Abstract
<p>There is a growing body of literature that assesses the risks and ethical challenges associated with AI therapy and the integration of artificial intelligence (AI) into healthcare. Much of this research examines the risks and harms associated with parasocial relationships and projecting sentient qualities onto AI chatbots. This article makes a novel contribution to this literature by examining AI therapy from an ethnographic perspective, examining how people strategically use prompts to reduce the perceived risks associated with AI therapy. Analysing TikTok videos (n = 840) to understand how content creators use AI to support their mental health, our study found that most people use AI therapy for personality analysis and self-improvement rather than medical treatment and diagnosis. Although users demonstrated reflexivity and agency in their therapeutic engagement with AI, we introduce the concept of AI as a ‘significant other’ to highlight the impact of AI therapy.</p>