The release of ChatGPT has brought the power of large language models (LLMs) to the public’s attention and sparked academic interest in how well such models can generate different types of textual content as well as how their outputs might impact individuals. While research to date has found that LLMs are capable of generating persuasive messages in political contexts and could be used to amplify mis- and disinformation given that they can easily be employed at scale, it is unclear how large the impact of AI-generated news stories is on the public. This paper investigates in an online experiment how persuasive AI-generated news stories are compared to human-generated news stories in inspiring individuals to take action against climate change and does so for four different emotion frames: hope, threat, anger and sadness framings. We use news stories from the Climate Connections podcast and generate synthetic news stories with the same title and length using GPT-4 to facilitate the comparison. Insights from this study will contribute to a better understanding of what types of news stories may or may not persuade individuals to take climate action and help to shape the policy debate of whether and how AI-generated news should be regulated. The research will also inform news organisations regarding where AI should (not) be used given its potential persuasiveness and how messages can be framed to nudge people to tackle societal challenges such as climate change.