Evaluation of Artificial Intelligence for Patient Self-Triage: Comparison of General-Purpose AI Platforms With the NHS 111 Online Symptom Checker in the United Kingdom

Emergency departments (EDs) in the UK face substantial pressure due to non-urgent attendances. This technical report evaluates the performance of two general-purpose artificial intelligence (AI) platforms (ChatGPT GPT-5 (OpenAI, San Francisco, California, USA) and Gemini AI v2.5 Flash (Google, Mountain View, California, USA)) for patient self-triage, compared with the NHS 111 online symptom checker. Ten simulated patient scenarios, including five emergency and five non-emergency cases, were assessed against National Institute for Health and Care Excellence (NICE) guideline-based gold standards. Both AI platforms correctly identified all emergency cases; NHS 111 under-triaged one acute emergency. For non-emergency scenarios, AI occasionally over-triaged, recommending emergency assessment for pyelonephritis, whereas NHS 111 correctly classified all non-emergencies. AI triage responses were faster and required fewer follow-up questions than NHS 111, although sometimes producing unclear recommendations. The findings suggest that general-purpose AI may serve as an adjunct to NHS 111, supporting patient self-triage and potentially reducing ED burden. Future work should include larger-scale testing, real patient data, and prospective safety evaluation to assess clinical feasibility.