Abstract
<sec> <title>BACKGROUND</title> <p>Digital peer support (DPS) services offer scalable mental health support. Including psychoeducational (PE) resources according to users’ readiness for change based on the Transtheoretical Model (TTM) may increase emotional benefits.</p> </sec> <sec> <title>OBJECTIVE</title> <p>Investigate the association between PE provision, delivered by AI-supported human moderators and adapted to the TTM stage, and DPS users’ sentiment scores based on expressed language.</p> </sec> <sec> <title>METHODS</title> <p>28,718 Supportiv DPS live chats from 17,237 distinct users (January-December 2024) were analyzed. Users were classified into non-exclusive TTM stages, and PE resource distributions were examined. Statistical analyses assessed changes observed in language-based sentiment scores.</p> </sec> <sec> <title>RESULTS</title> <p>The most common TTM stages were contemplation (n=26,594), preparation (n=8,296), and precontemplation (n=1,515). DPS use was associated with decreased negative sentiment, with greater changes when resources were distributed (p<.0001): anger (reduced by 48.09% without versus 57.48% with resources, Δ9.39%, rrb=.227), loneliness (38.72% versus 47.81%, Δ9.09%, rrb=.227), sadness (35.72% versus 44.23%, Δ8.51%, rrb=.219), depression (36.74% versus 44.48%, Δ7.74%, rrb=.227), anxiety (35.41% versus 41.28%, Δ5.87%, rrb=.216), stress (35.96% versus 44.11%, Δ8.15%, rrb=.209), and optimism increased (68.78% versus 110.10%, Δ41.32%, rrb=.245). Larger anger reductions in teens (p=.0368) and increased optimism in older adults (p=.0089) were observed.</p> </sec> <sec> <title>CONCLUSIONS</title> <p>DPS is associated with significant changes in expressed sentiment, and this association is greater when TTM-stage-tailored PE resources are provided. AI-supported human moderation of DPS may allow scalable and stage-tailored support across diverse populations.</p> </sec>