Zwischen Fortnite, TikTok und ChatGPT: Mediennutzung, Risiken und neue Nutzungstrends bei Kindern und Jugendlichen in Deutschland (Mediensuchtstudie OTTER, Ergebnisbericht 2025/2026)

Key Findings

• — 38.8% of 10 to 17 year olds used generative-AI chatbots at least weekly and 27.2% several times a week to daily; 33.3% had not used one in the past six months. Weekly-or-more use rose from 12.1% at age 10 to 58.7% at age 15 and stayed at that level.
• — Among users (N = 644) ChatGPT had been tried by 87.7%, Meta AI (WhatsApp, Instagram, Facebook) by 34.6% and Google Gemini by 33.4%; homework and school help was the most common motive, followed by learning or finding information.
• — 70% of users at least sometimes trust the chatbot's statements and 40.7% trust them often or very often; 10.9% say the chatbot understands them better than a real person; 10.4% tell it things they would tell nobody or only close friends; 9.8% use it to distract from negative feelings; 7.8% to feel less lonely; 3.4% miss the chatbot when not using it (Table 5).
• — Users in the top quartile of the ten-item bonding score (n = 135) had depression (PHQ-9), anxiety (GAD-2) and stress (PSS-4) levels about 0.6 standard deviations above the mean of all respondents, while low and medium bonding groups sat 0.15 to 0.25 SD below it; users with at least mild depressive symptoms (PHQ-9 of 10 or more, n = 73) more often used chatbots to reduce loneliness, distract from negative feelings and share things they tell nobody else.
• — 22.8% of users at least rarely share sensitive personal information (full name, address, photos of themselves or others) with chatbots (19.5% of 10 to 13 year olds, 24.8% of 14 to 17 year olds).

Methodology Notes

Online survey by forsa Politik- und Sozialforschung of 1,250 parent-child dyads (representative sample of 1,005 children aged 10 to 17 after excluding panel members who had aged out), fieldwork September and October 2025, weighted by region, sex and age to Federal Statistical Office figures; longitudinal panel with refreshment since wave 1 (September 2019). Chatbot bonding measured with ten newly developed items (trust and emotional bonding), summed and split into quartiles; associations with PHQ-9, GAD-2 and PSS-4 are descriptive, not adjusted. 28-page German-language results report; PDF created 19 March 2026 (DAK press release 24 March 2026). The gKI-chatbot base is users who did not answer 'never' (N = 644).

Sources