How do you write the events section of a local newsletter?

I don't write about events. Each one gets a single sentence that says what it is. That's it.

"I get AI to write it. What difference does it make? It's not a story, it's a description."

Events are also why people read. Finding good events in your community is nearly impossible without searching a million places. We do it for them, and they love it. I think that edge lasts at least another year, maybe two or three, so use it.

How long should each event listing be?

One sentence. What is this event? Readers scan the list for something they want to do this weekend. A paragraph per event slows them down and costs you hours. No description at all is too little.

Our format:

• Event name in bold.
• Location in italics.
• A page break and a "more" button.
• A one-sentence description.

Which dates, and how many events?

The day after the newsletter goes out, through the next seven days. If it goes out on the 23rd, it covers the 24th to the 31st. That's the exact rule I give the agent that builds my events section.

About eight usable events per newsletter. To get there I pull 20 to 25 and cut. Some are duplicates, some are stale, some are a bad fit.

Keep weekly recurring events, like the farmers' market, out of the regular list. Nobody wants to see the same thing every single week. Save them somewhere separate. If it's a light week, or you haven't featured the farmers' market in eight weeks, pull one or two back in.

Where do you find the events?

This is the hardest, most annoying part. Events are scattered all over the place, and a lot of them live on Facebook, which doesn't give you an API. Writing is the easy part now. Getting the data and finding the content is where we spend a ton of time.

What has worked for us over time:

1. An event aggregation service. In 2025 we paid one about $175 a month. It finds maybe 80% of what's out there. You put some effort into the other 20%, at the library and wherever else.
2. Agents instead of scrapers. Every website is structured differently, and sites update. You build a scraper for one, it doesn't work on another, and it all falls apart. Facebook and Instagram actively block scraping. A Claude-based agent adapts where a brittle automation keeps breaking. More on that in whether AI can run a local newsletter.
3. A local eye. My overseas team doesn't always have the context for what's a good event versus a bad one. Somebody who knows the town makes the final call.

Watch for canceled events. I've had canceled Eventbrite events still show up in an issue. I've weighed re-checking them right before send against over-engineering the whole thing.

Do you need the ten best events?

No. Six pretty good events found quickly versus the ten best: I'm pretty confident nothing changes for the reader. That's my prediction, not a test result. Spend the saved time somewhere readers can tell the difference.

You also can't feature everybody. People send us events all the time that are too small or too random. Your whole job is to curate what you think people want to see, so it's your judgment call. If somebody pushing timeshare-style pitches demands a listing, I don't feel obligated.

Where do events go in the issue?

The full list goes at the end. It's what people want to read most, so it makes them read the whole thing, the same reason grocery stores put the milk in the back. A big list of events at the top would also just be boring.

I used to put news at the top because news is what I personally care about most. That was a mistake. You're not the target reader. News now sits right above the events.

The exception is one featured event at the very top, really big. It gets the most clicks in the newsletter. That's what makes it sellable: go to an event venue and say "do you want to be the featured event of the week? It's $100." If you have no paying venues yet, feature one anyway so readers get used to the slot. When an advertiser wants in, give them the choice: one line at the bottom next to 30 other events, or the top with more information and way more clicks.

Which day should events go out?

Thursday. People are planning their weekend. Then package the same events for social as a weekend carousel. A guest on my podcast reported her weekend events carousel got almost 40,000 views, versus about 9,000 for the same events as a reel.

How do you get an AI agent to do this well?

Iterate with feedback. My events agent started around 70% good. Writing a better skill file got it to 90%. Now we push for 95 to 98%. Don't expect perfect on day one.

One mistake to fix right away: AI loves linking to a venue's general calendar instead of the event itself. Picture a reader who clicks on a concert next week and lands on a page of 30 concerts this month. That's really frustrating. Link to the exact page for that exact event, and put that rule in each brand's instructions.

Find local events happening from [day after send date] through [send date + 8 days].
Pull 20-25 candidates. Drop canceled, duplicate, recurring or out-of-area events.
Keep the best ~8. Write ONE sentence per event: what it is, when, where.
Link each event to its own event page, never a venue or monthly calendar.
No hype words, no emojis. Flag one event as the featured event.

That's a starting prompt, not a finished one. Your town will teach you the rest.