Daily functioning improves over the first four weeks of using Ash

A 2026 study (n=1,284) found new Ash users improved in life satisfaction, sleep, relationships, and activity over four weeks, with no rise in grandiosity.

Single-arm observational cohort study

University of North Carolina at Charlotte, University of California, San Francisco

How four weeks with Ash relates to daily functioning

Key Finding

Over their first four weeks using Ash, 1,284 new users showed significant improvements in life satisfaction, relationship satisfaction, sleep quality, and behavioral activation — with no rise in grandiosity — and how consistently people returned to Ash predicted those gains more than how many messages they sent.

Summary

A 2026 preprint (n=1,284) followed new users of Ash, Slingshot AI's purpose-built conversational AI for mental health, across their first four weeks of real-world use. In this single-arm observational cohort study, participants completed in-app single-item measures of life satisfaction, relationship satisfaction, sleep quality, behavioral activation, working alliance, and grandiosity at baseline and Week 4. All four functioning domains and working alliance improved significantly (all ps < .001), while grandiose self-perception was unchanged. Consistency of engagement — active days, total sessions, and session minutes — predicted Week 4 functioning, whereas the sheer volume of user messages did not. The authors frame this as preliminary evidence that purpose-built AI can support broad day-to-day functioning, not just clinical symptoms.

The Full Picture

Gains were modest in size but consistent across domains: sleep quality showed the largest average improvement (14.69%; d=0.26), followed by relationship satisfaction (11.63%; d=0.25), life satisfaction (10.16%; d=0.20), working alliance (7.73%; d=0.20), and behavioral activation (5.58%; d=0.14); grandiosity did not change (d=-0.02, p=.477). Over 28 days participants averaged 17.72 active days, 45.93 sessions, and 3,185 session minutes. In ANCOVAs controlling for baseline, active days, total sessions, and total session minutes each positively predicted Week 4 functioning (all ps <= .006; partial R2 ~ 0.58-1.15%), and all four engagement metrics predicted working alliance (partial R2 ~ 1.63-2.15%), but total user messages predicted no functioning outcome. Baseline scores described a distressed sample — life satisfaction 4.96/10 (below the 6.7 cross-national average) and sleep quality 4.38/10. As a single-arm, self-report study with no control group, the changes cannot be attributed to Ash and may partly reflect regression to the mean or natural recovery, and because completing the Week 4 survey was required, the sample skews toward more engaged users and is not representative of typical use.

Van Swearingen, K. M., Hull, T. D., Sarma, K. V., & Stamatis, C. A. (2026). Functional outcomes and naturalistic engagement with a purpose-built conversational AI for mental health (Ash) [Preprint].

Begin your journey

Take the first step today

GET IN TOUCH

support@talktoash.com

press@slingshotai.com

ACKNOWLEDGMENT

Ash is not designed to be used in crisis. If you are in crisis, please seek out professional help, or a crisis line. You can find resources at www.findahelpline.com.