Who benefits most from AI mental health companions?
The clearest answer comes from a large Japanese study of 14,721 adults, which found that AI companions (like Replika or Character.AI) were linked to higher well-being across three measures: life satisfaction, happiness, and sense of purpose [4]. The benefits were strongest for people who reported high loneliness — meaning those who feel socially isolated may get the biggest emotional lift from an AI companion. Interestingly, the study also found a U-shaped pattern: people with moderate levels of friend-based social support benefited most, while those with very high or very low social support saw smaller gains. This suggests AI companions work best as a supplement for people who have some social connections but still feel a gap.
A U.S. national survey of over 1,000 adolescents and young adults (ages 12–21) found that 19% had already used AI chatbots for mental health advice, and 91.7% of those users rated the advice as helpful [1]. Use was higher among females (twice as likely as males) and among 18–21-year-olds (3.6 times more likely than 12–14-year-olds). This age group is also more likely to have spoken with a doctor about mental health, suggesting they may be actively seeking support and finding AI a convenient first step. The same survey found that 63% of youth did not tell anyone they were using AI for mental health — so for those who are hesitant to talk to a person, AI offers a private, stigma-free option.
How do people actually use these tools — and what do they expect?
Real-world data from Headspace's AI companion (Ebb) shows strong engagement: among nearly 394,000 users, those who used an improved version (2.0) completed two sessions within a week at a rate of 51%, compared to 29% for the earlier version [2]. Users turned to the AI for in-the-moment support during stress or anxiety, often during morning routines, commutes, or winding down at night. Diary study participants saw the AI as a guide to navigate mental health resources and content, not as a replacement for human care. They emphasized wanting transparency about data safety, clinical guidelines, and clear limits — they wanted the AI to be a helpful tool, not a therapist.
A survey of 107 community members and 86 mental health professionals in Australia found that 28% of community members used AI for mental health, mainly for quick support (60%) and as a personal therapist (47%) [3]. While 77% found it beneficial, 47% also reported harms or concerns — including worries about reduced human connection, privacy, and misinformation. This dual finding reinforces that AI companions are most helpful when used as a supplement, not a substitute, and when users have realistic expectations about what the AI can and cannot do.
What are the risks and limitations?
The evidence also highlights important caveats. In the Australian survey, 51% of mental health professionals who used AI reported experiencing harms or concerns, and both professionals and community members expressed equal optimism and worry about AI's future in mental health [3]. The Headspace study explicitly built in safety mechanisms — like monitoring for overuse, detecting risk, and flagging crises — because general AI tools lack these safeguards [2]. A scale developed to measure attitudes toward AI mental health chatbots found that people weigh both technical advantages (like accessibility) and personal advantages (like feeling heard), but trust and fear are key factors [5].
The Japanese study, while showing benefits, was cross-sectional (a snapshot in time), so it cannot prove that AI companions cause better well-being — it could be that happier people are more likely to use them [4]. And the U.S. youth survey found that those who had already spoken to a doctor about mental health were more likely to use AI chatbots, suggesting AI may complement — not replace — professional care [1]. The bottom line: AI mental health companions show real promise for lonely individuals, young adults, and those with moderate social support, but they work best as part of a broader support system, not as a standalone solution.
About These Sources
This answer is built on 5 peer-reviewed studies — published from 2024 to 2026, 5 from 2024 or later, 3 in Q1 journals — selected as the most relevant from 5 studies that passed quality screening, drawn from 47 papers retrieved from a database of over 500 million.
Sources used in this answer
AI Chatbot Use and Disclosure for Mental Health Among US Adolescents and Young Adults.
In a nationally representative U.S. survey of 1,009 youth (ages 12–21), 19% had used AI chatbots for mental health advice; 91.7% found it helpful, and use was higher among females and 18–21-year-olds.
Real-World Use of a Mental Health AI Companion: Multiple Methods Study.
In a multiple-method study of Headspace's AI companion (Ebb) with nearly 394,000 users, the improved version (2.0) doubled two-session retention (51% vs 29%), and users saw it as a guide for in-the-moment support, not a replacement for human care.
Use of AI in Mental Health Care: Community and Mental Health Professionals Survey.
In an Australian survey of 107 community members and 86 mental health professionals, 28% of community members used AI for mental health (mainly for quick support and as a personal therapist), and while 77% found it beneficial, 47% reported harms or concerns.
AI companions and subjective well-being: Moderation by social connectedness and loneliness
In a cross-sectional study of 14,721 Japanese adults, AI companion use was linked to higher life satisfaction, happiness, and purpose; benefits were strongest among those with high loneliness and moderate friend-based social support.
Development and Validation of the Artificial Intelligence in Mental Health Scale: Application for AI Mental Health Chatbots
A validation study of a new scale (AIMHS) with 5 items found that attitudes toward AI mental health chatbots are shaped by both technical advantages (accessibility) and personal advantages (feeling heard), with trust and fear as key factors.