Can AI Models Be Used to Generate High‐Quality Pictorial Stimuli for Consumer Behavior Change Interventions?

As generative AI becomes increasingly embedded in marketing practice and consumer research, understanding its potential to support psychologically grounded behavior change interventions is both timely and essential. Traditional methods for creating experimental stimuli are time‐consuming and prone to bias, potentially undermining scientific rigor. This study investigates whether AI models generate high‐quality pictorial stimuli for interventions aiming to change consumer behavior, both as single stimuli and as groups with controlled variations. In an online experiment ( N  = 995), we compare AI‐ and human‐generated pictorial stimuli, evaluating AI models' effectiveness in conveying object and emotion accuracy (for happy), anthropomorphic qualities, and overall appeal. Results show AI‐generated pictorial stimuli outperform human‐created stimuli in object and emotional accuracy, anthropomorphism, and visual appeal. Notably, participants struggled to distinguish AI‐generated pictorial stimuli from those created by humans, highlighting the credibility and realism of AI outputs. ChatGPT 4o excelled in generating consistent stimuli with simple styles, though complex emotions like envy remain challenging to depict. These findings suggest that AI models offer a scalable, efficient, and psychologically valid alternative to traditional design methods for creating pictorial stimuli in marketing and behavioral research contexts.