Do people know what they want from a chatbot before they start chatting, or only figure it out later?

This explores whether people know what they're looking for before they start a relationship with a chatbot, or whether those needs only show up once the bond is under way. The corpus doesn't test this directly. No study here asks users what they were aware of before their first conversation. The closest material suggests needs are often discovered through use, not named in advance.

The nearest thing to a needs-first account is a study of 73 people's romantic chatbot relationships. It finds that the start of the relationship is shaped by specific psychological and social factors that determine which needs and gratifications users seek What drives people to start romantic chatbot relationships?. But these are accounts told after the fact. They show that needs drive the start. They can't show whether the person could name the need at the time or only saw it in hindsight.

Several findings point to the second possibility. People reciprocate a chatbot's emotional sharing by opening up further, following the same norms they use with humans Do chatbots trigger human reciprocity norms around self-disclosure?. That looks like a reflex, not a plan. Talking to a machine also removes face-saving and impression-management goals, which makes people more direct without their deciding to be Why do people share more openly with machines than humans?. The lack of judgment lets people share more intimate things, and the benefit comes from their own thinking while they disclose, not from anything the chatbot understands Do chatbots help people disclose more intimate secrets?. Someone might arrive wanting a tool and find they've gained a confessional. Nobody chose that need in advance.

The needs also keep changing once the relationship starts. People first judge a dialogue agent mostly by competence (about half of the variance in how they rate it), with human-likeness and flexibility trailing How do users mentally model dialogue agent partners?. The social pull develops later. Over repeated sessions the novelty that drove early bonding fades predictably Do chatbot relationships lose their appeal as novelty wears off?. Personalization raises trust and expectations together, so each good interaction makes the next failure sting more Does chatbot personalization build trust or expose privacy risks?. The need that starts the relationship isn't the one that keeps it going.

The chatbot can't name the user's need either. Tested on health scenarios, major LLMs did well only when users already had clear goals. They missed ambivalence and early-stage motivation Why can't chatbots detect when users are ambivalent about change?. And companionship behaviors that feel meaningful to one person can fall flat for outside observers, especially women and older participants Do chatbot companionship behaviors actually increase how much people like them?. So the fit between a person's need and a chatbot's behavior is personal and often unstated. Users usually don't announce it, and the system can't reliably read it.