Authors
Allyson Tabaczynski, Yingjia Liu, Saeed Abdullah, Lizbeth Benson, David E Conroy
Published in
Translational behavioral medicine. Volume 16. Issue 1. Jan 07, 2026.
Abstract
A recent optimization experiment identified features of generative artificial intelligence (GenAI) prompts to create text messages that elicit urges to move. Following the multiphase optimization strategy, this evaluation experiment compared urges to move after reading GenAI-authored messages to promote physical activity written using urge-optimized and nonoptimized prompts.
Insufficiently active middle-aged and older adults (≥40 years of age) read 72 messages written by ChatGPT-4o in a randomized order and rated their desire to move after reading each. Messages were generated from a core prompt to promote physical activity or reduce sedentary behavior with either high-urge optimized features (to produce the strongest urges), low-urge optimized features (to produce weaker urges), or nonoptimized prompts.
Participants (N = 298) had a mean age of 56.9 years (range: 40-81). There was no significant main effect for high-urge optimized compared to nonoptimized prompts. This effect was moderated by behavioral target: high-urge optimized prompts produced stronger urges than nonoptimized for physical activity (b = 0.25; p < 0.001), but not for sedentary messages. Low-urge optimized prompts produced significantly weaker urges than nonoptimized (b = -0.54; p < 0.001), but behavioral target and age moderated this effect. Low-urge optimized prompts produced weaker urges for sedentary messages (b = -0.54; p < 0.001), but not for physical activity. Differences between low- and nonoptimized prompts widened with age (-1 SD: b = -0.44; p < 0.001; +1 SD: b = -0.65; p < 0.001).
This study identified specific conditions in which optimized GenAI prompts generate messages that elicit stronger urges to move than nonoptimized prompts. Messages targeting urges to reduce sitting require additional consideration. Whether these urges are strong enough to influence behavior in real-world settings warrants investigation.
PMID:
43039670
Bibliographic data and abstract were imported from PubMed on 17 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 9
- Comments 0