Effectiveness of AI-Guided Meditation: EEG Analysis and User Experiences
Wen Xiao, Hui-Wen Huang, Daniel G. Dusza
2025 8th International Conference on Robotic Systems and Applications September 19, 2025 DOI: 10.1109/icrsa66467.2025.11383824 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Randomized controlled trial Qualitative Peer reviewed |
|---|---|
| Intervention | AI-guided meditation |
| Measures | electroencephalography (EEG), post-session interviews |
| Topics | Meditation |
| Key findings | EEG showed significant within-session changes across 5-minute meditation phases, but no statistically significant between-group differences in alpha-theta cross-frequency dynamics for AI-guided versus instructor-provided text-based meditation. Interviews suggest AI-guided meditation redirects internal attention and promotes relaxation, though some participants found the AI chatbot's tone less emotionally engaging than human-led or video-based meditation. |
Abstract
This study investigates the effectiveness of AIguided meditation compared to instructor-provided, text-based meditation, using electroencephalography (EEG) and post-session interviews. Participants were randomly assigned to either condition, and EEG data were analyzed to assess alpha-theta cross-frequency dynamics, a biofeedback marker of meditative states versus mind-wandering. Although EEG analysis revealed significant within-session changes across the $\mathbf{5}$-minute meditation phases, no statistically significant differences in these neural measures were observed between the groups. Qualitative interview data suggest that AI-guided meditation effectively redirects internal attention and promotes relaxation; however, some participants perceived the AI chatbot's tone as less emotionally engaging than that of human-led or video-based mediation. The findings suggest that AI-guided meditation is a scalable and promising tool for modulating neural correlates of emotion regulation. However, further refinements in AIsupported voice models should enhance user engagement.