TL;DR

Research from King’s College London finds that one in seven UK adults has put a health question to an AI chatbot rather than contacting a general practice. Among those who did, a fifth said the tool never suggested they get a professional opinion. A quarter of users gave NHS waiting times as their reason.

Substitution, not supplementation

The distinction that matters here is between consulting a chatbot alongside the NHS and consulting one instead of it. The survey, covering close to 2,100 adults in March, measures the second. Convenience and curiosity featured among the stated motivations, along with doubt over whether a symptom warranted a GP appointment at all — but the waiting-time answer is the one with policy consequences, because it describes demand being displaced rather than met.

The safety-netting failure is the more troubling number. Where a receptionist or 111 call handler would route a caller onward, a fifth of chatbot users were sent nowhere. That gap sits outside existing regulation: a general-purpose assistant answering a symptom question is not classed as a medical device, so the safeguards applied to clinical software do not reach it.

The searching layer has already moved

Stephanie Eltz, a former NHS surgeon who now runs the clinician review platform Doctify, argues the debate over whether patients should use these tools is “futile”, and that attention belongs on whether the information underneath them is accurate and traceable.

Her company’s own traffic supports the claim that behaviour has already shifted: visits arriving from AI-generated search results are up 680% year on year. OpenAI’s figures point the same way, with health accounting for roughly 5% of chatbot conversations globally and a quarter of ChatGPT’s 800 million regular users raising a health prompt each week.

Three bodies ran the work jointly: the university’s Policy Institute, alongside King’s Health Partners and Responsible AI UK. Their researchers stressed that people must still reach professional care when they need it.

Looking forward

Britain now has two health-advice systems running in parallel, and only one of them is accountable. NHS England has been rolling out AI to clinicians on the supply side; this research describes patients adopting it faster on the demand side, without the clinical governance that surrounds the official channel.

The practical response is unglamorous — making authoritative NHS guidance easy for models to retrieve and cite, and pressing vendors on escalation behaviour. Waiting for patients to stop asking is not a strategy.