Will AI Replace Journalists? The 2026 Risk Picture
Journalists score 7.0/10 on AI exposure - a significant risk score for a profession that historically resisted automation by claiming complexity as a moat. AP's Automated Insights system generates thousands of earnings and sports articles per month with no human writers. BLS OEWS May 2025: 44,000 US news analysts, reporters, and journalists at a median annual wage of $55,960. ILO ILOSTAT 2024: approximately 1.7 million workers in media and communications roles globally. The risk is real but uneven across journalism types.
Key findings
- ISCO 26 (journalists and media professionals) scores 7.0/10 on AI exposure
- AP's Automated Insights generates thousands of earnings, sports, and data articles monthly - no human writers
- US: 44k reporters and journalists at median $55,960 (BLS OEWS May 2025)
- Global: ~1.7M media and communications workers (ILO ILOSTAT 2024)
- Structured data journalism (earnings, scores, weather, local stats) is largely automatable
- Investigative journalism, source-led reporting, and foreign correspondence retain strong human requirements
- Robotics risk: 1.0/10 - purely a language AI threat
The AP model: AI at scale in newsrooms
Associated Press (AP) has used Automated Insights' Wordsmith platform to generate quarterly earnings reports since 2014 - one of the earliest documented large-scale deployments of AI in a major news organisation. By 2025, AP was generating several thousand automated articles per quarter in areas including corporate earnings, minor league baseball, and college sports. The articles are factually accurate and indistinguishable from basic human-written roundups of structured data.
Bloomberg uses Cyborg, its in-house AI system, to handle financial news automation. Reuters uses its own AI tools for similar structured data outputs. The common factor: all three organisations found that AI can produce accurate, publishable content faster and at lower cost than human reporters for any story that is primarily a transformation of structured data into prose. Quarterly revenue figures. Points scored. Weather station readings. Vote tallies.
The 7.0/10 score reflects this accurately: a substantial portion of total journalism output is structured data reporting. The remainder - analysis, investigation, source development, feature writing, editorial judgment - retains higher human requirements. But "substantial portion" means real job volume.
What AI is already doing in journalism
| Journalism task | Tool / approach | Status | AI risk |
|---|---|---|---|
| Earnings and financial summaries | Automated Insights (AP, Bloomberg) | Largely automated | 9.0/10 |
| Sports score recaps | AI narrative generation | Largely automated | 8.5/10 |
| Local government statistics | AI data-to-prose tools | Partially automated | 7.5/10 |
| Press release processing | LLM summary tools | Largely automated | 8.0/10 |
| Interview transcription and summary | Whisper, Otter.ai, Descript | Fully automated | 9.0/10 |
| Beat reporting (politics, courts) | AI-assisted research, human-led | Augmented | 5.5/10 |
| Investigative journalism | Data analysis tools assist; human-led | Low automation | 3.0/10 |
Why investigative journalism is safer
The defining skill of investigative journalism - source development - is not automatable with current AI. A source who leaks documents does not do so to an AI system. A whistleblower calls a journalist they trust, who they believe will understand the significance of what they are sharing and protect their identity. That relationship is built over months or years of beat reporting. AI cannot build it.
Document analysis is partially automatable: AI can now process and identify patterns in large document sets (Muckrock, Document Cloud, and custom LLM pipelines are used in investigative newsrooms for exactly this). But the judgment call - which documents are significant? what does this pattern mean? who do I need to talk to? - remains a human function. The Pentagon Papers required someone to recognise what they had and decide to act. AI would have indexed them and moved on.
Foreign correspondence is similar. A journalist in Kyiv, Kabul, or Kinshasa providing on-the-ground reporting cannot be replaced by an AI reading wire copy. The physical presence, the source relationships, the contextual understanding that comes from living in a place - these are not language model skills. Foreign correspondents face real financial pressure as media organisations cut costs, but that pressure is economic (budgets), not technological (AI capability).
The safest journalism roles from AI
| Role type | Why lower risk | AI score |
|---|---|---|
| Investigative journalist | Source development, document judgment, accountability reporting | 3.0/10 |
| Foreign correspondent | Physical presence, source relationships, contextual expertise | 3.5/10 |
| Political beat reporter | Source access, editorial judgment, off-record intelligence | 5.0/10 |
| Feature writer / long-form | Voice, narrative judgment, human interviews | 5.5/10 |
| Data / structured report writer | Core AI use case - earnings, scores, statistics | 9.0/10 |
What this means for you
If you write primarily data-driven or structured content - quarterly earnings roundups, sports recaps, local statistics, press release processing - the honest picture is that AI can produce equivalent output at near-zero marginal cost. Organisations that have not yet deployed automated writing tools for this work will over the next 24 months. The headcount implication is direct: these specific roles compress or disappear.
The skills that hold value are those AI cannot replicate: source relationships, editorial judgment, narrative voice for complex human-interest stories, and the physical and relational presence required for original reporting. A journalist with a deep source network in a specific beat - healthcare policy, financial regulation, criminal justice - is substantially more resilient than one who covers a mix of structured topics without deep source access.
The structural economic problem for journalism is separate from but compounded by AI: digital advertising revenue has already decimated newsroom headcount. AI accelerates this by further reducing the cost of content production for the commodity end of journalism. The surviving journalism market is bifurcated: automated commodity content at scale, and premium original reporting at higher cost. The middle is where most journalists currently sit, and it is the most exposed.
See how media workforce AI risk varies by country
Explore AI exposure across 206 countries including the media and communications sector.
Open the Explore Tool →Was this analysis useful?