Do “AI-assisted” labels reduce perceived quality of scenarios? An exploratory experimental study

Generative AI tools have been discussed in the foresight literature as promising aids for developing narrative scenarios. However, it is less known whether disclosing AI assistance affects how audiences evaluate scenarios. This matters because the perceived quality of scenarios directly influences their effectiveness, especially in policy contexts. This exploratory study examines the possible influence of an AI-assistance label on readers’ evaluations of scenario quality across several dimensions. A single original narrative scenario on the future of education in 2046 was developed and presented in an online experiment to 125 participants. Participants were randomly assigned to one of two groups: the control group, in which the scenario was presented as “developed by a group of experts”, or the experimental group, in which the same scenario was presented as “developed based on expert input with generative AI assistance (ChatGPT)”. After reading, participants rated the scenario on multiple dimensions of quality using 5-point Likert scales. These findings provide initial experimental evidence that labeling a scenario as “AI-assisted” reduces its perceived credibility, perceived expertise, and desirability, with small-to-medium effects ( r = − 0.32 to − 0.20). Moreover, the effect on perceived credibility was moderated by gender, with a stronger negative impact among female participants. This underscores the need for further research into how integrating generative AI into foresight processes affects audience evaluations of scenario quality.