Informed consent is foundational to medical ethics, yet its quality or validity may be compromised when the consent-seeking process is delegated to junior doctors – a common practice in the UK, among other jurisdictions – due to inconsistent training, out-of-date knowledge or other factors. Generative AI systems, such as large language models (LLMs), could be used to enhance the consent process by offering standardised, comprehensive, and accessible information delivery, while reducing time pressure on patients and clinicians. However, little is known about public attitudes toward the use of AI in this context and how it might affect perceptions of consent validity, trust, and hospital liability.
This study investigates how laypeople in the United Kingdom (N=376, nationally representative for age, race/ethnicity, and gender) evaluate surgical consent obtained with the assistance of AI systems (Consent-GPT) compared to human clinicians, either junior doctors or treating surgeons. While participants broadly agreed that consent obtained through Consent-GPT was valid, they rated its perceived validity significantly lower than consent obtained by human clinicians. Notably, participants expressed substantially lower satisfaction with AI-mediated consent compared to human-led processes. When assessing litigation scenarios, participants were slightly more inclined to support legal action when consent was obtained through AI rather than human clinicians. However, the strongest predictor of support for litigation was whether complications had been properly disclosed during the consent process, regardless of which agent obtained consent. These findings highlight the need for public engagement and further research to guide the ethical integration of AI into informed consent processes.