow manufacturers to pre-define future changes without requiring full re-certification each time, a potentially significant runway for scribe vendors updating their models in production. The draft regulations are expected to be adopted in December 2026 and come into force in June 2027.

On the medicines side, the MHRA launched a parallel AI sandbox in June 2026, creating a controlled environment to test AI tools for predicting drug safety. Adverse drug reactions send around 250,000 people to hospital in the UK every year at a cost of over £2 billion to the NHS. Up to five AI-driven approaches will be tested in the programme's first phase, starting summer 2026.

What the Market Data Shows

AI Business Dispatch analysis of Companies House and Intellectual Property Office data, as of July 2026, adds commercial texture to the regulatory urgency. Just 56 new SIC 86.10 (hospital activities) companies were incorporated in Q3 2026, down 82.1% versus the prior period, suggesting the acute-care-tech startup wave is consolidating rather than expanding. Class 5 UK trademark filings (pharmaceutical and medical preparations) fell to 558 in Q3 2026, a 78.2% drop on the prior period.

Perhaps most striking: 98.8% of active SIC 86.10 companies hold no Class 5 trademark at all. In a sector where product credibility and brand protection matter, that is a significant governance blind spot. Companies building AI scribe tools, clinical decision-support software, or ambient documentation products should be watching this closely as regulatory scrutiny intensifies.

Consent Is the Differentiator

For every trust deploying scribes at pace, the consent framework they set today is the one regulators may have to revisit later. Healthwatch recommends notice at booking, respect for patient requests to switch scribes off during sensitive topics, and consistent national standards.

As one recent clinical study on AI scribe risks put it, reliability in a demo is not reliability in a noisy clinic. The vendors who move from leading on transcription accuracy to leading on audit trails, consent flows and review workflows may find themselves on the right side of whatever the MHRA's National Commission recommends.