["Journal of Applied Philosophy, EarlyView. ", "\nABSTRACT\nSelf‐conceptions provide a framework through which we can make sense of ourselves, interpret and navigate the world, plan our lives, and relate to others. Relational influences can greatly shape them, for instance, when others react to us or offer advice. What if this ‘other’ is not a human being, but an AI? AI companions (AICs) are dialogue systems that offer highly agreeable, low‐risk, and personalized ‘friendships’ or ‘romantic partnerships’ which can influence users' identities. This article identifies risks and opportunities AICs pose to users' practical interests regarding their self‐conceptions. Risks arise from overly positive or missing feedback, incoherence between belief and experience, lack of relational depth and empathy, third‐party dependence, and social stigma. At the same time, AICs may support identity interests by fostering self‐reflection, providing affirmation and enabling safe identity exploration. Based on this analysis, the article offers design recommendations to help align AIC development with users' identity needs.\n"]