<p>Climate scepticism remains an important barrier to public engagement with accurate climate information, because sceptics often actively avoid information that contains climate science facts. There still lacks a scalable, repeatable intervention to boost sceptics’ engagement with climate information. Here we show that generative artificial intelligence can enhance engagement with climate science among sceptical audiences by subtly modifying headlines to reduce anticipated disagreement, regret and negative emotions, without compromising factual integrity. Headlines of climate science articles modified by an open-source large language model led to more bookmarks and more upvotes, and these effects were strongest among the most sceptical participants. Participants who engaged with climate science as a result of this intervention showed a shift in beliefs towards alignment with the scientific consensus. These results show that generative artificial intelligence can alter the information diet sceptics consume and holds promise for advancing public understanding of science when responsibly deployed by well-intentioned actors.</p>