Efficacy of a web application as a self-help intervention for nightmares: a randomized controlled trial

A web-based self-help application called DreamMend, which uses an AI-powered chatbot to deliver imagery rehearsal therapy, reduced nightmare frequency and distress more than a waitlist control over 8 weeks. Nightmare frequency decreased in the intervention group while remaining constant in the control group, and distress decreased significantly more in the intervention group. The findings suggest that digital, AI-supported imagery rehearsal therapy is a feasible and effective self-help intervention for nightmares, though larger and more diverse trials are needed.

Study at a glance

Abstract

Abstract Background Approximately 5% of people suffer from nightmares at least every 2 weeks. One cognitive behavioral intervention for recurrent nightmares is imagery rehearsal therapy (IRT), in which individuals recall, modify, and rehearse a less distressing version of their dream. Objective This randomized controlled trial evaluated the efficacy of a web application called “DreamMend” as a self-help method for reducing nightmares and their associated burden. It delivers digital IRT via an artificial intelligence (AI)-powered chatbot. Materials and methods A total of 42 participants were randomized to the intervention ( n = 18) or control group ( n = 24). The intervention group received access to DreamMend for 8 weeks. Analyses were conducted in 34 participants with available outcome data (intervention: n = 16; control: n = 18). Results Nightmare frequency in the intervention group decreased over the course of 8 weeks, while it remained constant in the control group. Linear mixed models (LMMs) revealed a statistically significant time × group interaction, showing that the intervention group profits from the intervention. Nightmare distress decreased in both groups over the 8 weeks. The LMMs revealed a significant time × group interaction, indicating that the intervention group experienced a significantly more marked decrease in distress. Conclusion The DreamMend web-based application significantly reduced nightmare frequency, distress, and overall nightmare burden compared with a waitlist control, even with variable adherence and minimal intervention exposure. These findings demonstrate that AI-supported digital IRT is a feasible and effective self-help intervention. Larger, more diverse trials are needed to confirm efficacy and examine long-term outcomes.