Key points are not available for this paper at this time.
Abstract Policymakers and environmental advocacy organizations are increasingly using AI-generated images of climate disasters to advocate for climate interventions. Here we show that highly realistic AI-generated climate disaster images do not increase support for climate action. Three large-scale experiments ( N = 2580) provide evidence that these images intensify emotional responses but also elicit resistance in the form of reactance and reduced trust in the message source. These adverse effects are especially pronounced for highly realistic images that are suspected to be AI-generated or explicitly labeled as such. Moreover, when their AI origin is suspected but not disclosed, such images significantly reduce individuals’ willingness to make personal sacrifices for climate action. Overall, our findings call for an informed use of generative AI in climate advocacy that accounts for unintended effects and challenges the assumption that highly realistic AI-generated disaster images are inherently persuasive.
Bünzli et al. (Wed,) studied this question.