Last updated: December 2025
Journalism is under pressure from every direction: shrinking budgets, faster news cycles, audiences that expect instant coverage. A Reuters Institute report found media leaders expect an additional 43% traffic decline over the next three years as AI summaries eat into click-through rates. AI doesn’t replace journalists (the judgment, ethics, and source relationships are irreplaceable), but it handles the mechanical work that eats up 60% of a reporter’s day.
Here’s what newsrooms are actually using, from local papers to national outlets.
Research
Good reporting starts with good research, and this is where AI saves the most time. I watched a reporter at a mid-size paper cut her background research from two hours to twenty minutes. The tools don’t replace journalistic instinct, but they eliminate the tedious parts.
Perplexity: The Reporter’s Best Friend
Perplexity replaced Google for most research tasks in newsrooms I visited. The sourced answers save the “open 15 tabs and piece together the story” workflow.
How journalists use it:
- “What are the key arguments in the current debate about AI regulation in the EU? Cite recent sources.”
- “Summarize the financial performance of [company] over the past 3 years.”
- “What has [public figure] said about [topic] in the past 6 months?”
Each answer comes with citations. Click through to verify. The research that took 2 hours now takes 20 minutes.
Academic mode searches scholarly papers, useful for science and health reporting where peer-reviewed sources matter.
Cost: Free → Pro $20/month
Claude: Deep Analysis
For longer research tasks, Claude handles document analysis that Perplexity can’t. Claude is particularly strong at working through dense material:
- Upload a 200-page government report → “Summarize the key findings and identify the most newsworthy elements”
- Upload financial filings → “What changed between this quarter and last quarter? Flag anything unusual”
- Upload court documents → “Explain the legal arguments in plain language”
Cost: Free → Pro $20/month
Transcription
Transcription used to be the most soul-crushing part of journalism. Hours of rewinding, typing, rewinding again. AI killed that workflow entirely, and I haven’t met a single journalist who misses it.
Otter.ai: Interview Transcription
Every journalist I spoke with uses AI transcription. The days of manually transcribing hour-long interviews are over.
Otter.ai transcribes accurately enough for routine interview work, identifies speakers, and generates summaries. Record an interview, get a searchable transcript in minutes. Search for specific quotes instead of scrubbing through audio. One thing to be aware of: a 2025 lawsuit raised privacy concerns about how Otter uses customer data for model training, so check their current data policy if you’re handling sensitive sources.
The workflow: Record interview → Otter transcribes → search transcript for key quotes → copy quotes directly into your story → verify against audio for accuracy.
Time saved: A 1-hour interview takes 4-6 hours to transcribe manually. Otter does it in real-time. Even with review time, you save 3-4 hours per interview.
Cost: Free (300 min/month) → Pro $8.33/month (billed annually) or $16.99/month
Whisper (Free, Local)
For sensitive interviews where you don’t want audio uploaded to a cloud service, OpenAI’s Whisper runs locally on your computer. Same accuracy, complete privacy.
Cost: Free
Writing Assistance
This is where things get ethically tricky. AI can absolutely help with writing, but the line between assistance and replacement matters in journalism more than any other field. Here’s how I see newsrooms drawing that line.
What AI Should and Shouldn’t Write
AI should help with:
- First drafts of routine stories (earnings reports, sports scores, weather)
- Headline options (generate 10, pick the best)
- Story structure suggestions
- Rewriting for different audiences (web vs. print vs. social)
- SEO optimization of headlines and descriptions
AI should NOT:
- Write opinion pieces or analysis (that’s your voice and judgment)
- Generate quotes (ever, and this is fabrication)
- Replace fact-checking (AI hallucinates)
- Write stories without human review and editing
The Ethical Line
The AP, Reuters, and most major newsrooms have AI policies. The common principles:
- AI-generated content must be reviewed by a human editor
- AI should not fabricate quotes, sources, or facts
- Readers should be informed when AI played a significant role
- Journalists are responsible for the accuracy of AI-assisted work
Fact-Checking
Misinformation spreads faster than corrections. Always has. But AI gives fact-checkers a fighting chance by automating the identification of claims worth checking and structuring the verification process.
ClaimBuster
AI that identifies check-worthy claims in speeches, press conferences, and articles. Feed it a political speech and ClaimBuster highlights the factual claims that should be verified. Saves fact-checkers from manually identifying claims in hours of content.
Cost: Free (research tool)
Claude for Verification
"A politician claimed that 'crime has increased 40% in the
past year.' Help me fact-check this:
1. What are the actual crime statistics from official sources?
2. What time period and geographic area matter?
3. What context might make this claim misleading even if
technically accurate?
4. What questions should I ask to verify this claim?"Claude doesn’t verify the claim itself (it might have outdated data), but it structures the fact-checking process and identifies what to look for. The journalist still does the actual verification.
Data Journalism
Data stories win awards and drive impact, but they used to require specialized skills that most reporters didn’t have. AI changed that equation. Now any journalist who can ask a clear question can pull insights from a dataset.
ChatGPT Code Interpreter: Data Analysis
Upload datasets (CSV, Excel) and ask questions in plain English. “Show me the trend in police use-of-force incidents by district over the past 5 years. Highlight any districts with significant increases.” ChatGPT generates charts and statistical analysis without requiring Python or R skills.
Impact: Data journalism used to require a dedicated data team. Now any reporter can analyze datasets and find stories in the numbers.
Datawrapper: AI-Assisted Visualization
Create publication-ready charts and maps from data. AI suggests the best chart type for your data and generates accessible, responsive visualizations that work in any CMS.
Cost: Free (basic) → $599/month (newsroom)
Social Media and Audience
Publishing the story is only half the job now. If nobody sees it, it doesn’t matter how good the reporting is. AI helps newsrooms distribute content across platforms without hiring a dedicated social media team.
AI for Social Distribution
How newsrooms use AI for social:
- Generate platform-specific posts from articles (X/Twitter thread, LinkedIn post, Instagram caption)
- Identify trending topics relevant to their beat
- Monitor social media for breaking news signals
- A/B test headlines for engagement
Claude approach:
"We published an article about [topic]. Generate:
1. An X/Twitter thread (5 tweets) highlighting key findings
2. A LinkedIn post for professional audience
3. An Instagram caption with hashtags
4. A push notification (under 100 characters) for our app"The Journalist AI Stack
Here is the practical recommendation by budget and role. The beat reporter stack covers daily reporting needs. The investigative stack adds more horsepower for deep dives. The newsroom stack is the setup I would pitch to an editor-in-chief equipping the whole team.
Beat Reporter ($20-28/month)
| Tool | Cost | Use Case |
|---|---|---|
| Perplexity Pro | $20/mo | Research |
| Otter.ai Free | $0 | Transcription (300 min) |
| Claude Free | $0 | Analysis, writing assist |
| Total | $20/mo |
Investigative Reporter ($40-48/month)
| Tool | Cost | Use Case |
|---|---|---|
| Claude Pro | $20/mo | Document analysis, deep research |
| Perplexity Pro | $20/mo | Sourced research |
| Otter.ai Pro | $8.33/mo | Unlimited transcription |
| Total | $48/mo |
Newsroom ($100-200/month per journalist)
| Tool | Cost | Use Case |
|---|---|---|
| Claude Team | $25/user/mo | Analysis, writing |
| Perplexity Team | $20/user/mo | Research |
| Otter.ai Business | $20/user/mo | Transcription |
| Datawrapper | ~$50/mo shared | Data visualization |
| Total | ~$115/user/mo |
The Trust Question
AI in journalism raises a fundamental question: can audiences trust AI-assisted reporting?
The answer depends on how it’s used. AI that helps a journalist research faster, transcribe accurately, and analyze data thoroughly produces better journalism. AI that generates stories without human oversight produces unreliable content.
The journalists who use AI well are more productive, more thorough, and break better stories. The ones who use AI as a shortcut produce work that erodes public trust. The tool is neutral. The journalist’s judgment determines the outcome.
Related guide: Can AI write SEO content?.