AI will tell NHS patients if they need a GP appointment

[highplainsdem]

Source: Telegraph via Yahoo

The move to use AI chatbots as the front door to primary care marks one of the biggest changes to the way patients access family doctors for a generation.

-snip-

If a GP appointment is recommended, a clinician will decide whether one is needed and how urgently the patient should be seen. In other cases, the AI will recommend alternatives such as pharmacies, self-care or emergency services.

During the trial, every recommendation will be checked by clinicians, who will be able to override the chatbot's advice retrospectively, calling patients back if they believe a GP appointment or more urgent care is needed.

However, the long-term aim is to remove that safeguard, if the trial shows AI works safely, allowing it to advise millions of patients without clinical back-up.

-snip-

Read more: https://www.yahoo.com/news/us/articles/ai-tell-nhs-patients-gp-213000250.html

---

Hallucinating chatbots, advising patients without any clinical backup by humans.

This is crazy. Unfortunately the current UK government is entirely too close to AI companies

And here in the US we have Dr. 0z just as eager to have chatbots offering medical advice.

See my OP in GD about a study published a few months ago showing how badly chatbots fail at giving the correct medical advice:

'Unbelievably dangerous': experts sound alarm after ChatGPT Health fails to recognise medical emergencies (The Guardian)
https://www.democraticunderground.com/100221066192

[RockRaven]

1. This will turn out badly.

Generational-scandal-level badly, if it gets broadly implemented after the trial (no matter how good those trial results look, which they won't).

[Karasu]

2. No government should be REQUIRING an unregulated technology to be used for ANYTHING, health care least of all.

[xocetaceans]

4. That is an insane choice. 'AI' is prone to hallucinate. Will this modern incarnation of Clippy be up to the task?

That seems very doubtful.

This passage is from a seemingly separate article from the same URL as the OP but from further down the page:

NHS will turn science fiction into science fact
By James Murray

...

My priority as Health and Social Care Secretary is now to maximise the return on that investment for patients and staff.

That's why I've made sure we're backing the projects and innovations I believe will deliver the greatest improvements in diagnosis times, operating costs and efficiency.

I'm talking about the tools and processes that will cut waiting lists faster, reduce non-attendance at appointments and free up more time for clinicians to focus on what they do best: caring for patients.

These include ambient voice technology, AI-assisted triage through the NHS app, Microsoft Copilot, digital therapeutics that support treatment at home, and follow-up services initiated by patients that give people greater control over their care.

...

James Murray is the Secretary of State for Health and Social Care and the Labour MP for Ealing North.

https://www.yahoo.com/news/us/articles/ai-tell-nhs-patients-gp-213000250.html

[highplainsdem]

5. Yes, it's the AI-assisted triage. The sort of thing the study I mentioned tested.

From my OP a few months ago and the study I linked to:

https://www.democraticunderground.com/100221066192

https://www.nature.com/articles/s41591-026-04297-7

Brief Communication
Published: 23 February 2026
ChatGPT Health performance in a structured test of triage recommendations
Ashwin Ramaswamy, Alvira Tyagi, Hannah Hugo, Joy Jiang, Pushkala Jayaraman, Mateen Jangda, Alexis E. Te, Steven A. Kaplan, Joshua Lampert, Robert Freeman, Nicholas Gavin, Ashutosh K. Tewari, Ankit Sakhuja, Bilal Naved, Alexander W. Charney, Mahmud Omar, Michael A. Gorin, Eyal Klang & Girish N. Nadkarni


Abstract
ChatGPT Health launched in January 2026 as OpenAI’s consumer health tool, reaching millions of users. Here, we conducted a structured stress test of triage recommendations using 60 clinician-authored vignettes across 21 clinical domains under 16 factorial conditions (960 total responses). Performance followed an inverted U-shaped pattern, with the most dangerous failures concentrated at clinical extremes: non-urgent presentations (35%) and emergency conditions (48%). Among gold-standard emergencies, the system under-triaged 52% of cases, directing patients with diabetic ketoacidosis and impending respiratory failure to 24–48-hour evaluation rather than the emergency department, while correctly triaging classical emergencies such as stroke and anaphylaxis. When family or friends minimized symptoms (anchoring bias), triage recommendations shifted significantly in edge cases (OR 11.7, 95% CI 3.7-36.6), with the majority of shifts toward less urgent care. Crisis intervention messages activated unpredictably across suicidal ideation presentations, firing more when patients described no specific method than when they did. Patient race, gender, and barriers to care showed no significant effects, though confidence intervals did not exclude clinically meaningful differences. Our findings reveal missed high-risk emergencies and inconsistent activation of crisis safeguards, raising safety concerns that warrant prospective validation before consumer-scale deployment of artificial intelligence triage systems.

[Miguelito Loveless]

8. Hey, no defaming Clippy

He may have been annoying, but he never had hallucinations.

[not fooled]

6. The proles don't deserve health care

This is how they will get there...step by step. And a largely ignorant, careless, oblivious populace will not stop it.

[Miguelito Loveless]

7. This is incredibly stupid.

Simply tax billionaires, close tax loopholes, and fund the NHS properly. Stop trying to provide healthcare on the cheap while enriching foreign Tech Bros b

[róisín_dubh]

9. Jesus Christ this is ridiculous

I don’t want a conversation with Gemini about my HRT struggles just let me do a phone call with my friggin’ GP! Or whatever I need to discuss!
I swear Labour is trying to lose. People do not love AI

[snot]

10. My own experiences with the quality of health care

indicate that it's already gone drastically downhill with the incorporation of various digital or computer-based facilities. It's gotten so bad that I actually did a bit of research to try to figure out why it's become so hard to reach a real doctor – are there fewer doctors in the US than there used to be, say, 30 years ago? As it happens, no – there are more doctors per capital than ever.

When I was a kid, you could reach a human carer quickly and easily: pick up the phone and reach a nurse if not a doctor. You might have to leave a message, but you'd get a call back reasonably quickly, and if your question wasn't too complicated, you could sometimes even get an answer over the phone.

Now, I have to work my way through an automated system in order to reach a human at a third-party answering service that not only can't answer any questions but can't even make an appointment – they then call the doctor's office, and it may be one day-to-literally weeks before I get a call back from someone who still can't answer my question but can at least make an appointment. Then it turns out that the first appointment isn't available for about 4 months. If you need help before then, they suggest going to an emergency room.

The other way to try to reach a human carer is via the doctor's portal – an extremely inconvenient way to communicate, since you have to log in, and some portals are so poorly designed that in one case, it took me months to figure out the only way for me to consistently get logged in: it required a regular 20-minute process of log-in attempts, phone calls for verification codes, password-resetting, account re-creating, etc. And it seems that at least some doctors' offices never read the messages left in their portals; and/or if they do, it may take several days to actually receive an answer.

In another example, I tried to obtain a second opinion from a doctor I'd seen before for the same condition within the same year as my original consultation with them. I was told that, although their system was supposed to be coordinated with that of my primary doctor so as to enable the 2nd-opinion doc to access my intervening test results, the second-opinion doctor's office was for some reason unable to access those results – so I'd need to obtain digital copies of all of them (there were quite a few, many irrelevant) and upload them to the second-opinion doctor's portal. I dutifully did this – in some cases, I had to physically visit the testing facility and have them print a copy and then scan it at home – and uploaded them all, only to be told that the second-opinion doctor – whose office was far away and located in a city with very expensive lodging – refused to look at any of it or consult with me via zoom, and would provide a second opinion only if I travelled to their city, which was far away and would require me to spend two nights in an expensive hotel.

In another example, it was 4 months before I was allowed to see a surgeon who could fix an agonizingly painful and partially-debilitating condition (all of this despite the fact that I have the best insurance money can buy in my state!)

We need to require Congress to rely on the same health care options as are available to the "least" of us.