Artificial intelligence is starting to sit quietly in the consultation room. Ambient voice technology, sometimes called an AI scribe, listens to the conversation between a clinician and patient and produces a draft clinical note or summary for the clinician to review.
The attraction is obvious. Less typing could mean more time listening to the patient, better eye contact and less time finishing notes after the consultation. The technology is also moving quickly into NHS use. In July 2026, NHS England announced a regional procurement of ambient voice technology across the Midlands, covering 1,239 GP practices and more than 70,000 clinicians across 15 acute and community trusts. NHS England has also developed national guidance to support adoption. (england.nhs.uk)
Most discussion so far has focused on face-to-face care. There is another question worth asking:
What happens when AI is listening to a consultation that is already virtual?
The potential is clear
Clinical documentation takes time, whether care is delivered face-to-face, by telephone or by video. Early evidence suggests ambient scribes can reduce some of this burden.
A 2026 study in general practice found that ambient scribing reduced documentation time and clinician workload, although it did not significantly reduce overall consultation length (van Linschoten et al., 2026). (nature.com)
A UK survey of 1,003 GPs found similar benefits. Among those already using ambient AI, 80% reported spending less time on documentation and 70% reported reduced cognitive load (Blease et al., 2026). (pubmed.ncbi.nlm.nih.gov)
For virtual consultations, this could be particularly useful. When clinicians assess patients remotely, communication becomes even more important. If AI reduces the need to constantly type notes, the clinician may be able to focus more closely on what the patient is saying.
Virtual care already involves less information
A virtual consultation is not simply a face-to-face consultation delivered through a screen. The information available to the clinician changes.
On the telephone there is no visual information. With video, what the clinician can see depends on the camera, connection quality, lighting and what the patient is able to show.
Assessment can therefore become more dependent on history-taking, patient or carer accounts, visible cues where these are available, and recognising when there is not enough information to continue remotely.
This matters when AI is introduced. Ambient scribes are good at capturing conversation. Clinical care includes information that may never be spoken.
A clinician may notice that someone looks pale, appears breathless, struggles to move or looks towards a relative before answering a question. An audio-based system cannot document something it has not heard.
A 2026 study comparing audio-only and vision-enabled AI scribes in simulated medication histories found fewer omissions when visual information was included. Accuracy was reported as 98% with video input compared with 81% using audio alone (McKinnon et al., 2026). (nature.com)
The study was simulated, so it should not be applied directly to routine virtual care. It does, however, reinforce a simple point:
The quality of an AI-generated note depends on the information available to the AI.
The notes still need checking
One risk with AI-generated documentation is that well-written text can appear more accurate than it really is.
The UK GP survey found that 32% of users reported errors occurring often or always. Fourteen per cent reported errors they considered potentially significant or critical. Problems were more common in complex histories, consultations involving several people and consultations in languages other than English (Blease et al., 2026). (pubmed.ncbi.nlm.nih.gov)
That is particularly relevant in community healthcare.
Consultations may involve carers, family members or interpreters. Patients may have several long-term conditions. Conversations can move between symptoms, medication, social circumstances and safeguarding concerns.
The question is therefore not simply whether an AI scribe can generate a note. It is whether it can generate an accurate enough draft from that particular consultation.
NHS England is clear that clinicians remain responsible for checking AI-generated documentation before it becomes part of the clinical record. (digital.nhs.uk)
The AI assists with documentation. It does not take responsibility for it.
Patients also need to be considered
Patients need to know when ambient scribing is being used and should have the opportunity to object. NHS England also recognises that highly sensitive consultations may not always be suitable for ambient scribing. (digital.nhs.uk)
This feels especially important in virtual care. Imagine discussing mental health, safeguarding, continence or end-of-life care over video and then being told that an AI system is also listening.
Some patients may be completely comfortable with that. Others may become more careful about what they say. So implementation cannot be based only on whether the technology works technically. It also needs to consider whether patients still feel able to speak openly.
It comes back to clinical fit
Perhaps the most useful question is not whether ambient AI works.
It is:
Where does it work, for whom and under what circumstances?
A straightforward follow-up consultation may be well suited to ambient scribing. A multilingual consultation involving several family members and a patient with complex needs may be much harder.
The same applies to virtual care more broadly. In my own research into virtual consultations, one issue I keep coming back to is whether the technology fits the patient, the clinical task and the consultation taking place. Ambient AI is likely to be similar.
NHS organisations therefore need to evaluate more than time saved. They should also look at errors, omissions, clinician corrections, patient experience and how well these tools perform in telephone, video and complex community consultations.
Ambient AI has real potential. It may reduce documentation burden and give clinicians more time to focus on patients. It also introduces new risks, particularly where the consultation already provides limited clinical information.
So the question is probably not:
Should we use AI scribes in virtual consultations?
A better question is:
Which virtual consultations are they suitable for, what might they miss, and how do we make sure they support rather than interfere with clinical care?
References
Blease, C. et al. (2026) ‘Ambient AI in primary care: an exploratory mixed methods survey of UK general practitioners’, BMJ Health & Care Informatics.
McKinnon, R.A. et al. (2026) ‘Vision-enabled AI scribes reduce omissions in clinical conversations: evidence from simulated medication histories’, npj Digital Medicine.
NHS England (2026) Using AI-enabled ambient scribing products in health and care settings.
NHS England Midlands (2026) Midlands leads the way on ambient voice technology. 15 July 2026.
van Linschoten, R.C.A. et al. (2026) ‘Ambient scribe in general practice: a multi-perspective before-after longitudinal mixed-methods study’, npj Digital Medicine, 9, 299.
