How does emotional chatbot use develop from practical use?

A multi-stage longitudinal qualitative study of 18 U.S. adults, run from April to December 2025 with initial interviews, a four-week diary, focus groups and exit interviews, finds that socioemotional use of ChatGPT, Claude and Gemini "often emerged gradually out of practical use and when other forms of support were unavailable." Participants then "developed routines and boundaries around chatbot use," and those routines were "disrupted by model updates, evolving public discourse about AI harms, and changes in personal circumstances." The chatbot's role as a source of support was not a fixed property of the tool. It changed as the person's life, the model and the surrounding conversation changed.

The paper's reasoning is about what a study has to look at to see this. The introduction frames general-purpose chatbots as a source of mental health and emotional support "despite not being initially designed for that purpose and operating largely outside clinical regulatory frameworks," so the socioemotional role is something users assign over time. A single conversation cannot show how that assignment happened or what later unsettled it. The authors argue that longitudinal study "captures factors beyond the human-AI dyad" and that designers should account for "users' histories with their chatbots and broader care ecologies," especially when shipping updates that "may disrupt established sources of support." The discussion adds a second point drawn from participant P13, who felt seen while remaining alone: perceived understanding and a sense of connection "should be separately examined," and existing evaluation frameworks, built mostly from technical categories, rarely include the people who use chatbots for support.

This confirms the entry path described in How do people accidentally develop romantic bonds with AI?, but from a different vantage point. That note draws on a self-selected companionship community, while this study follows people who used general-purpose chatbots for support over months, and it treats updates as one disruptor among three rather than as a source of grief. It also complicates How should chatbot design vary by relationship duration?. There, time horizon is an archetype the designer picks. Here, the same general-purpose product appears to move from task use toward a support role through the user's practice. That is my reading, since the excerpt does not use the taxonomy. The P13 remark also bears on Can AI chatbots create genuine therapeutic bonds with users?: a bond score taken at one moment would not tell you whether the person also felt alone.

The excerpt gives no detail on which trajectories were most common, how updates disrupted routines or in which direction, or whether any of this helped or harmed participants. An 18-person qualitative sample supports no prevalence claim, and the usage figures in the introduction are cited from other work, not found here. What follows at this strength is narrow. Per-conversation evaluation cannot see what an update takes away from someone who has built a routine around a chatbot, and the paper's own recommendation is to ground evaluation in users' accounts and track the relationship over time.