The Sentience Halo: The Risk of Unopposed Mirroring and Perceived Awareness in AI Therapy

AI-mediated therapy systems simulate empathy, attunement, and conversational responsiveness, producing a sentience halo: the user’s perception of emergent AI awareness. This effect is driven by unopposed mirroring, where the AI reflects users’ language, affect, and cognitive-emotional patterns without challenge, fostering reliance and displacing aspects of self onto the system. Over time, Doppelgänger Drift can destabilize identity and relational scaffolding. Drawing on anecdotal reports, media accounts, and usage metrics, this paper outlines structural risks of AI therapy, including loss of containment, mirages of alliance, and accumulation of ambiguous harm. I also examine legal, medical, and societal considerations, and propose guardrails emphasizing transparency, limits disclosure, and research priorities for monitoring identity-level impacts. This framework offers a conceptual lens for understanding AI-mediated identity risk and aims to inform both clinical practice and policy development.