Generative debunking of climate myths using AI

October 6, 2025

Christian Downie’s forthcoming book, Climate Obstruction, taking a global view, will complement Marian Wilkinson’s The Carbon Club (2020), which exposed the powerful forces shaping Australia’s poor response to climate change. The unrelenting misinformation has been effective.

A 2020 University of Canberra study, cited at the current Senate inquiry into Information Integrity on Climate Change and Energy, found that Australia ranks third in the world for climate denialism — 8% of the population — behind the US and Sweden. The global average is just 3%.

While Downie is right to highlight the role of PR firms spreading climate misinformation for fossil fuel clients, the Stockholm Resilience Centre warns that generative AI could create a “perfect storm” of climate falsehoods.

Yet AI might also be part of the solution: researchers at the University of Melbourne are successfully training large language models that take a climate myth as input and generate a debunking using the fact-myth-fallacy-fact (“truth sandwich”) structure. In the battle against climate misinformation, this kind of innovation could prove vital – using the same powerful technology that spreads falsehoods to instead reinforce scientific truth.