I Scanned Every Chiropractor in Wilmington. Six of Them Own the Answer.

Ask ChatGPT for the best chiropractor in Wilmington and it hands you six names. There are twenty-four practices in this town.

I know that because I checked all twenty-four, one at a time, on a Saturday.

If you want the actual list, plus the three things separating the six who get recommended from the eighteen who don't, that lives on its own page: the Wilmington list. No email, nothing gated.

This post is the other half. The method, the two places I got it wrong, and the one claim on that page I genuinely cannot prove. Because if I'm going to tell twenty-four business owners what's broken on their websites, the least I can do is show my work.

Why I did this at all

Honestly? Because my ads weren't working and I was annoyed about it.

I'd spent $101.91 over thirty days sending strangers to a free scan tool. Ninety-six of them landed on the page. Zero did anything. Meta's own scoring told me the ad was fine and the thing after the click wasn't, which is a polite way of saying I was asking people to do homework about a problem they didn't know they had.

So instead of writing another ad, I went and looked at my own town. Turns out the finding was sitting right there the whole time, and it needed exactly zero persuasion. I wasn't selling anybody an opinion about AI. I was showing somebody a list they weren't on.

Where the list came from

I pulled every business in the Wilmington metro filed under NAICS 621310.

NAICS is just the government's filing system for what a business does. 621310 is "Offices of Chiropractors." It's not glamorous, but it's how you get a list that isn't cherry-picked, which matters a lot when you're about to make claims about a whole industry in a small town.

That export gave me nineteen practices. Five more surfaced during the recommendation searches themselves, practices the data didn't have but the internet clearly did. Twenty-four total, Wilmington plus Hampstead, Leland and Surf City.

How I decided who "gets recommended"

I asked the way a patient would ask.

Not one query. Several phrasings, across a few engines, on the same day. "Best chiropractor in Wilmington NC." "Who should I see for lower back pain in Wilmington." "Chiropractor near me Wilmington." The specific words changed. The six names barely did, which is what makes it a finding rather than a fluke.

The other thing that showed up every single time: national directories. Healthgrades, WebMD, ThreeBestRated, and a franchise chain. Four sites that have never treated one person in this town, absorbing the top of the answer anyway.

That's the part people get wrong about AI search. It isn't that the engine picked your competitor over you. It's that the engine picked a directory over both of you, and your competitor at least made it into the directory's shortlist.

What I checked on each site

Same nine things, every time, so the comparison stays honest:

The title tag. The text on your browser tab, and the blue line in Google results.

The meta description. Present, absent, or arguing with the headline.

Location signals. City, state, street address, phone, in the visible text.

Word count on the homepage. How much there is for anything to read.

Whether the page serves any text at all without JavaScript running.

Whether both hostnames resolve. The bare domain AND the www version.

robots.txt and the sitemap it points at.

llms.txt, the file that tells AI crawlers what you are.

Whether a blog or an FAQ exists, and whether it lives on the page that ranks.

Nothing exotic. Every one of those is checkable by the owner in under a minute, which was the point. Nobody argues with their own footer.

The check that surprised me most

The JavaScript one, by a mile.

Some of these sites look completely fine when you open them. Nice photos, clean layout, everything where it should be. Then you fetch the same URL the way a crawler does, without running a browser, and you get back one line of viewport code and nothing else. No title. No headline. No address. No phone.

To a person, a working website. To a machine, an empty room.

That's not a small SEO ding, that's the business not existing to the thing your next customer is asking. And you would never catch it from the inside, because you look at your site in a browser, like a human, the way it was designed to be looked at.

The two places I got this wrong

Both of these happened during the scan, before a single word went out to anybody, which is the only reason I'm comfortable writing them down.

On one practice I decided there was no location information on their site. Then I found a town name on their Facebook page. So I decided there was no street address. Then I found it, complete and correct, sitting on their privacy policy page. Four separate times I went looking for something, didn't find it, and quietly treated "I didn't find it" as "it isn't there."

Every one of those errors was the same shape. I was claiming something was absent.

Here's the rule I came out the other side with, and I'd take it seriously if you ever do this yourself. A single page fetch can prove that something IS there. It can never prove that something isn't. "Your footer says 2012" is a fact. "You have no address" is a guess wearing a fact's clothes.

The email that eventually went to that practice never mentioned a missing address, because by then I knew better. What it actually said was that their site gives two different answers about where they are... and then quoted all three, including the complete one buried on the privacy policy page, which is the page nobody reads and search engines care about least. That's a finding somebody can go verify in thirty seconds. The version I almost sent would have been an accusation, and a wrong one, from a stranger.

So on the public page, every absence claim is either hedged or cut. The ones that survived are all positive observations: a live 404, a dead DNS record, a copyright year, a published sentence.

The second one is smaller but it would have made me look like a jerk. Two practices have domains that resolve, answer, and 404 on every single path. My first instinct was "your website is gone, let me help." Then I sat with it and realized a totally dead website is very often a closed business, and leading with "your site is broken" to somebody who already knows because they shut the doors is a genuinely awful thing to do to a person. Those two got moved to a "confirm they're still trading first" pile.

The one thing I can't prove

Schema markup.

Schema is structured data, a block of code that tells a search engine "this is a business, here's the address, here are the hours" in a format built for machines instead of people. It usually lives inside a script tag.

My fetch strips script tags. Which means for most of these twenty-four sites, the honest answer on schema isn't "absent"... it's UNVERIFIED. I did not see it. That is not the same thing as it not being there, and I'm not going to tell somebody they're missing structured data when what I actually mean is that my tool couldn't look.

If you want to check yours properly, right-click your homepage, choose View Page Source, and search for "application/ld+json". Or run the URL through Google's Rich Results Test, which is free and does it in about ten seconds.

I'm putting that limitation in bold print on the public page too. A report that tells you where it's blind is worth more than one that pretends it isn't.

How to run this on your own industry

You don't need me for any of this, and I'd rather you had it.

Pick your trade and your town. Write down every competitor you can name, then add the ones you find in a maps search. Aim for the whole set, not the ones you think about.

Ask three or four engines the question a customer would ask. Vary the phrasing. Write down who comes back by name, and how many of the results are national directories rather than actual local businesses.

Open each site and read the browser tab. That's the title tag. Does it say what they do and where?

Press Ctrl+F on each homepage and search for the city, the address, the phone. All three should be in the visible text.

Type www in front of each domain in a private window, then try it without the www. Both should load.

Now do all of the above to your own site, last, when you already know what good looks like.

An afternoon, tops. And at the end of it you'll know something concrete about your market that no dashboard was ever going to tell you.

Why I gave the whole thing away

Because the alternative is being the fifth stranger this week who found a problem and won't say what it is until somebody pays.

I've been on the other side of that. Somebody runs an audit, tells you there are "several critical issues affecting your visibility," and then holds them hostage behind a discovery call. It's a rotten way to meet a person, and it's especially rotten when the fixes turn out to be a two-minute title change and a footer nobody updated since 2012.

The last big client I landed came from handing over a fifty page audit with exact links, why each thing mattered, and how to fix it. They interviewed over a dozen companies. They picked the one that had already given them the answer.

I'm not going to pretend I don't want the work. I do. But you should get to walk away better off even if you never write back, and that's the actual test I hold this stuff to.

The list is right here. Take it and fix your own site if you want to. That's allowed.