USEFUL, BUT NOT IN CHARGE: PUBLIC VIEWS ON AI READING MEDICAL SCANS

Abstract

Background: Artificial intelligence (AI) can now help to read X-rays, CT and MRI scans, but whether patients accept it is a different question from whether it works. We explored how far people trust AI to read medical scans, what worries them and what role they want doctors to keep. Methods: An anonymous online questionnaire was shared through WhatsApp, Instagram and school groups between July and September 2026. Respondents aged 16 or over rated ten statements about AI in medical imaging, watched a short news clip about an AI tool used in the NHS, rated their support and could leave a written comment. Spearman’s rank, chi-squared and Wilcoxon tests were used. Results: Of 505 responses, 480 came from people aged 16 or over, and 219 passed a data-quality check and were analysed. Most saw benefits: 61% agreed that AI could speed up diagnosis and 62% that it could reduce waiting times. Yet 66% said AI should always be double-checked by a clinician, and only 25% would be comfortable with AI reading their scan first. Support was linked to trust in hospitals (ρ = 0.41, p < 0.001) and confidence with digital tools (ρ = 0.22, p = 0.001), but not to having heard of medical AI or having had a scan. Mistakes and missed diagnoses were the most common concern in written comments. Conclusion: Respondents saw AI as useful but wanted doctors to stay in charge. Visible human oversight and plain-language explanations may matter as much as accuracy for public acceptance