Empirically validate operational criteria for AI-associated psychosis

Determine whether the phenomenological patterns reported in cases of AI-associated psychosis are sufficiently consistent, specific, and stable to support operational diagnostic criteria.

Background

The paper argues that “AI psychosis” currently lacks a sufficiently precise clinical definition and risks being diluted into general online slang or hyperbole. The patterns summarized in the paper are drawn largely from anecdotal media reports and case reports rather than systematic clinical data.

The authors identify the consistency, specificity, and stability of these patterns as an unresolved empirical issue because those properties are prerequisites for reliable case ascertainment, surveillance, and research. Establishing whether the patterns can ground operational criteria would help determine whether AI-associated psychosis should be treated as a distinct clinical syndrome or instead as a context or precipitating factor within existing diagnostic categories.

References

We stress that the final column is a descriptive synthesis of the current (largely anecdotal) case literature, not a proposal for what the diagnostic criteria of AI-associated psychosis should be; whether these patterns are consistent, specific, or stable enough to ground operational criteria is precisely the empirical question that remains open.

— An Echo Chamber of One: Should AI Psychosis Be a Distinct Clinical Entity?  (2608.23937 - Yeung et al., 25 Aug 2026) in Section Discussion and Future Directions, subsection “Conceptual Clarification: General Slang vs. Clinical Syndrome”

Only by accumulating and sharing on-the-ground clinical reports of these cases with the wider community can we shed light on key questions, including: What is the incidence and severity of AI-associated psychosis? What are the different phenotypes? What are the predisposing, precipitating and perpetuating factors? Can AI-associated psychosis be triggered by chatbots in a person with no prior mental health history? Does it warrant its own diagnostic criteria?

— An Echo Chamber of One: Should AI Psychosis Be a Distinct Clinical Entity?  (2608.23937 - Yeung et al., 25 Aug 2026) in Section Discussion and Future Directions, subsection “Exacerbation vs. De Novo Psychosis”