Can AI write local news from city council meetings and permits?

This was my idea a year ago. If you had asked me the week before, I'd have told you it wasn't realistic, that it would take three to six months. Then a new frontier model came out, and within days it was ready.

"The tech watches for city council meetings and then proactively writes the article by researching. There's seven main softwares in the United States that host city council content. It wrote adapters for all seven."

Why are council meetings such a good news source?

Because that's where the news gets discussed first, and nobody goes. A competitor in my area does breaking news, so I asked a chamber leader how they did it. Pretty simple: they send a reporter to city council and city meetings. That's where you hear about the new Chick-fil-A or the freeway getting extended three miles.

It's all public, but nobody knows because nobody posted about it. My city doesn't put meetings on YouTube, but the video and the transcripts are on its website, ready to download, about six pages deep. That's how you become the source that breaks the news.

How do I do it by hand today?

Back in April 2025, this was my workflow, and it still works:

1. Make a Claude project with instructions to write articles for your local newsletter in Smart Brevity style. Upload the style guide PDF the company publishes.
2. Get the transcript. From the city site, or a free website that posts council transcripts for a lot of cities, which someone in my newsletter group shared.
3. Paste it in and ask for as many articles as possible on the topics discussed.

One paste gave me four articles that were amazingly good about new stuff that had just happened. A story on our council declining to support a feasibility study for a contested road took about 10 seconds to write, structured as why it matters, the big picture, details, reactions, what's next and bottom line.

My check on that one was a reread, because I was only asking the AI to summarize the transcript I gave it, not to go research. In that setup it usually summarizes without errors. That's my light-touch approach, not a verified accuracy rate.

What other public sources can AI watch?

• Permit and licensing filings. Another local outlet posted that a chicken chain was taking over a closed restaurant, and they found it in the Texas licensing records. That's scrapeable.
• City department Facebook pages. A lot of city governments post to Facebook because it's easier than updating a website. Pull what the fire department releases, rewrite it, post it.
• Police postings and obituaries. A peer I worked with pulls the police department's arrest postings and funeral home obituaries. Some people love that stuff. Now you're the full media company in your town, like the local paper was 30 years ago.

Think about AI as a reporter. A reporter finds the information, rewrites it and submits it. It's the same thing, except now you just have to be the editor.

How does larkfeed turn a council meeting into an article?

1. It watches public sources: council meeting minutes and agendas, permit filings and other public data.
2. It connects through adapters for the seven main council-hosting platforms in the US.
3. It researches and drafts articles on what happened. For one meeting, that can be three: the big story, the secondary story and an overview.
4. A person reviews it, edits it, or pushes publish.

Meetings hosted only on YouTube need a paid third-party transcript service. That gets passed on as an option.

The bigger vision I've described since late 2025: you show up in the morning and the article is already written in your voice, with a shorter version for your newsletter and Instagram carousels from yesterday's news built on pretty templates. You just say yes. Most local media software today is legacy, built for 2009, and you can't even figure out how to use it.

What does the human do?

You become an editor. Here's how I describe it to publishers:

"Last night they talked about X at city council. We wrote an article on the most important thing. Go review it, edit it, or push publish... The stuff that doesn't need you as a person writes itself."

Same with a permit: Chick-fil-A files on the Texas permit site to open in Georgetown, the system writes it up, and you publish if you like it. First-person columns still come from you, by brain-dumping your take out loud and letting the AI shape it.

And local judgment still matters. The value isn't that the restaurant is coming. It's knowing the intersection was already backed up. AI can't know that, so the editor adds it.

What the editor shouldn't do is the old way. Sitting in a meeting for three hours, handwriting notes and then writing the story just isn't viable anymore, and it's unnecessary because AI does it really well. Being the perfectionist is going to get you crushed, whether you feel good about doing it that way or not.

Uriah Kaiser told me on my podcast that he feeds public meeting transcripts into AI and labels those articles as produced with AI and reviewed by humans for accuracy. He said it let him tell more in-depth stories after burnout nearly made him sell his business in 2023. That's his experience, and it matches what I see.

Isn't everyone already doing this?

No. Even the people who are ahead have a 10-step workflow that breaks all the time. Plenty of local operators understand the idea but have nothing integrated. The difference is putting the workflow inside the software. I've shown it to about eight people, including large local publishers, and they all wanted it. I think this, not the newsletters, is the real money maker.

How fast can you launch a local news site now?

I'd already bought a few city domains months earlier. I told the system: just make me one. It generated logos (one with a Statue of Liberty head I never asked for), maps and templates. It took me about 10 minutes of my own time to launch four sites.

larkfeed runs many sites from one backend, and cities can combine into a single brand covering two neighboring towns. One system, one AI agent and one team can create content for every site automatically. For the newsletter side of the same idea, see can AI write and run a local newsletter.