Why local news startups fail: four mistakes first-time publishers make

A few months into running a local newsletter, the same picture shows up. The town's website has a business directory where half the listings sit empty, because owners never got around to filling them in. The city council has met four times and the publisher made it to none of them. And the pages that did get published were built for Google alone, so the AI assistants people now ask for local answers have nothing they can read. This is the point where a lot of first attempts quietly stop.

The failures are predictable, and so are the fixes. Here are the four that matter most.

Mistake 1: A directory that waits for owners

A business directory only earns money when its pages exist. Most publishers build the empty shell, then wait for owners to sign up and fill in hours, photos and descriptions. Owners are busy. The directory stays thin, visitors stop trusting it, and there is nothing to sell. The fix is to load the businesses first. Every local business gets a real page with photos, reviews, hours and a map, whether or not the owner has claimed it yet. The structured pages exist from day one rather than trickling in as owners sign up.

The claim is the front door to revenue. An owner claims their page for free, and once the page belongs to them, the publisher can sell an upgrade. An upgraded page might run $49, $100 or $300 a month, and the publisher sets every price. Payment runs by card inside the platform, so nobody chases invoices. The publisher keeps the pricing power; the assembly work is what goes away.

Mistake 2: Pages built for Google only

People increasingly ask ChatGPT and other AI assistants where to eat, what happened at the last council meeting and who is hiring in town. Those tools answer from pages they can parse and verify. A page with no structured data, no author and no date is invisible to them, no matter how good the reporting was. Structured data, a standardized set of machine-readable tags, tells a search engine or an assistant exactly what it is looking at: this is a restaurant, these are its hours, this is the author, this is the date the story ran.

The fix has three parts. Mark up every page, not just the news. Add a guide file for AI crawlers so the tools can find their way around the site. And put sources, dates and authors on every article, because an assistant needs those before it will attribute a claim to a publication. A story that says who wrote it, when it ran and where the facts came from is citable. One that says none of those things is not.

Mistake 3: One product, one ad slot

A newsletter that only sells sponsor slots has one product. When an advertiser passes, a banner sits unsold and the business has no fallback. The fix is a site with inventory. An events calendar that takes free submissions fills itself with things to do, and the publisher sells featured spots and Event Spotlights. A job board that takes free listings keeps local openings flowing, and employers pay for premium roles at the top. Upgraded business pages, sponsor bands and ad spots round out the list. The free submissions do the work of making the site look alive; the paid placements give buyers a reason to spend. A newsletter sitting on that kind of site has several things to sell instead of one.

Mistake 4: Council coverage nobody has time for

The publisher who misses four straight council meetings does not lack interest. They lack hours. A single meeting can come with a 300-page packet. Reading it, finding the three decisions that matter, and drafting the story is the work of an evening, repeated every month, across every city on the beat. Software can take the reading. AI handles a long packet well. It does not decide what to publish.

The pattern that works: a publisher picks the cities to watch, the agendas and meeting videos come in, a sourced draft lands in the inbox, and the publisher reads it, fixes what they want, and publishes. The same idea covers permit filings. When a business files to build in town, the filing itself says what is coming, and a draft from the permit can beat everyone else to the story. Nothing goes live until a person says so.

Each of these fixes can be assembled by hand: load a directory with a scraper, add structured data with a plugin, write a rate card, and read council packets at night. larkfeed bundles all four on one site, which is the reason to look at it before building any of them yourself. It loads the local businesses into the publisher's directory, marks up every business, article, author and place page for Google and AI search, ships a guide for AI crawlers, watches the council meetings, agendas and permit filings a publisher picks, and holds every draft until a person approves it. Its article writer turns a brain dump of notes, photos and links into a designed draft with sources, and a second AI edits and scores it before a human reads it. WilCo Guide, serving Williamson County, Texas, was the first site running on it.