<p>The global demand for high-quality, robust and up-to-date evidence to guide decision-making has never been higher. The vast quantity of scientific literature being produced and made accessible presents an unparalleled opportunity for evidence-based decision-making to become a widespread reality. In addition, the world has at its fingertips cutting-edge technologies, such as AI, to make sense of this extensive knowledge base and deliver insights more quickly to decision-makers most in need. AI-powered evidence syntheses promises to be transformative, saving many lives and enhancing livelihoods globally. However, achieving this requires substantial cultural shifts in the evidence community, including amongst both AI developers and users to shape both trustworthy AI and trust in AI. Current efforts to establish best practices are emerging, but progress is hindered by the lack of clear consensus on what constitutes trustworthy AI for evidence synthesis. Philanthropic investments in trustworthy AI systems, alongside robust evaluations of trust in AI for evidence synthesis, must be prioritised to determine the conditions required for an enabling environment. Mainstreaming AI for reliable, faster and cheaper evidence synthesis demands a better understanding of trustworthy AI and trust in these systems. Funders should prioritise aspects of trustworthiness and trust whilst balancing the drive towards ongoing innovation.</p>