Awarefy Survey Finds Generative AI Is Becoming a Familiar Source of Support, Not Just an Information Tool

Awarefy and its Mental Health Research Institute announced the results of a February 2026 survey on the relationship between conversational generative AI and people, conducted among about 1,000 residents of Japan who had used conversational generative AI. The findings suggest that as usage frequency and demand for consultation increase, AI is moving beyond its role as an information-gathering tool and is increasingly becoming a familiar source of support for users. Compared with Awarefy's previous survey in August 2025, the share of people using AI every day rose from 21.2% to 30.9%, while 61.9% said their usage frequency or time had increased compared with six months earlier. The share of respondents using AI at least once a week reached 86.5%. Most users, 82.5%, use free personal plans, suggesting that AI is becoming established as an everyday partner that does not add to living costs. Information-related needs remain the most common reasons for using AI, including wanting AI to teach users things they do not know (79.5%) and help generate ideas (65.0%). However, needs related to emotional and psychological support are expanding notably. The share of respondents who want AI to listen to their concerns rose from 33.0% to 42.8%, those who want AI to be a conversation partner rose from 18.1% to 24.7%, and those who want AI to serve as emotional support rose from 11.6% to 18.5%. Around 12% of those seeking emotional support expect AI to act like a psychological counselor, indicating that a certain segment expects advanced professional-style support from AI. The survey found that AI's appeal is not limited to its knowledge and research capabilities. Respondents also valued being able to speak without hesitation because the other party is AI, and being able to talk anytime and as much as they want. Some users reported positive changes, such as becoming better able to put negative feelings into words, express their thoughts, and feel a sense of relief. Others reported negative changes, including becoming more suspicious, feeling irritated by AI responses, or accepting AI-generated search results without checking sources. When asked who they can casually consult, 87.9% selected conversational generative AI, ahead of close friends, mothers, and spouses. AI led the second-ranked option, close friends, by more than 40 percentage points, and the gap has widened by more than five points since the previous survey. However, the researchers note that AI consultation does not currently appear to be replacing or obstructing consultation with people. The ease of consulting other people has changed little over the past six months, suggesting that AI is better understood as a new support resource rather than a substitute for human relationships. From a mental health perspective, notable changes are emerging as AI use becomes more routine. The share of respondents who would feel moderate or greater anxiety if conversational generative AI became unavailable tomorrow rose from 43.7% to 49.4%. Those who had felt they might be dependent on AI rose from 27.5% to 37.1%. Among people using AI at least once a week, 41.9% reported moments of possible dependence, compared with 5.7% among those using it less than once a week. Examples included wanting to talk to AI first when something unpleasant happens, consulting AI before thinking through a problem, and continuing to interact with AI until feeling calm when overwhelmed. The research team says the feeling of possibly being dependent on AI should not automatically be read as negative. It may also indicate that users are aware of changes in their own behavior and usage patterns. Further research is needed to determine whether this awareness is a healthy sign of self-monitoring or a risk that could interfere with daily life and health. The institute said it will continue research and information-sharing to help more people use AI appropriately, emphasizing the view that mental health is protected not by choosing between AI and people, but by having diverse sources of support.