I typed a sentence into ChatGPT: best acupuncturist in St. Petersburg, Florida.

I got an answer. Not ten links to sort through. An answer, naming a handful of practices, written with the confidence of a friend who’d been to all of them.

Then I did it 26 more times. Nine different questions, the kind real patients ask, run across ChatGPT, Google AI Overviews, and Gemini. I built a list of 28 independent acupuncture practices across St. Pete, Tampa, and Clearwater, and I tracked who got named.

Eleven of the 28 were never mentioned. Not once, on any platform, for any question. Two-thirds of every recommendation went to six practices.

These are not bad businesses. One of the invisible ones has 133 Google reviews and does everything a marketer would tell you to do. It simply does not exist as far as these systems are concerned.

*Why this is a St. Pete problem and not a marketing problem*

For twenty years, getting found meant ranking on a page. You could be tenth and still get seen. Somebody scrolled.

Nobody scrolls an answer. When a machine names three practices, the fourth one wasn’t ranked lower. It wasn’t there.

That shift lands hardest on exactly the businesses this city is built on. A national chain has an agency watching this. A two-person shop on Central Avenue has the owner, at 10 p.m., after close. And nothing in their analytics tells them it’s happening. There’s no report that says a customer asked about you today and got sent somewhere else.

I’ve been in marketing and web for more than fifteen years. I’ve watched plenty of platform shifts. I haven’t seen one arrive this quietly.

*What actually separated the visible from the invisible*

I audited all 28 practices against seven basic signals. Two of them explained almost everything.

*Pages about specific problems.* Fifty-three percent of the practices that got cited had a page devoted to a particular condition. Among the practices that never got cited, that number was nine percent. This was the single largest gap in the study.

The pattern held in the queries themselves. Asking for the best acupuncturist in St. Petersburg gave me a different top answer on all three platforms. Asking about back pain, or migraines, or fertility gave me the same names over and over. Specific questions get stable answers, and the businesses that had written something specific were the ones getting named.

Recent reviews, not total reviews. Eighty-eight percent of cited practices had three or more Google reviews from the previous 90 days. Among uncited practices, 45 percent did. Total review count mattered far less: 59 percent versus 45 percent, a much narrower gap.

Fifty reviews from 2019 read as a business that used to be busy. Four reviews from last month read as a business that’s open.

*And one thing that surprised me*

The practices that were invisible were, on average, slightly better at the standard checklist items. More of them had a complete Google Business Profile than the cited group did.

One practice scored a perfect seven out of seven on every signal I measured and got cited zero times. Meanwhile the most-cited practice in the whole study was missing two of the seven, and got named in 16 of 27 tests on the strength of 349 reviews and a deep presence across other websites.

So the checklist is necessary and it isn’t sufficient. What these systems seem to be doing is looking for a business other sources already agree exists and can vouch for. You can build the perfect house and still not be on the map.

*What you can do this week*

Ask the machines about your own business, but do it in a private or incognito window, always. If you’re signed in to your own accounts, these systems already know who you are and where you’ve been, and they will show you a flattering version of reality. You’ll see yourself and conclude everything is fine. A stranger asking the same question gets a different answer, and the stranger is the customer.

So: incognito window, no account signed in, then type the question a customer would type into ChatGPT and Gemini. See whether you come up at all, and whether what it says about you is even true. It takes ten minutes and most owners have never done it.

Then write one page about the specific problem people actually come to you for. Not a services list. The problem, named, in the words a customer would use, with your city in it.

Then ask three recent customers for a review this month, and do it again next month.

None of that requires hiring anyone. It requires knowing there’s a door there.