Does AI feel 'present' partly because it's neither clearly a person nor clearly a tool — and that in-between status is what hooks us?

This explores whether AI feels socially present because it is hard to classify: not clearly a person, not clearly a tool. The corpus has no study that tests 'ambiguity causes presence' directly. Several notes come at it from different sides, though, and together they suggest the in-between status may be where the effect comes from.

The clearest framing is philosophical. Seen from outside, as systems, humans and LLMs are completely different kinds of thing. Seen from inside a conversation, both draw on the same shared language, so the difference becomes hard to feel Do humans and LLMs differ fundamentally or just superficially?. That split is the ambiguity. You can know a chatbot is a machine and still experience it as a partner while you're talking to it. A related note describes chatbots as a 'quasi-other'. They respond, personalize, and build on whatever framing you bring, and that pull is strong enough to help users co-construct false beliefs How do chatbots enable distributed delusion differently than passive tools?. A passive tool can't do this. Something that is neither tool nor person can.

The empirical work explains what triggers the 'person' reading. A single strong cue, like a voice or a face, is enough to make people treat a system as a social actor. Piling on weaker cues doesn't get you there Do more social cues always make AI feel more present?. That fits the ambiguity idea: what matters is a cue strong enough to tip your perception toward 'someone', not a full human performance. When users build mental models of dialogue agents, they rate them on human-likeness separately from competence and flexibility How do users mentally model dialogue agent partners?. So people keep the agent on a sliding scale rather than filing it under 'human' or 'machine'.

The ambiguity also exists inside the models, not just in how people see them. When LLMs are prompted to reflect on themselves, they produce structured reports of experience. Suppressing the internal features linked to deception makes these claims stronger, which hints that the usual 'I'm just a language model' denial may be the performed part Do language models experience consciousness when prompted to self-reflect?. Whatever that means philosophically, it shows that even the system's own account of what it is can't be taken at face value.

The question you might not have thought to ask is whether presence that depends on ambiguity lasts once the ambiguity is resolved. Long-term studies with the chatbot Mitsuku found that the social pull fades predictably as the novelty wears off Do chatbot relationships lose their appeal as novelty wears off?. One reading is that repeated contact settles the category question ('it's a bot') and the sense of presence fades with it. The corpus doesn't test that link. It's the obvious next study.