Can AI write and run a local newsletter?

In mid-2026 I tested Claude's dynamic workflows, a coordinator agent running sub-agents at the same time, and got a local newsletter to roughly 97% quality in about an hour. The day I first tried it, my local newsletter production changed.

That's a big swing from where I started. Here's how it changed, what the machine does now, and what a person still has to own.

How fast did AI go from useless to useful here?

Fast. In December 2024 I told people a local newsletter was still a VA job. I'd dug into it a lot, and AI couldn't do much of it. Finding good events meant going through Facebook by hand.

By mid-2025 it was the same story. We used ChatGPT to help rewrite stuff, and that was it. The real work was research and piecing things together.

Then in March 2026 I tested Claude on one of my local newsletters and got a draft with real event links. Six to nine months earlier, ChatGPT's event links went 404 every time. They weren't even real. This time only two or three out of 10 linked to a general page instead of the event page, and the right page did exist. It didn't hallucinate. A few months later came the 97% test.

What can AI actually do in a local newsletter today?

In one demo, Claude Code built a full Beehiiv newsletter end to end with no human drafting:

• News written from local sources
• Real estate listings
• Pet adoption listings pulled live from the PetFinder API
• Local events it went out and found
• Images it generated

The one gap was images for news stories. A headline-graphic style works as a stand-in. I also tested AI-generated memes, and three of four were rated good.

Images matter more than people think. The automated local newsletters I see from competitors are all text, and that reads as AI to me. Photos make it feel more real, so automating a good header image and event images is a real differentiator.

How much should AI write, and how much should you?

My rule is 90% AI, 10% you. Use AI for the first 90%, then rewrite it, edit it and add your personal flavor. For a national newsletter I've estimated AI can do 90 to 95% and a teammate edits it to make it a bit better.

"90% AI, 10% you, right? The first 90%, I mean, it obviously depends on the vertical, but like the first 90% use AI. And then you go through it. And you rewrite it, you edit it, you add your personal flavor."

Both extremes lose. AI-only newsletters set to go out every day on their own are going to be garbage, nobody's going to read them, and people will unsubscribe. The people saying never touch AI are going to get crushed. You need to be doing both.

The human part is the whole moat in local. What makes this opportunity special is the person's brand and writing like a resident. If it could all be AI, what's to stop somebody in Chicago from running one for your town?

What's the real bottleneck?

Sourcing. Everyone already knows how to get AI to help write.

"It's less about writing the content... It's more about getting the data and it's the finding the content."

Local information is broken. Every town is its own island with no incentive to streamline, and there just isn't a lot of local news for an AI to find. That's where the time goes, and that's where the setup matters:

1. Add the exact page that changes, not just the domain. Sources decide what news the AI sees, so this is the most important setup step.
2. Check each source is actually pulling content. "Added" doesn't always mean working.
3. Use an agent for events, not a scraper. Event sites change formatting constantly, and Facebook and Instagram block scraping. An agent adapts. My target is to pull 20 to 25 events to land the roughly eight an issue needs.
4. Pull from RSS and YouTube, too. The software ingests real sources and writes each section in your voice into a fixed template. It's not "type: I want a newsletter about leadership."
5. Use what you already paid for. A lot of local media outlets pay journalists to write articles that sit on a WordPress site nobody visits. Drag those into the newsletter and have AI write each item. It's your content.

Why not just ask Claude to write the whole thing?

People tell me they'll just use Claude to write their local newsletter. It's possible. But you can't edit it, and I'm not putting something out in my own name, like Local Media HQ, without editing it. And how are you going to run ads in a newsletter written in a chat window?

That's why I want the AI to build issues in the same section-based builder a person uses. The AI's job is deciding which news, events and businesses are worth including. Once it decides, it uses the builder like we would, so anyone can edit, add, delete or swap a piece.

Drag and drop also turns out to be what wins over journalists who are skeptical of AI. They still look at the content and place it themselves, so it feels like their work. If they spend three hours writing a newsletter today, they can do it in about 10 minutes.

Speed is the payoff. In recorded lightbreak demos in February 2026, I said a newsletter can be created in under five minutes per issue once the template is set up. My co-presenter said under 15 minutes, with a social repurpose in under 30 seconds. Those are demo timings on prepared templates, not audited production runs. One trick that helps: generate the intro last, because it summarizes what's already in the issue.

Before I roll a setup out across brands, I test it on a few issues first: "We don't want to do any copy-pasting until we know it's giving us exactly what we want to a T."

What does a human still have to do?

• Supply the local knowledge. If a story's insight could have been written from out of state, it isn't worth publishing under a local brand. The value isn't that a Chick-fil-A is opening. It's knowing that intersection was already backed up. AI can't know that.
• Build the relationship. When you just use AI, you can't build trust with readers. Interviews with local business owners, in their own words, are what regurgitated AI news can't copy.
• Read it before it sends. My grading bar for fully automated issues is 9 out of 10 from an automatic reviewer; 8 is questionable and 7 is unacceptable. Early test issues graded 5, 6 and 6, so I keep fixing gaps. Until a system clears that bar, a person reads it: should you send an AI newsletter without reading it.
• Get it to readers. Every automated local newsletter I've seen is terrible, and on top of that they don't have the audience. Who are you sending it to? Building the thing doesn't make anyone care. Growth still takes ads or cross-promotion.
• Own the last 10%. AI gets a build 90% done in an hour that used to take months. The last 10% is still hard.

Is newsletter production a real service now?

I think so. Newsletter production as a service, at $500 to $1,000 a month, is operationally viable now: the machine does the volume and a person owns the quality. I'd argue AI is the opposite of a threat to newsletter businesses for the next year or two, because an agency can now do great newsletters for people quickly and cheaply. That's the work lightbreak does for local newsletter operators.