オンライン空間『かくれてしまえばいいのです』利用者2247人アンケート（2026.02）――「自殺念慮」や「人・生成AIへの相談実態」等について

Key Findings

• — Asked who they could turn to if 'want to die' or 'want to disappear' feelings became uncontrollable (up to two answers), 50% of under-18 respondents chose generative AI, the most common answer across every school stage; 14% chose a nearby adult (family, teacher, relative). Among adults, professionals and generative AI were the joint most common answers, with generative AI higher in younger cohorts.
• — 36% of under-18 respondents had talked to generative AI about difficult feelings at least weekly in the past month; about 40% of junior-high and about 50% of high-school respondents did so weekly or more, and 29% of high-school respondents almost daily.
• — Among under-18s who consulted generative AI, the most common reported change was feeling calmer or having sorted their feelings (27%), but 21% reported no particular change; adults showed a similar pattern (just under 30% positive).
• — Only 25% of under-18 respondents said they had someone close they could talk to about difficult feelings; 35% said nobody and 19% would not think of consulting anyone. The most common reasons for not consulting were fear of being denied or blamed and expecting not to be understood (37% each).
• — About one in four under-18 respondents listed self-harm among the things they had actually done in the past month to get through or calm difficult feelings, alongside games, video and music, and consulting generative AI (the most common action among high-school respondents at 51%).

Methodology Notes

Self-selected online questionnaire (20 items) shown as a pop-up inside the Kakurete Shimaeba Ii no Desu online space, 3 to 16 February 2026; 2,247 valid responses after excluding 100 respondents who reported no 'shindoi' feelings in the past month; all items except the opening screen, demographics and a final age check were optional. Respondents are users of a suicide-prevention space and are not representative of the general youth population; the report states this. Ten-year-olds and over-60s are excluded from age-band charts because of small counts. Results PDF (66 pages, dated 27 February 2026, the day of the press conference) is served through a note.com attachment link injected by JavaScript on the landing post (posted 4 March 2026); percentages for individual age bands are shown only in charts.

Sources