CBT for psychosis

Evidence and outcomes

NICE recommends individualized CBT within care for psychosis and schizophrenia, delivered by appropriately skilled practitioners (NICE, 2014b, recommendations 1.3.4.1 and 1.3.7.1). A Cochrane review of first-episode and recent-onset psychosis found modest average advantages for overall symptoms and functioning when CBT was added to standard care. Certainty varied by outcome. Its relapse estimate crossed no difference, and evidence about adverse effects was very limited (Mayer et al., 2024). These distinctions prevent an overbroad statement that CBT reliably prevents psychotic relapse.

Sources: NICE, 2014b; Mayer et al., 2024

Applicability

The Cochrane population was predominantly young and early in the illness course, and the underlying search ended before publication. Its findings should not automatically be applied to all presentations, ages, or chronic illness. Specialist care also includes assessment, treatment coordination, and other interventions that a stand-alone reflective tool does not supply. Differences in usual care and therapist support affect what an added-treatment comparison means.

Sources: Mayer et al., 2024

Digital boundaries and research priorities

The reviewed evidence concerns specialist treatment alongside care, not an AI chatbot as a treatment for psychosis. A proposed evaluation of a conversational system would examine responses to unusual beliefs without affirming them as facts, routes to human support, and limits of unsupervised use. A clinical trial would need specialist partners, an appropriate risk protocol, and patient-important outcomes. These are proposed research priorities, not evidence that the current apps are safe or effective for psychosis.