Artificial intelligence (AI) is becoming more widely used in healthcare by both patients and NHS practitioners.
Clearly, AI technology has the potential to help speed up elements of medical care, whether that’s reducing paperwork, interpreting information or managing workload. However, there are worries about whether patient safety, transparency and accountability are keeping pace.
Imagine going to your GP with a worrying symptom. In some GP surgeries now, as well as casualty departments, during your appointment a computer programme can listen to the conversation and write up your medical notes.
Your doctor may welcome the extra time this creates to focus on you. But what if the programme gets something wrong? A symptom may be missed, a medication could be recorded incorrectly, or an important detail disappears from the notes.
It might seem like a small mistake and something to easily blame on the algorithm or tech, but medical records can shape future decisions about your health. These AI generated notes are meant to be carefully checked. However, if an error goes unnoticed, it could have consequences long after the original appointment.
Similarly, what if you get so fed up of waiting to see your GP or a specialist that you put your symptoms or concerns into AI platforms, only to be told the wrong information and undergo unnecessary worry or treatment.
The government has recently backed the full recommendations by the AI Commission which sets out how AI in healthcare will be delivered across the UK. That means it won’t be going away anytime soon. So, if AI is going to play an increasing part in our healthcare, patients need to know that their safety, their voices and their right to answers won’t be overlooked.
What is AI being used for in healthcare?
AI use in the NHS can vary from assistive technology (supporting in decision making), to autonomous (making decisions without human oversight).
Some of the tasks that are commonly being completed by machines include:
- Diagnostic tools such as imaging recognition for radiology
- Predictive analytics to identify patients most at risk of deterioration
- Decision support systems for treatment recommendations
- Administrative or operational tools such as AI receptionists, triage, note taking scribes or chatbots
Clinicians’ use of AI varies sharply by task. There has been strong uptake in admin and documentation support, compared to limited use in clinical decision making.
AI could give doctors more time for patients
With NHS staff under considerable pressure, on the surface it makes sense why over 80% of NHS staff support the use of AI to automate those routine and admin tasks that take up much of their time. Reducing unnecessary paperwork could make a meaningful difference to their working lives and to the care patients receive.
A January 2026 article from Skills for Health highlights research into AI note-taking tools, sometimes called AI scribes. These tools can listen to consultations and help produce written records, reducing the need for clinicians to type up notes themselves.
When introduced appropriately, AI could help healthcare professionals spend less time on administration and more time with the people they are treating. According to the Skills for Health report, programmes like Copilot could save each staff member 43 minutes per day, with an estimated 400,000 hours freed up per month if it was to be deployed across the NHS.
A trial led by Great Ormond Street Hospital reported promising results. The use of AI scribes was associated with 23.5% more direct interaction time between clinicians and patients and appointments that were 8.2% shorter. In an A&E setting, the trial also reported a 13.4% increase in patient output per shift.
However, there is an important distinction between technology that helps a doctor do their job and technology that is allowed to go unchecked. Saving time is valuable only if patients continue to receive safe, effective care.
Neither will it save time if doctors are having to review all transcripts for errors, especially if they haven’t taken notes in appointments to refer back to for evidence of what was said.
Doctors and other healthcare professionals already make difficult decisions, often under pressure. Introducing a new tool should not leave them guessing about whether its output can be trusted, how it should be checked or what to do if something appears wrong.
A computer can make a mistake
These tools can be useful, but they are not infallible. An answer that looks convincing is not necessarily correct. If a busy clinician accepts that answer without checking it against the patient's symptoms, medical history and circumstances, an opportunity to spot a mistake could be lost.
Healthwatch England has shared patients’ experiences of AI-generated medical notes containing inaccuracies including incorrect diagnoses.
One woman was left upset after being wrongly told she had serious nerve damage that can lead to Multiple Sclerosis (MS). She was an NHS professional so questioned the results of MRI. There had been a typo which should have actually said ‘null’ demyelination.
Another case saw a patient prescribed a different drug with a similar name. This was picked up by the patient and not the doctor. In another example, an AI generated summary letter didn’t say the doctor told a patient to seek a repeat prescription for migraine medication. This error could have left them unable to get the medication they needed.
Individual experiences cannot tell us how often these problems occur. Nor does the available evidence establish that AI-generated notes are generally less accurate than notes written by clinicians.
However, these accounts illustrate why checking matters. A patient may not know what has been recorded about them. Another healthcare professional may rely on those notes weeks or months later. An error that seems minor at first could become more significant if it influences a future decision.
Staff need to understand when AI can help, when its output needs checking and when they should not rely on it. That may sound obvious, but it raises a practical question: can healthcare professionals realistically check AI-generated information if they have not been given the training, time or guidance to do so?
Responsibility cannot simply be placed on an individual clinician and left there. Healthcare organisations need appropriate procedures, training and support to help staff use these tools safely.
Patient exposure to misinformation is rising
As well as healthcare professionals using AI in their clinics, patients are also increasingly turning to AI for health advice. Around 60% of clinicians say that patients are arriving at consultations with AI-generated health information.
After recently spending an evening in our local Urgent and Emergency Care Centre, there were many patients waiting who were overheard as saying ‘What does ChatGPT say?’ You may have even put your own symptoms into an online platform to see what ‘Doctor Google’ has to say.
According to Healthwatch England, people are generally using AI to support their health because of difficulty accessing NHS services or due to poor experiences with the NHS.
In their review, they found that people turned to AI tools due to numerous reasons, from frustration trying to get advice from NHS 111 to patients who were struggling to get a diagnosis from their GP.
People were generally happy or impressed with the information and advice AI tools provided. There were, however, some concerning reports of AI tools providing inaccurate information.
In a post shared on Instagram, the gynaecological cancer charity, Eve Appeal reported that four in ten people have reported seeing misleading health information online in the last year, up from one in ten in 2023. This has already had a detrimental effect, with a report by The London School of Hygiene and Tropical Medicine finding that a third of people aged 18-24 know someone who has suffered serious harm as a result of making a health decision based on false or misleading information.
There are currently 115 health charities in the Patient Information Forum that are calling on the government to develop a framework for safe AI generated health information so that people are always directed to trusted, evidence-based information and advice online.
Patients want to know when AI is involved
For patients, a healthcare appointment can be a deeply personal experience and, while over half of the public support the use of AI when safeguards are in place, the majority still want human connection.
People may be discussing pain, mental health, a frightening diagnosis or circumstances they find difficult to talk about. They need to feel comfortable speaking openly, knowing that what they say will be understood and recorded properly.
Research by Healthwatch England suggests that people want greater openness about the use of AI in their appointments.
In polling of over 4,000 adults, 81% said they wanted healthcare professionals to tell them when AI note taking tools were being used and ask for their consent. AI frequently involves processing confidential and personal data, so consent should still be logged.
Some patients did say they wouldn’t feel comfortable discussing certain subjects if AI was involved, particularly sensitive subjects like domestic abuse, mental health or sexual wellbeing.
Patients should not have to wonder whether a computer is listening, what it is recording or whether anyone will check that it has got things right.
The National Commission into the Regulation of AI in Healthcare has recognised the importance of transparency. Its recommendations call for a proportionate approach to informing patients about AI use, including consideration of when people should be able to opt out.
Doctors have concerns about who is responsible when AI goes wrong
Patients are not the only people asking questions. Skills for Health also discusses research from the Nuffield Trust involving 2,108 GPs that found that, among GPs who did not use AI tools, 89% identified professional liability and legal risk as major concerns.
This suggests that some healthcare professionals are not just reluctant to embrace new technology, but are concerned about how to use it safely and what might happen if an AI-generated answer is wrong.
NHS Resolution has published guidance and FAQs on indemnity and liability considerations when AI enable tech is used in the NHS. It states that it is covered by existing indemnity, however if harm arises from the failure or malfunction of AI systems then the manufacturer may be liable under product liability law.
Conversely, AI scribes – of which there are 27 different types being used by the NHS – are not classed as medical devices by the MHRA so there is no oversight to ensure they are safe and effective to use.
Irrespective of product liability, the use of AI for the purpose of care, diagnosis or treatment will not negate the duty of care owed to patients.
It is far more likely that claims will be made against the NHS organisation rather than the AI developers or manufacturers. However, the law around the standard of care delivered using AI is developing and not yet clarified by relevant case law.
In the interim, there must be clear guidance about who is responsible for checking information, how errors should be reported and how concerns should be investigated.
If something goes wrong, patients deserve answers
Perhaps the biggest concern arises when a patient has been harmed.
If an AI tool contributes to an incorrect medical record, a missed diagnosis or a delay in treatment, there will be many questions to ask:
- Did the clinician appropriately rely on AI or fail to apply their own judgment?
- Was AI used in accordance with the guidance, or was it without appropriate safeguards?
- Were there any known limitations or performance concerns about AI that were disregarded?
- Or did several things go wrong at once?
The answer will depend on the circumstances. It would be wrong to assume that every AI-related error amounts to medical negligence. But it would also be wrong for patients to find themselves facing unanswered questions simply because technology was involved.
Patients should still have access to financial redress if they are harmed because the care they receive fell below what they should reasonably expect. The involvement of AI should not make that process more confusing or leave patients feeling that nobody is responsible.
The Government has promised action. Now it must make a difference.
As part of its 10-Year Health Plan, the government wants to make the NSH the most AI-enabled care system in the world. To do so, the National Commission into the Regulation of AI in Healthcare has been set up to look at the potential benefits to patients – so it is not a case of if AI should be used but how.
Chaired by consultant ophthalmologist Professor Alastair Denniston, with vice chair Professor Henrietta Hughes, a GP and patient safety commissioner for England, the Commission has sought opinion from the public, patients, healthcare professionals and policy makers. Using their findings, they have made 44 recommendations about how AI should be regulated and monitored as its use grows, all of which were accepted by the government in October 2026.
One of its key points is that checking a product before it is introduced is not enough. Its performance also needs to be monitored once it is being used with real patients. The Commission wants clearer ways to identify and report problems, better information about the performance of AI tools and stronger arrangements for deciding who is responsible at different stages of their use.
It also highlights the importance of training healthcare staff and ensuring patients can access redress when care falls below what they should reasonably expect. These are welcome recommendations, but accepting them is not the same as putting every safeguard into practice.
The real test will be whether they lead to meaningful changes in hospitals, GP surgeries and other healthcare settings. Can staff recognise when something is wrong? Will concerns be acted on? Will patients receive clear explanations when technology has played a part in their care?
The government may want to see healthcare catch up with other sectors, but patients deserve the reassurance that AI will be introduced responsibly and that accountability remains clear when concerns arise.