Earned exposure, SEO & AI search
How do I run an automated SEO article system for local sites without publishing AI slop?
Short answer
Go slow and make every article specific. I cap automated output at about one article a day per brand, often every other or third day, with detailed prompts per brand. Research tools feed the writer real local details, every location article names real neighborhoods and cities, and a final pass strips out AI tells.
I want my local sites to publish a lot of useful articles without anyone writing them by hand. The goal is a system where it happens automatically. The risk is obvious: a pile of near-identical AI pages that helps nobody.
These are the rules I've set for my own systems.
How often should it publish?
At most one article a day per brand, if not one every other day or one every third day. The content has to be super high quality, so each brand gets its own very specific prompts and plan. Volume isn't the point.
How do you keep location articles from being copies?
A lot of my sites need the same kind of article for different places. That's exactly where AI content goes bad.
"Every article, even if it's about the same topic in a different location, has to be custom, pulling custom info about custom neighborhoods or cities that make it very specific, calling out names and cities."
- Feed the writer real research. Tools like Firecrawl or Exa pull information so the writer actually understands the area.
- Name real places. Neighborhoods, cities, local names.
- Start where you live. I start my location sites in Williamson County or Austin, then the San Antonio area. It's easier to recruit partners and sell leads in places I actually live.
Does AI-sounding writing hurt you?
I assume it can. When I relaunched a content site, I had the AI edit every article specifically for AI language and AI tells. My worry wasn't mostly human readers noticing. It was Google or the LLMs knocking the site down because the language reads too AI.
Then I had it audit its own work. The goal is a site that's as valuable as possible to both humans and AI. The human parts that make that possible are in the 10-80-10 rule.
Where do the photos come from?
For first-person articles, I take my own photos when we go places. Broad informational articles won't have our own photos, so those images have to come from somewhere else. My requirement: the agents do it automatically, not a person by hand.
Design matters too. One AI article preview I got was straight text with a blank line between every line and a couple of headers. That's not what an article should look like. I wanted real design with photos, cropped closer to square instead of super wide.
What about the technical side?
I'm not an SEO expert. On one local directory, I told Claude Code to optimize the site for SEO and it handled the data and the SEO work, including local guide pages for every restaurant type in every city. The SEO score came out very high.
For my product sites, I ask for schema, JSON-LD and robots.txt set up well, especially on the homepage and the articles page. Articles go on the public marketing site where anyone can read them, never behind a login. Domain choices are in subdomain vs. root domain.
Do small fixes add up?
That's my bet. Each thing we do probably isn't very valuable on its own. Keep doing all of them, over and over, and they add up to something much bigger. WilCo Guide's SEO scores and impressions are already much better than my sites hosted on my email platform, so I think we're headed in the right direction. I can't promise rankings or AI citations from any of this.
Answered by TJ Larkin. From my own build notes, calls with my team and work on WilCo Guide and other local sites, 2026. Last updated .