A Pilot Two-Arm Randomized Controlled Trial Comparing an AI Mental Health Tool and Psychoeducational Control

Slingshot AI is the developer of Ash, a conversational AI mental health tool designed to provide accessible, evidence-informed support for adults experiencing common mental health concerns such as low mood, anxiety, stress, and sleep difficulties. Common mental health concerns are highly prevalent in the United States, and access to traditional psychotherapy remains constrained by cost, geography, and clinician supply. A central open question is how much of any observed benefit from a tool like Ash is attributable to its personalized, conversational, and adaptive properties, as opposed to the psychoeducational content and the structure, cadence, and expectancy that any well-designed digital program provides. Psychoeducation alone has documented effects on depressive symptoms, so a psychoeducational comparator delivered in the same interface, at the same cadence, and with the same reminder schedule isolates the contribution of personalized conversational support.

This protocol establishes a pilot, two-arm, parallel-group, open-label, superiority randomized controlled trial comparing two conditions over a 4-week period in adults with moderate or higher symptoms of depression (PHQ-9 total score >= 10 at screening): full access to Ash, including conversational chat, memory, and personalized content; and a psychoeducational control consisting of 12 static psychoeducational modules released three times per week within the same Ash application, without conversational chat or personalization. Participants randomized to the Ash arm receive 9 months of free Ash access beginning at randomization. Participants randomized to the Psychoeducation arm receive the 12 modules over the first 4 weeks, with Ash conversational chat deferred until after the Week 4 primary endpoint, at which point their 9 months of free Ash access begins. Both arms receive an identical cadence of reminders and notifications. This deferral preserves a clean comparison of personalized conversational support versus static psychoeducational content during the intervention window while ensuring that every participant ultimately receives the same study benefit.

Target enrollment is approximately 500 adults, randomized 1:1 to Ash (approximately 250) and Psychoeducation (approximately 250). The SAP pre-specifies a blinded sample size re-estimation at a fixed interim, based on pooled (arm-blinded) variance and observed retention. Eligible participants are new Ash users with active accounts, aged 18 or older, residents of the United States, English-speaking, and able to provide informed consent, among other eligibility criteria; individuals currently engaged in individual therapy or higher-level mental health treatment, and those with severe psychiatric presentations (e.g., positive mania screen, positive psychosis screen, frequent heavy substance use, suicidality, or recent psychiatric medication changes), are excluded.

Primary aim: Evaluate whether Ash produces greater reduction in depressive symptoms (PHQ-9) than the psychoeducational control from baseline (Week 0) to Week 4. The between-arm difference in PHQ-9 change (Ash vs. Psychoeducation) will be estimated with a two-sided 95% confidence interval and tested for superiority at a two-sided alpha of 0.05 in the intention-to-treat population, as pre-specified in the preregistration and statistical analysis plan (SAP). Within-arm change from baseline to Week 4 will also be estimated for each arm.

Secondary aims: Estimate within-arm change and the between-arm difference in anxiety symptoms (GAD-7), functional impairment (WPAI-GH), global health (PROMIS Global Health), sleep quality (SQS), behavioral activation (BADS-AC), perceived stress (PSS-4), worry (PSWQ-3), and PTSD symptoms (PC-PTSD-5); and evaluate the feasibility of recruitment, in-app consent, randomization, retention, survey completion, and psychoeducational module engagement to inform the design and powering of a future definitive trial.