A collaborative project between the NIHR Policy Research Unit in Healthy Weight , the NIHR Innovation Observatory, and the NIHR Policy Research Unit in Healthy Ageing

Background

Systematic and rapid reviews are important tools for making decisions in health and social care. New AI technologies are helping speed up these reviews, saving time and effort. One such tool is called Elicit, which helps with reviews but still lets researchers control each step.  Some studies have shown that Elicit is useful for exploring topics and acting as a second reviewer. However, it’s not clear if it can fully replace reviews led by humans. There are concerns about using AI in this area, such as bias, especially affecting underrepresented groups, lack of clarity in how AI works and finds information and questions about how accurate and repeatable its results are.  To improve our understanding of how Elicit might support responsive and rapid research scoping and evidence synthesis for policy we are working with the NIHR Policy Research Unit in Healthy Weight and the NIHR Innovation Observatory on a collaborative project. 

Aims and Objectives

We want to learn more about how Elicit works in a fast-paced policy research environment, specifically looking at two main potential use cases:

1. Helping to develop and improve complex policy questions
2. Acting as a tool for quick evidence reviews

Aims

• Can Elicit be used as an aid to develop, refine and translate complex health and social care policy questions?
• Can Elicit be reliably used as a rapid review evidence synthesis tool in the context of responsive policy research?

Methods

The project will test how well an AI tool called Elicit can help with policy research. In the first phase, it will use real policy questions from two research areas to see if Elicit can summarise existing evidence and highlight gaps. We will check how accurate these summaries are, and policy teams will say how useful they find them. In the second phase, the team will test whether Elicit can carry out a quick evidence review by repeating a 2025 study on promoting physical activity among older, disadvantaged people, a topic influenced by factors such as income, ethnicity, and gender groups. This topic is complex and includes many factors like income, ethnicity, and gender.

Policy Relevance

Fast and thorough reviews can help policymakers make informed policy decisions. AI tools like Elicit could help speed up this work and help develop better questions and gather evidence. Elicit is already being used in some government and health settings, but we still don’t know how reliable or accurate it is. Policymakers need solid evidence about its strengths and weaknesses to use it confidently.

Delivery dates

November 2025 – January 2026

Related