By 2060, one in four Australians will be over 60, intensifying demand for scalable and cost-effective strategies to promote healthy ageing and independence. While digital health interventions such as BetterBrains have demonstrated feasibility in reducing dementia risk factors, their scalability is limited by the heavy reliance on human staff to support participants. Advances in generative AI now provide a unique opportunity to reduce this burden through empathetic, accurate, and timely chatbot responses.

This project will be the first to systematically evaluate AI-generated chatbot responses against human responses within a large digital health RCT. Using over 200 real participant queries from BetterBrains, the study will compare two chatbot training approaches—Prompt-Engineered AI and Retrieval-Augmented AI—assessing their ability to respond across administrative, technical, and emotional support categories. Responses will be rated by older adults and researchers for quality, empathy, and suitability.

Expected outcomes include:

1. Evidence on whether AI chatbots can deliver responses of equivalent or superior quality and empathy compared to study staff.

2. Identification of the most effective AI training approach (Prompt-Engineered vs Retrieval-Augmented) for supporting participants in digital trials.

3. Insights into chatbot effectiveness across different query types, informing best practices for AI integration in health research.

4. Pilot data to support future funding applications, enabling larger-scale trials and deployment of chatbot-assisted digital health programs.

Ultimately, this project will generate critical knowledge on the role of AI in supporting older adults in dementia prevention programs, addressing barriers of scalability, accessibility, and cost. If successful, it will pave the way for AI-driven engagement strategies across diverse digital health interventions, helping older Australians live healthier, more independent lives while reducing strain on the healthcare system.