Before an AI says a single word, chat bubbles, typing dots, and pauses already convince you 'someone' is there.

This is about the anthropomorphism that's already built in before the first reply: what the container, the setup and the conventions tell you about the thing you're about to talk to. The clearest answer in the corpus is that the chat interface does much of this work by itself. Turn-taking, artificial 'thinking' delays and messaging-app conventions (speech bubbles, typing indicators, a back-and-forth layout) all signal 'a someone is on the other end' before any text is generated How much does chat interface design contribute to anthropomorphism?. The practical point is that making the model's language less human-sounding won't fully fix anthropomorphism. The layout itself is part of the cause.

Why is this framing so powerful? One note argues that conversational design switches on skills people have built up over a lifetime of talking to other people. Once something looks like a conversation, users bring all their habits about intention, shared context and repair, even though the system isn't communicating the way a person does Why do users fail with AI interfaces designed like conversations?. So the trigger isn't a single feature like a name or an avatar. It's the genre of the interaction. A related critique questions whether chat should be the default at all. For many goal-directed tasks, clicking and direct manipulation would serve better, and choosing chat is a design decision with costs, not a neutral given Is chat the wrong default interface for AI products?.

A second, less visible layer sits in the setup the user never sees: the system prompt and alignment training. In Shanahan's framing, the dialogue prompt establishes a character, and the model then produces text consistent with that character. Folk psychology ('it wants', 'it believes') fits the character being played, not the underlying system Should we treat dialogue agents as role-playing characters?. Alignment training also fixes a single, stable communicative identity across all interactions Can language models adapt communication style to different contexts?. That consistency is exactly what makes something feel like a persistent 'someone'. The persona exists before the user types anything.

The more surprising takeaway is that anthropomorphism can be treated as a dial. One proposal places interfaces on a spectrum from anti-anthropomorphic to human-like. It suggests 'materiality' metaphors that reveal the data, infrastructure and human labor behind the model, as a counterweight to uncritical trust Can interface design reduce harmful anthropomorphism of language models?. Research on how users mentally model dialogue partners adds a useful caution. Perceived competence shapes impressions more than human-likeness does How do users mentally model dialogue agent partners?. So turning down the human cues doesn't have to make a system feel less capable.

A gap worth naming: the corpus doesn't directly study specific pre-conversation cues such as names, avatars, gendered voices or onboarding copy. What it covers well is the structural level: the chat genre, timing conventions and pre-set personas. If you want the strongest single doorway, start with the note on chat turn-taking and thinking delays.