How many people are already using AI for emotional support?

Public First's global survey — more than 18,000 respondents across 15 countries — finds that personal, emotionally inflected use of AI is already mainstream, running well ahead of how comfortable the same population is handing AI practical authority. "1 in 5 said they have used an AI for emotional advice, and 1 in 6 that they have a relationship with an AI," the report states. Read against the survey's other findings — only 29% comfortable giving an agent bank-account access, and net sentiment in the US falling from +0.73 to +0.18 — the emotional-use figures describe a different axis of engagement: not trust in AI's competence or reliability, but a willingness to treat it as a confidant or relationship partner regardless of broader unease about the technology.

The report does not separate "emotional advice" from "relationship" as categories, nor does it define either term for respondents beyond the self-report question itself; the two figures (1 in 5, 1 in 6) come from the same survey block and should be read as two self-reported behaviors — use for emotional advice, and self-identified relationship — rather than one phenomenon measured twice. The report's own framing elsewhere, that "much of this remains within our control, and dependent on what choices we make," is applied to economic adoption decisions, not emotional ones; it does not extend that "choices matter" argument to the relationship finding, leaving the emotional-use data as a standalone descriptive statistic in this excerpt.

This sits in tension with How soon do AI researchers expect artificial general intelligence?: the experts surveyed there hold meaningful credence on extinction-level risk from the same technology that a sixth of ordinary respondents here already describe as a relationship partner. The two surveys measure different populations and different questions, so this is not a contradiction, but it is a juxtaposition worth holding onto — lay emotional reliance on AI appears to be scaling in parallel with, not behind, expert concern about AI's downside risk. It also pairs with the sibling note on this same survey, How willing are people to let AI control their finances?: emotional trust and financial trust move independently in this data, so a design or policy response calibrated to one should not assume it transfers to the other.

The excerpt gives no country-level breakdown, no demographic split, and no trend line, so it cannot say whether emotional reliance on AI is rising, who it concentrates among, or what "a relationship with an AI" meant to respondents answering the question. The implication the evidence supports is modest but real: a meaningful minority of the surveyed population — not a fringe — already relates to AI in personal terms, and any product or policy conversation about AI companionship should treat that as a present-tense, double-digit phenomenon rather than a speculative future one.