How do you build a local directory that AI search will cite?

About a month after launch, nobody knew WilCo Guide existed except Google. I hadn't spent an ounce of effort promoting it. My own audience didn't even know about it. We were still getting roughly two businesses a day filling out the form to submit themselves, plus event submissions, and an AI research audit told me Google was already citing the site.

Around the same time I was told the site had something like 900,000 Google impressions in that first month. The most-found section was the job board. It's the case for building a directory for AI search instead of for browsing.

What makes a directory "AI-first"?

Most directories that exist today look bad to humans and look bad to AI. Mine is built to be crawlable and structured from day one.

"By being the one that's AI first, you're, you're the only one."

A normal directory stores name, hours, description. Mine stores answers to the exact way people ask questions, like whether a practice will send a doctor to a senior's house for an ear infection. That kind of specific, unusual answer is what AI can match to a real question. Almost nobody else is doing it, because most directories aren't built around data and lots of content.

How do you seed the directory with the right businesses?

If you just tell an AI agent to go scrape businesses, you can end up with a thousand pet centers and Mexican restaurants, and that's your directory now. Ask for a representative mix by category and city instead. You should have more Mexican restaurants than ophthalmologists, because there are more of them and people care more.

The plan my agent came back with looked like this, per city:

• 20 Mexican, 15 Chinese, 5 sushi, and so on for every food category
• 15 main doctors, 10 pediatricians, 5 urgent cares
• De-duplicate, then verify nobody is listed in the wrong city

It also said I needed about 250 "best" pages: best restaurants in Liberty Hill, best Mexican restaurants in Liberty Hill, best Chinese restaurants in Leander. It estimated six hours. It took 30 minutes. Separately, a teammate gave me an SEO prompt for the directory that made about 50 pages of code changes and took the site's score from 66 to 88 in 30 minutes. That's infrastructure. You do it once.

What data should every listing have?

My standard for each business in a new city:

Item	Rule
Photos	3 to 5 real ones, uploaded by the business or by people. Street View doesn't count.
Reviews	5 written reviews required, 10 max. Star-only ratings don't count.
City	Required. The whole system is organized by city.
Scrape budget	About $45 to $50 per new city

The city rule came from a surprise. Almost every owner who claimed a listing filled in hours and a description and nothing else, not even what city they're in. Now it's a required field with a star.

Don't publish empty pages. A business page with nothing on it either gets filled in by scraping or stays off the site. Same with categories: a page called "all the pest control companies in the area" with one company on it looks thin. Store the category as a tag so it's ready, and give it its own page once there are enough businesses.

What goes on each business page?

• The basics, correct. Same facts the business uses everywhere else.
• The owner's own words. I give owners a free page and a short greetform interview about their business, and we post it on their page. What the LLMs care about is trust, and they want custom content an LLM couldn't write itself. Everybody's asking Claude for 50 articles about insurance. A teammate made a sharp point here too: human authorship alone doesn't make content authoritative. It needs a real owner and real entity signals, so keep the owner's name and link the answers to the business.
• A stack of articles. I call it the gold. A roofer with 50 articles about roofing in his town, on a domain that search and AI already favor, gets brand-level search value. Articles live as subpages of the directory page, with no limit.

On pace, I've changed my mind. I first planned several new articles a month in the paid tier and cut it to one, because you can hurt SEO by overdoing things. See the 10-80-10 rule.

Why would AI trust it? Google and ChatGPT know everybody is manipulating their own sites, so they lean on third parties. Over time, I think that means trusting local media brands. If the domain earns trust and your business shows up on it many times, through the directory page, articles, the newsletter and homepage slots, your odds of being cited go up. Your words on a third-party site are worth more than the same words on your own site.

That's the pitch for the paid tier: about $50 a month to borrow a well-optimized domain's standing. I went back and forth with Claude on it, and its blunt answer was that it won't work as perfectly as I want, but for the price it's a home run. I agree with both halves.

How do you get businesses to join?

Free. Let owners add their business for free and upsell them later. A free listing removes the objection, and it gives you a reason to talk to the owner: I built this directory for AI traffic, I built your page for free, I just want to make sure it's correct. Now they're surprised you did that for free, and you have a relationship.

For a public adjuster directory, the cold email says: we built a system that helps people find you in AI chats, sign up free, and we'll interview you and post custom content so your chances of being cited by AI go up.

One growth idea I'm planning: tag local businesses in social posts with a branded comment and no link, so platforms don't flag it, then follow with roundups and a "Featured on" badge that earns backlinks. Advice I got: pilot it by hand in one city for two weeks, and publish roundups before single-business pages so you don't create thin content.

Does the domain matter?

A lot. WilCo Guide is one website with the news, the directory, a job board and local guides for two suburb newsletters, and each newsletter keeps its own name. For a national senior site I'm planning the bigger version: one domain, with each business tagged by category and place. An estate planner in Miami shows up on the national estate planning page, the Florida page and the Miami page. Details in subdomain vs. root domain. Useful tools like calculators and trip planners help too. Google and AI both love them.

Why does a media brand have the advantage?

Website traffic for news is dropping because AI gives readers the headline without a click. I wouldn't build a local media plan on story traffic. The directory is the piece that makes my website model different, because its job is to help a business get cited by AI as the best roofer or the best insurance guy. That AI traffic might be worth more than human traffic.

A standalone directory is mostly an SEO bet. You hope people find it, and the only thing to sell is a paid spot at the top. Two big directory operators on my podcast said that once you have enough traffic you can sell listing slots for $50 or $100 a month. A media company already has newsletter readers, so it can push people to the directory instead of pulling them. I don't need to make money directly from the directory. It's one placement inside a package with newsletter ads, Instagram reels and event sponsorships. More on that in how to monetize a local directory.

Ken Jee told me on my podcast that his free business directory has 300+ signups, about half said yes or maybe to sponsoring, and he has 30+ sponsorships this year with featured listings around $20 a month. Those are his numbers, not mine.

Some people think directories are done because of AI. From my own research, I think they may get more valuable. But just building one doesn't make it work. Plenty of people will spin up a directory in Claude in three hours, leave it up, and hope. If you're out in the community talking to businesses, it can have a massive impact. Early citations are a good sign, and none of it is guaranteed.