British Gen Z trusts AI less than Boomers — here's the data

A new survey by The Turing Institute finds that British Gen Z (ages 18-24) trusts AI less than Boomers, with 74% supporting mandatory watermarking on AI-generated media and 61% backing licensing for frontier models. The longitudinal panel (n=4,200) reveals a 'trust decay' curve where trust drops 9 points after a user's first hallucination experience, and 58% of Gen Z worry AI will automate entry-level roles while only 19% have taken formal prompt engineering courses.

British Gen Z trusts AI less than Boomers — here's the data Three things stand out when you cross-tab the crosstabs. First, usage ≠ adoption. The 18-24 cohort treats ChatGPT /en/tags/chatgpt/ like a search engine with better UX — quick summaries, code snippets, essay scaffolding — but they've all been burned by hallucinated citations. One respondent in the qualitative follow-up put it: "I use it to start, never to finish." That's a workflow distinction most enterprise buyers miss when they pitch "AI assistants" to grad hires. Second, the regulation appetite is fierce. 74% of that same age band want mandatory watermarking on AI-generated media, and 61% back a licensing regime for frontier models — higher support than any other demographic. They've grown up with deepfakes in group chats; they don't need a white paper to understand provenance risk. Third, the skills anxiety is real but misdirected. 58% worry AI will automate entry-level roles in their field, yet only 19% have taken a structured prompt engineering or eval course. Most are self-taught via TikTok tutorials and Discord leaks. That gap — high anxiety, low formal upskilling — is where UK universities and bootcamps are failing. The Turing Institute's longitudinal panel n=4,200, weighted also tracks a "trust decay" curve: trust drops 9 points after a user's first personal hallucination experience, then plateaus. Means the first bad output does disproportionate reputational damage. For anyone building LLM products aimed at students or junior devs, that's your activation metric — not MAU, not token volume, but first-hallucination latency . What's missing from the discourse: nobody's measuring how this cohort evaluates agentic workflows versus single-turn chat. The surveys still ask "do you trust AI?" like it's a monolith. My guess? They'll trust a coded agent with a visible reasoning trace and rollback button before they trust a chat interface that hallucinates confidently. If you're hiring juniors in London or Manchester right now, the signal isn't "they know AI." It's "they know where AI fails." Berkeley Law just banned AI by default across every classroom 16h ago /en/news/7131/ Twitch is now using your stream data to train generative AI by 8d ago /en/news/6161/ Generative AI is basically the Guitar Hero of the creative world 14d ago /en/news/5421/ AI-Designed Bacteriophages: Engineering a Novel E. coli Killer 14d ago /en/news/5330/ The boos that greeted the AI-assisted staging at this year's 17d ago /en/news/4905/ Paying artists for AI training: enough to change minds? 18d ago /en/news/4840/ Next The consciousness debate feels like a distraction tactic → /en/news/7184/