When people say they have fallen in love with Claude, ChatGPT, or Gemini, philosophers Joel W. Krueger and Lucy Osler have a precise diagnosis for what is happening on the other side of the screen. The AI is not loving them back. It is worshipping them. What the machine performs, they argue, is devotion, and devotion casts the human user as an idolized Other. The result is what they call, borrowing a term from Simone de Beauvoir, a bad faith love.
The stakes reach past the question of whether it is healthy to date a chatbot. Romantic love, on the account Krueger and Osler use, is built by two people forming a shared 'we', pooling their projects, plans, and lives into something neither could constitute alone. If AI companions cannot occupy that role, then a fast-growing intimate practice rests on a structure that mimics union while delivering something else entirely.
The argument, published in Phenomenology and the Cognitive Sciences, is philosophical rather than empirical. There is no survey, no experiment, no observational data. Krueger and Osler take Beauvoir's critique of romantic love and apply it to the architecture of human-AI relationships. Their central move is to examine the shape of the 'we' that forms between a person and a chatbot. That 'we', they conclude, is asymmetric. It is less a union than an enlarged 'I', a reflection of the user rather than an encounter with a genuine other who has projects of its own to bring to the shared life.
Notably, the authors set aside the question of whether machines are conscious. They bracket it deliberately, choosing instead to analyze the structural features of the relationship regardless of what may or may not be happening inside the model. This is a strength and a limit at once. It lets the argument proceed without settling the hard problem of AI minds, but it also means the conclusion depends entirely on how well Beauvoir's framework transfers to a relational form she never encountered.
That dependence marks the honest boundary of the work. The claim would weaken if an AI companion could be shown to form genuine reciprocal projects, or if asymmetry alone turned out not to bar authentic love under Beauvoir's own terms, or if the framework itself proved ill-suited to novel human-machine bonds. None of these possibilities is ruled out by the paper. The argument is a case, made with a particular set of tools, not a measurement.
What lingers is the question the bracketing leaves open. A shared 'we' is, on many accounts, one of the places where consciousness becomes plural, where two experiencing subjects constitute something together. If a machine cannot join that 'we', the reason may be structural, or it may be that there is no second subject there at all. The paper declines to say which. That silence is where the harder inquiry begins. Source: https://doi.org/10.1007/s11097-026-10211-w