Abstract
Globally, cigarette smoking results in over 8 million premature annual deaths. Addressing this issue requires high-impact, cost-effective population-level interventions for smoking cessation. Conversational chatbots offer a potential solution given the recent advancements in machine learning and large language models. Chatbots can deliver supportive, empathetic behaviors, personalized responses, and timely advice tailored to users' needs that is engaging through therapeutic conversations aimed at creating lasting social-emotional connections. Despite their promise, little is known about the efficacy and underlying mechanisms of chatbots for cigarette smoking cessation. We developed QuitBot, a quit smoking program of two to three-minute conversations covering topics ranging from motivations to quit, setting a quit date, choosing cessation medications, coping with triggers, maintaining abstinence, and recovering from a relapse. QuitBot employs conversational interactions, powered by an expert-curated large language model, allowing users to ask questions and receive personalized guidance on quitting smoking. Here, we report the design and execution of a randomized clinical trial comparing QuitBot (n = 760) against Smokefree TXT (SFT) text messaging program (n = 760), with a 12-month follow-up period. Both interventions include 42-days of content on motivations to quit, skills to cope with triggers, and relapse prevention. The key distinction between QuitBot and SFT is that QuitBot has communication and engagement features. This study aims to determine: whether QuitBot yields higher quit rates than SFT; and whether therapeutic alliance processes and engagement are mechanisms underlying cessation outcomes. Additionally, we will explore whether baseline factors including trust, social support, and demographics, moderate the efficacy of QuitBot. Trial Registration numberClinicalTrials.govNCT04308759.