AI-generated mindfulness exercises: The role of attitudes towards AI and expected effects on acceptance and effectiveness

• Subjective effectiveness and acceptance of AI-generated mindfulness exercises are increased when exercises are presumed to be human-created • Attitudes towards AI predicts effectiveness and acceptance of AI-generated mindfulness exercises • Expected effects mediate the effect of attitudes towards AI on effectiveness and acceptance Mindfulness exercises can improve psychopathology and mental well-being. Generative Artificial intelligence (AI) allows the flexible creation of tailored and individualized mindfulness exercises. The present study investigates the feasibility (acceptance, subjective, mindfulness) of AI-generated mindfulness exercises depending on participants’ presumption on whether the exercise is human or AI, expectancy effects, and attitudes towards AI. In an online experiment, 411 participants conducted four AI-generated mindfulness exercises and were told that exercises were either human or AI-generated. Exercises were assessed on acceptance, mindfulness, and subjective expected and experienced effectiveness. Linear mixed models and ANOVAs were used to investigate the effect of the manipulation and number of presumed human mindfulness exercises on measured variables. Results show that when participants were told exercises were human (vs AI-generated), acceptance ( η ² p = .10), experienced effects ( η ² p = .34), and mindfulness ( η ² p = .09) of were significantly higher. The number of mindfulness exercises thought to be human significantly predicted acceptance ( η ² p = .08), experienced effects ( η ² p = .10), and mindfulness ( η ² p = .02). Attitudes towards AI significantly predicted acceptance, experienced effects, and mindfulness scores, partially mediated by expected effects. Belief that exercises are human improved ratings, predicted by attitudes towards AI and expected effects. Negative attitudes and expected effects may dampen acceptance and effectivity. The implementation of AI-enhanced tools may be hindered by pre-existing attitudes which acceptance-enhancing interventions may counteract. AI-generated mindfulness exercises can be adaptable tools for mental healthcare. Future research may investigate the use of AI-generated mindfulness exercises in clinical trials.