TL;DR: Muck Rack’s 2026 State of Journalism report surveyed 897 journalists through early March 2026 and found 82% use at least one AI tool — up from 77% last year. But concern about unchecked AI also rose, by 8 percentage points, to 26%. High adoption numbers rarely tell you what’s actually changing in the work. This one is no exception.

Key Takeaways

• 82% of journalists use at least one AI tool, up from 77% in 2025, based on 897 responses collected January 30 – March 2, 2026.
• ChatGPT remains the most commonly used tool among the surveyed group.
• Concern about unchecked AI rose 8 percentage points year-over-year to 26% of respondents.
• Only 21% of journalists say social media is very important to their work — down 12 percentage points since 2024.
• Disinformation and lack of funding tie as the top perceived threats to journalism, each cited by 32% of respondents.
• Two-thirds of newsrooms report AI efficiencies have not saved any jobs, according to the Reuters Institute’s parallel 2026 survey of 218 news leaders.

What 82% AI Adoption in Newsrooms Actually Looks Like

The number sounds like a verdict. 82% of journalists use AI — surveyed by Muck Rack across 897 respondents through the first week of March 2026. If you read the headline and move on, that’s the takeaway: adoption is mainstream, the debate is settled, figure out which tools to buy. That framing is not wrong, exactly. It just is not particularly useful for anyone deciding how to structure their team’s actual work around these tools.

I’ve been watching survey data on AI in newsrooms since 2023. The structure of the findings tends to be the same each year: more people using AI, similar uncertainty about outcomes, persistent concerns from a minority that keep growing. What’s different in this report is that the concern number moved meaningfully — up 8 percentage points in one year. That’s worth sitting with before moving to the tool recommendations.

What 82% Actually Means in Practice

At 82%, AI use is no longer a differentiator — it’s closer to a baseline. But the survey does not break down how frequently respondents use AI tools, for what tasks, or whether the tools are integrated into regular workflows or used occasionally for specific problems. ChatGPT is the most commonly used tool in the survey, which tracks with what I hear in conversations with editorial teams: it’s where most people start, often for research assistance, summarizing documents, or generating first-draft structures.

The limitation is that use of at least one AI tool covers everything from a single Perplexity search per week to Claude handling draft review before every publication cycle. Those are not the same thing. When newsrooms make resource decisions — which tools to pay for, which workflows to redesign — the headline adoption number does not tell them much. What matters is frequency, integration depth, and whether the output is going through substantive editorial review or being published after a quick scan. The report does not cover those questions. Most of them do not.

The Concern That Grew Alongside Adoption

26% of journalists say they’re concerned about unchecked AI — up 8 percentage points from 2025. In absolute terms, it’s still a minority position. But the direction of movement is unusual. Normally, as a technology becomes more familiar, anxieties about it level off or decline. Here, adoption is rising and concern is rising at the same time. The most plausible explanation is that more journalists are now using these tools long enough to notice the specific ways they fail.

AI transcription tools have gotten significantly better at converting audio to text — but they still struggle with multi-speaker meetings, accented speech, and domain-specific terminology. Summarization tools can compress 10,000 words to 500 reasonably well for straightforward documents, but they flatten nuance and occasionally introduce confident errors in technical content. Journalists who have been using these tools daily for six months know this from experience. The concern number going up likely reflects that growing familiarity, not panic about the abstract idea of AI.

What Journalists Are Actually Using AI For

Across the Muck Rack data and the Reuters Institute’s 2026 survey of 218 news leaders, AI use in newsrooms breaks into roughly three categories. Back-end automation — transcription, metadata generation, copyediting assistance — is the most widely adopted use case, cited by 64% of newsrooms in the Reuters Institute survey. Research assistance and document analysis are growing fast. Content generation for finished editorial work remains limited and heavily reviewed where it exists at all.

Google’s NotebookLM has picked up significant adoption in research-heavy newsrooms over the past year — it handles large volumes of mixed-format documents in a way that most chat interfaces do not. Deep research modes in ChatGPT, Claude, and Gemini have changed how some journalists approach initial source gathering for longer investigations. The tools saving time in 2026 tend to handle the preparatory and administrative layer of the work, not the editorial judgment layer. That is consistent with where the tools are genuinely reliable and where human oversight remains practical.

AI Tools for Editorial and Newsroom Work: A Practical Comparison

All five tools have improved meaningfully in the past twelve months. None of them are at the point where editorial output can skip human review without real risk.

Is Your Newsroom Ready to Prioritize New AI Tools?

• You have at least one person who can evaluate AI output against your editorial standards before it goes public
• You have identified 2-3 specific workflow bottlenecks AI could address — not just general efficiency
• Your team has time to test tools before committing to a subscription or integration
• You have a policy on AI disclosure to readers, even a draft one
• You have audited which AI tools your team is already using — at 82% adoption, people are probably using tools you have not formally reviewed
• You are prepared to revisit decisions in 6 months, not lock in a workflow for a year

When You Should NOT Invest in New AI Tools for Editorial Work

If your team is already stretched across existing tools and workflow changes, adding new AI tools rarely helps. The productivity benefit of most AI tools requires a period of regular use to understand where they are reliable and where they are not — and that period costs time you may not have. If there is no one available to absorb that learning curve and translate it into guidance the rest of the team can act on, the tool will likely be used inconsistently and quietly abandoned.

The same applies if your accuracy requirements are high and your editorial review capacity is thin. AI summarization and research tools produce errors confidently. In a newsroom where every output needs to be attribution-ready before publication, adding a tool that requires fact-checking of its own output can add steps rather than remove them. The efficiency case for AI is real in many contexts — but it depends on having enough review capacity to catch the cases where the tool is wrong, which happens more than the demos suggest.

FAQ

Is 82% adoption consistent across newsroom sizes and beat types?

The Muck Rack survey included 897 journalists primarily based in the US, with additional representation from the UK, Canada, and India. The report does not break down adoption by newsroom size or beat. Adoption rates likely vary significantly between large outlets with dedicated product teams and smaller local newsrooms with fewer resources for tool evaluation.

Which AI tool performs best for meeting transcript summaries?

Researchers testing LLMs on local government meeting transcripts found ChatGPT-4o delivered the most reliable summaries among tools tested. All tools underperformed against human benchmarks on longer summaries. For short summaries of structured content, ChatGPT-4o is a reasonable starting point — with the caveat that speaker attribution is still unreliable across all tools.

Does the Muck Rack report cover AI-generated content disclosure practices?

No — the report focuses on AI adoption levels and journalist concerns, not disclosure. Separate reporting from the Reuters Institute’s 2026 survey found that editorial transparency around AI use remains inconsistent across newsrooms, with few organizations having formal disclosure policies in place.

What does unchecked AI mean to the journalists who expressed concern?

The Muck Rack report does not define the term or probe what respondents mean by it specifically. Based on adjacent reporting, concerns cluster around AI-generated misinformation, AI replacing journalists without equivalent quality, and the absence of editorial oversight in AI-assisted publishing workflows.

Conclusion: Next Steps

The Muck Rack 2026 report is useful data, not a decision framework. 82% adoption tells you the tools have become part of the professional landscape. The 8-point jump in concern tells you that familiarity with these tools is also producing more specific skepticism — which is probably the healthiest thing in the report. The social media number is the one worth watching most carefully: only 21% of journalists say it is very important to their work now, down 12 points since 2024. If journalists are pulling back from social distribution, that has implications for where editorial resources go next that have nothing to do with AI.

If you are an editor or content lead trying to figure out what to actually do with this: read the Reuters Institute’s parallel 2026 analysis alongside the Muck Rack data. The two together give you a more complete picture than either does alone. Then look at what your team is already using — the 82% adoption rate suggests they are probably already using something, with or without a formal policy in place. That is the more urgent question for most editorial leaders right now.