kcl.ac.uk via Reddit

KCL Study: 1 in 7 UK Adults Skip GP for AI

healthcare ai assistants healthcare-ai consumer-adoption

Key insights

  • One in seven UK adults have used AI instead of contacting a GP, per a nationally representative KCL survey.
  • One in five AI health users said it discouraged them from seeking professional in-person care.
  • Women oppose NHS clinical AI use at nearly double the rate of men, with safety and accuracy as the top concern at 39%.

Why this matters

AI health substitution is no longer a hypothetical liability -- it is a measurable behavior pattern affecting NHS demand curves and diagnostic delay rates right now. For founders building health AI, the one-in-five discouragement finding is a direct regulatory target: any product that reduces care-seeking for serious conditions will attract MHRA and CQC scrutiny, especially post-Cass Review sensitivity around mental health tools. For technical leaders deploying general-purpose LLMs, this data is evidence that users are already using consumer chatbots in high-stakes medical contexts regardless of whether those products are designed or tested for it.

Summary

King's College London's nationally representative survey puts a number on what many suspected: one in seven UK adults have already substituted an AI chatbot for a GP visit, and one in ten have used AI in place of professional mental health support. Convenience was the top driver at 46%, but 39% cited uncertainty about whether their symptoms were serious enough to warrant a doctor -- which is precisely the scenario where a wrong AI answer carries the highest cost. One in five respondents said AI actively discouraged them from seeking in-person care. Essentially: (King's College London, NHS) are now contending with a population that is quietly rerouting around formal healthcare infrastructure. - 39% of respondents cited safety and accuracy anxiety as their top concern about AI use in NHS clinical settings. - Women oppose clinical AI adoption at nearly twice the rate of men, a split that will complicate any NHS rollout narrative. - Public opinion on NHS clinical AI use is almost exactly even, meaning there is no clear mandate in either direction. The underlying dynamic isn't about chatbot quality -- it's about a healthcare system under enough access pressure that a statistically significant share of the population now treats AI as a credible first-line substitute.

Potential risks and opportunities

Risks

  • NHS integrated care boards could see increased A&E attendance from patients who used AI during a deteriorating condition and delayed presentation -- a cost and safety exposure that lands on trusts, not AI vendors.
  • Consumer AI companies (Google, OpenAI, Microsoft) face heightened UK regulatory risk if MHRA or the CQC move to classify general-purpose chatbots as medical devices based on documented substitution behavior.
  • Mental health charities and crisis services may see users arrive later and in worse condition if AI tools are absorbing early-intervention contact that would previously have reached Samaritans, MIND, or NHS Talking Therapies.

Opportunities

  • Clinical AI vendors with auditable triage logic (Babylon Health successors, Kry, Doctorlink) can use the KCL safety-concern data to differentiate on explainability and escalation design in NHS procurement bids.
  • UK digital health insurers (Vitality, Bupa) could productize structured AI triage that routes to human clinicians above defined symptom thresholds, turning the discouragement finding into a compliance-safe selling point.
  • Research funders and NHS England have a clear mandate to commission outcome studies tracking the cohort who substituted AI for GP contact -- a gap the KCL authors themselves flagged, creating a near-term grant opportunity for health AI evaluation teams.

What we don't know yet

  • Whether the survey captured which specific AI tools (ChatGPT, Gemini, NHS-integrated tools) users substituted for GP contact, which would clarify regulatory scope.
  • Whether delayed or avoided care in the surveyed population produced measurable downstream health outcomes -- the study captures intent and behavior, not clinical consequence.
  • How the 1-in-7 substitution rate compares to equivalent behavior in countries with easier GP access, which would separate NHS access-pressure effects from general AI adoption effects.

Originally reported by kcl.ac.uk

Read the original article →

Original headline: One in Seven UK Adults Have Used AI Instead of Seeing a GP, King's College London Study Finds — One in Five Say It Discouraged Them From Seeking Professional Care