|
|
Was this forwarded to you? Get The Fourth Estate AI Brief in your inbox →
|
|
|
|
View this issue in your browser
Tuesday Brief
Your Journalism Is Already Training the Machines
The interesting thing this week with AI and publishing — your content is already the biggest source feeding AI answers, and you probably didn't sign up for that. A Muck Rack study found journalism accounts for 27% of all AI citations, beating out corporate blogs, Wikipedia, and everything else. Meanwhile, readers are starting to ask ChatGPT about your town instead of visiting your site. The data this week paints a picture of an industry whose work product is being harvested at scale while the harvesters build tools that could redirect traffic away from the source. Here's what showed up that matters for publishers like us.
|
|
WHO'S FEEDING THE MACHINES
The Rundown: A new Muck Rack study of 25 million AI response links found news content is the single biggest category AI tools pull from when answering questions.
The details:
- Journalism accounts for 27% of AI citations — more than corporate blogs (24%), aggregators (17%), or any other category.
- 99% of links cited in AI answers come from non-paid, earned media sources.
- ChatGPT includes citations in 96% of responses; Claude cites sources only 55% of the time but averages 13 citations per answer.
- Hospitality, real estate, and government beats are among the sectors most reliant on journalistic content in AI answers.
- Forbes is the only outlet appearing in the top five for multiple AI providers.
Why it matters for us: Your reporting on city council, real estate deals, and local business is likely already being scraped and cited by AI tools. Knowing which beats show up in AI answers can help you prioritize coverage that maintains visibility — or decide whether to block crawlers entirely.
Read at mediapost.com →
|
|
|
THE TRAFFIC REDIRECT
The Rundown: Reuters Institute's 2026 Digital News Report shows AI chatbots are becoming a real channel for news consumption, especially among younger readers.
The details:
- Weekly use of AI chatbots for news grew from 7% to 10% globally since last year.
- 17% of the youngest age group uses chatbots for news — three times the rate of the oldest group.
- Only 1% say AI is their main news source, suggesting it's currently complementary rather than replacing direct visits.
- Usage is highest among 'news lovers' (18%) and those extremely interested in news (19%).
- Growth is driven largely by parts of Asia, Africa, Latin America, and Southern and Eastern Europe.
Why it matters for us: Your most engaged readers — the ones who actually care about local news — are the same ones experimenting with ChatGPT for summaries. A 25-year-old asking Gemini 'what happened at the school board meeting' instead of visiting your site is a trend worth watching now, not after it's normalized.
Read at reutersinstitute.politics.ox.ac.uk →
|
|
|
NOTEBOOKLM LEVELS UP
The Rundown: NotebookLM got a major upgrade — it can now write code, create editable files, and search the web for sources directly in the conversation.
The details:
- NotebookLM can now output editable PDFs, PowerPoints, Excel spreadsheets, and PNG/SVG charts.
- Web source discovery is integrated into the main chat — no more multi-step import process.
- The AI now shows its reasoning process in real-time instead of just a loading spinner.
- New features currently require Google AI Ultra plan at $100-200/month.
- The tool still cannot generate or pull images from the web, and AI responses still need human verification.
Why it matters for us: If you've got a new reporter covering courts for the first time, feeding NotebookLM a stack of case documents and asking for a formatted backgrounder could save a day of onboarding. The $100/month price tag is steep for a small shop, but worth testing if you're drowning in document-heavy beats.
Read at lifehacker.com →
|
|
|
THE TRUST ASYMMETRY
The Rundown: Knight Columbia's analysis of AI-generated content argues that disclosure requirements alone won't solve the authenticity crisis — but that's actually an opportunity for community publishers.
The details:
- The asymmetry between creating synthetic content (cheap, fast) and detecting it (expensive, imperfect) favors deception at scale.
- Traditional trust mechanisms break down when AI can impersonate humans at scale.
- The piece argues authentication and provenance systems may become essential infrastructure.
- Proposed solutions include fraud doctrines for AI content and mandatory disclosure requirements.
- Current voluntary labeling faces 'humanwashing' risks — companies claiming human authorship when AI did the work.
Why it matters for us: When national outlets are fighting internal battles over whether AI-written content gets human bylines, your named reporters with photos who show up at Friday's football game become a trust signal that's hard to fake. Lean into it.
Read at knightcolumbia.org →
|
|
|
THE UPDATE
Quick hits from the week
The Rundown: Smaller items that crossed the wire this week.
The details:
- The Muck Rack study found industry trend queries (46%) are the most common type of question people ask AI — useful to know if you're thinking about what content to produce.
- Media/Entertainment (36%), Finance/Insurance (32%), and Sports (31%) are the sectors where AI leans hardest on journalism for answers.
- Claude cites sources only 55% of the time but averages 13 citations when it does — more than double ChatGPT's average of 5.
Why it matters for us: These data points matter when you're deciding whether to block AI crawlers or position your content for citation.
|
|
|
What is?
NotebookLM
What it is: Google's free research tool that turns a pile of your own documents into something you can interview. You upload the sources — PDFs, reports, court filings, a year of council minutes, a fat records-request dump — and NotebookLM becomes an instant expert on just those files. Ask it a question and it answers using only what you uploaded, with a citation pointing to the exact page it pulled the answer from. It can even generate a podcast-style audio discussion of your documents, and the upgrade in this issue lets it export spreadsheets, slides, and PDFs too.
Why publishers care: It's built for exactly the journalist's nightmare: a mountain of documents and a deadline. Drop a 500-page budget or a stack of public records into it and ask "what changed?" or "where does it mention the new contract?" — and it answers in seconds, citing the source page so you can verify before you publish. Because it reads only your documents and not the open web, it hallucinates far less than a general chatbot, and every claim is traceable. For a small newsroom, it's the closest thing there is to a research assistant who read everything overnight.
Visit notebooklm.google.com →
|
|
|
|
The through-line this week is clear: your journalism is already feeding the machines, whether you consented to it or not. The question isn't whether to engage with AI — it's whether to do it on your terms or theirs. Next Tuesday we'll look at what's next. Until then, if you're experimenting with any of this at your paper, hit reply and tell me what's working.
Trevor Slette runs Quadd.ai — AI tools built for publishers.
|
|
— Trevor
Trevor Slette
Co-founder, Quadd.ai · 28-year community publisher
trevors@quadd.ai
·
Book a 15-min call
·
LinkedIn
P.S. P.S. Curious whether AI tools are already citing your paper? Reply "check" and I'll show you the 5-minute way to find out — and what to do with the answer.
|
|
|
|
|
AP style without the AI mush.
Quadd's AP-Style Proofreading flags grammar, style, and headline issues with accept-or-reject suggestions. It tightens copy without inventing facts, flattening your voice, or turning every story into AI mush. The accept-or-reject part matters: you stay the editor. The AI just catches what a tired second pair of eyes would catch — if you had a second pair of eyes.
See how proofreading works →
|
|
|
