Why AI search ignores most dental practices

Five hundred twenty-seven dental practices. One number to start with: 96.2% of them — five hundred seven out of five hundred twenty-seven — got zero mentions from ChatGPT or Perplexity in our June 2026 scan. Not a low ranking. Zero mentions. This video is about what those five hundred seven have in common — and just as important, what they don't. We're not showing you this to scare you. We're showing you because the fix that matters most turned out to be cheaper and less glamorous than most people assume.

Here's the obvious question every practice owner asks first: is it their reviews? Is their website blocking AI crawlers? We checked both. Neither one explains the gap. By the end of this video you'll know exactly which one actually matters, and which one you can stop worrying about.

In June 2026, we scanned five hundred twenty-seven dental practice websites across seven US metros — Austin, Charlotte, Columbus, Nashville, Raleigh, Salt Lake City and Tampa — plus nine smaller towns. Every site was checked the same way: star rating on file, robots.txt access, and structured data — what's called schema markup. Then we asked ChatGPT and Perplexity directly whether they'd name each practice. We picked this spread of metros on purpose, from a fast-growing hub like Austin to a smaller market like Columbus, so the pattern couldn't be explained away as one city's quirk.

Two honest numbers before anything else. ChatGPT returned a usable reading for four hundred fifteen of those practices — the rest hit a rate-limit error, so we dropped them instead of counting them as invisible. Perplexity returned a usable reading for two hundred ninety-seven. A failed request just means the check didn't run — it doesn't mean the practice is invisible.

First assumption: bad reviews. We checked. 96.7% of the practices with a rating on file — five hundred six out of five hundred twenty-three — already sit at 4.3 stars or higher. In SOCi's 2026 local visibility index, the locations ChatGPT actually recommended averaged 4.3 stars — an average of what got recommended, not a cutoff you have to clear. And it's not a thin signal: the median practice among those 523 carries 250 reviews and a 4.9 star average. This isn't a handful of five-star ratings propping up a weak sample.

When 96.7% of practices are already standing above a bar, that bar isn't doing any sorting. Nearly everyone in the room holds the same ticket, so the ticket can't be what decides who gets in. Reviews are not the blocker.

Second assumption: blocked crawlers. Also checked. 93.4% of the practices we scanned — four hundred ninety-two out of five hundred twenty-seven — had robots.txt open to AI crawlers. A small minority were actively blocking one. Separately, in a broader national scan, 11.4% of 6,497 dental sites whose robots.txt we could actually read blocked at least one of thirteen AI and data crawlers we tested. Narrow that to just the crawlers OpenAI, Anthropic and Perplexity actually run, and it drops to 7.6%. Either way, for the overwhelming majority, crawler access is not the blocker.

Look city by city and the same split shows up. In Charlotte, 154 practices, ChatGPT named two. In Columbus, 58 practices, ChatGPT named none. In Austin, 111 practices, Perplexity returned a usable reading for all of them and named seven — but ChatGPT could only get a usable reading for twenty-one of those same 111. In Tampa, 42 practices, ChatGPT named none, and Perplexity couldn't be read at all in that scan. Three cities, three flavors of the same gap, and one pattern underneath all of them: that's two different problems hiding under one word, "invisible." One is "we looked and didn't pick you." The other is "we couldn't even read you" — and that second one is the cheaper problem to fix.

Two suspects, both mostly cleared. So we kept looking — not at what practices say about themselves, but at whether a machine can actually parse who they are. That's where the picture changes.

There's a layer most owners never see, called schema markup. It's a block of code that spells out the basics for a machine — the type of business, its location, the conditions it treats — so an AI doesn't have to reconstruct that from paragraphs of marketing copy. Think of it as a label on a can instead of a description buried somewhere in the marketing on the box — a machine doesn't have to guess what's inside.

53.9% of the five hundred twenty-seven practices we scanned — 284 of them — have no LocalBusiness or Dentist schema at all. That's not a small gap in an otherwise complete picture — it's the majority of the sample, unreadable at this layer, regardless of how good the site looks to a human visitor.

Only 3.4% — eighteen out of 527 — have complete structured data. That's the one measurable trait we found shared by the practices that weren't getting named. Not their reviews. Not their crawler settings. Whether a machine could read them.

One more reason not to treat schema as a magic trick. Earlier this year, a researcher named Mark Williams-Cook ran a small experiment: he planted a fake address inside intentionally invalid schema code on a test page, while nothing in the readable text said the same thing. Both ChatGPT and Perplexity repeated that fake address back to him anyway. What that suggested to him: both assistants were pulling from the words on the page itself, not actually parsing the schema tag sitting underneath it. Schema states facts in a form a machine can use — it doesn't compensate for what's actually written on the page.

Here's what all of this looked like on the engine side. ChatGPT named three of the 415 practices it could read for — 0.7%. Perplexity named eighteen of the 297 it could read for — 6.1%. Three names and eighteen names. That's not enough to build a formula from — which is exactly why this video looks at the majority instead of the winners.

Now the part most agencies skip. We are not telling you that adding schema will get you recommended. We haven't run the kind of controlled test that would prove schema by itself is the reason. What we found is a shared trait among the practices that went unnamed — not a lever with a guaranteed payoff. Correlation, not causation. Anyone who tells you otherwise is selling you a number they can't back up. Selling you a guarantee here would be easy. We're choosing not to, because the data doesn't support one.

Why does that distinction matter to you? Because the honest version is still useful. If more than half of practices are missing basic machine-readable information about themselves, that's worth fixing regardless of whether it guarantees a citation tomorrow. It's infrastructure, not a magic trick. Think of it the way you'd think about keeping your hours and address current on your own website — not because it guarantees foot traffic, but because getting it wrong costs you for free.

So, to recap. 527 practices scanned. Reviews cleared the bar for 96.7% of them. Robots.txt was open for 93.4%. And 96.2% — 507 of them — got zero mentions from either engine. The trait almost all of them shared: 53.9% had no schema markup at all.

Every number in this video, the full methodology, and the raw CSV are open at citevio.com/data. We'd rather be corrected than quoted wrong — if you think a figure here is off, the contact is right there on the page. This channel exists to publish exactly this kind of data: measured, dated, and open to checking — not to promise rankings we can't deliver. Thanks for watching.