Personalizing AI art boosts credit, not beauty

While image-generating artificial intelligence (AI) increasingly democratises art creation, people tend to devalue AI-generated content. Recent work suggests that human use of personalized AI models, trained on a user's past work, can increase credit attributions to the human user for achieving beneficial text-based outputs. We investigated whether this effect extends to visual artistic outputs and further examined the relationship between credit attribution and aesthetic appreciation. Across two studies (N = 774), UK participants evaluated identical paintings that were described as being created either by hand, with a standard text-to-image generative AI system, or with an AI system personalized to the artist. Personalization significantly improved both achievement credit and authorship attribution towards the human user compared to standard AI use. However, it failed to enhance either aesthetic appreciation of the image itself or willingness to categorize the output as "true art"—revealing a striking disconnect between judgments of artistic contribution and artistic value. Our findings suggest that although personalized AI may help bridge the "achievement gap" in credit attribution not only for written works, as demonstrated previously, but also for artistic visual images, it cannot overcome fundamental barriers to aesthetic appreciation of AI art. This challenges assumptions about the relationship between artistic input and aesthetic value, with implications for understanding art categorization and human-AI cooperation in creative pursuits.