The Machine Can Find You. It Can't Confirm You're Real.
Machines can reach almost every site we measured. They can confirm barely half of them. That gap is where small businesses lose recommendations they never knew they were up for, and closing it takes filing work, not marketing work.
Two numbers from the same 54 sites
We measured 54 private therapy practice websites across ten US metros. Two of the numbers sit right next to each other and tell opposite stories.
On reachability, the cohort scored 97 out of 100. Robots fetch the pages, the servers answer, no wall stops them. That problem has already solved itself, and most owners still worry about it.
On verifiable identity, the same cohort scored 57.8.
So the machine arrives, reads the page, and then cannot establish who wrote it, whether that person is licensed, where they actually work, or whether the business on this page is the same business it saw somewhere else. It has your content. It does not have you.
What "confirm" actually means to a machine
An assistant recommending a local service is doing something closer to a background check than a search. It wants the same fact to show up in more than one place, attached to a name, without contradicting itself. When it can't get that, the safe move is to recommend somebody else. You are not rejected. You are skipped.
Here is what moved the score in the cohort, and what it costs to fix.
| Signal | What we looked for | Where most sites lost points |
|---|---|---|
| A named human | A real person's full name on the page that sells the service | "Our team of caring professionals" and no names anywhere |
| Credentials in text | Licence type and number written as text, not baked into an image | A scanned certificate, invisible to anything that doesn't read pictures |
| One address, everywhere | The same address and phone on the site, the map listing and the directories | Suite number on one, missing on another, old number on a third |
| Something outside your own site | At least one independent page that names the business the same way | Every claim living only on the domain that makes it |
None of that is content strategy. It's closer to making sure the paperwork agrees with itself.
The part that surprises people
Consistency scored 88.6 in the same cohort, well above verifiable identity. Sites were usually consistent about the little they said. The failure wasn't contradiction. It was silence.
A practice that never states who works there is perfectly consistent and completely unconfirmable. Getting a bit more specific is what moves this number, and specificity is free.
You can watch this happen in about a minute. Ask an assistant to name the owner of your business and cite where it found that. If it hedges, or names the wrong person, or cites a directory you've never updated, you just watched the check fail. Try it on the practice down the road too. That comparison tends to land harder than any score we could hand you.
One caution about that test: assistants sometimes fill gaps with a confident guess. A wrong name is not proof the name exists somewhere wrong. It's proof the machine had to invent one, which is the same problem wearing a friendlier face.
What to do this week
Open your services page and read it as a stranger who has to write one sentence about you, with a name in it. If you can't, the machine can't either.
Then check three things: your name and role in text on that page, your licence or registration written out rather than pictured, and your address matching your map listing character for character. Suite numbers are the usual culprit. So are phone numbers from two offices ago.
If you want to see how this fits with everything else a machine checks, we published the full index of those 54 sites, including the four we couldn't measure at all. And if your worry is the earlier step, whether your pages can be quoted in the first place, that's a different problem with a different fix.
Questions we get
Does this mean I need a schema markup plugin?
It helps, but it isn't the starting point. Structured data restates facts a machine can already read. If the fact isn't on the page in plain text, marking it up gives the machine a cleaner version of nothing.
My credentials are on my About page. Isn't that enough?
Often not. The machine usually judges the page that sells the service, and assistants don't reliably carry facts across your site. Put the name and the licence where the offer is.
I run a solo business and I don't want my address public. What then?
Use a service area and a business registration number instead of a street address, and keep that identical everywhere. The goal is a fact that can be checked against an outside source, not your front door.
How long until this shows up?
We don't know, and anyone who gives you a number is guessing. What we can say is that these are the cheapest fixes on the list, and unlike content they don't need to be redone every quarter.
Is 57.8 out of 100 bad?
It's the average. The median came in lower, at 55, so half the cohort sits below that. The useful read isn't the grade, it's the company: this isn't a race against a standard, it's a race against the practice down the road, and most of them haven't done it either.
One thing worth saying out loud
We sell to this market, so read our numbers with that in mind. We published the method so you can rerun it and argue with the result. The cohort held 58 sites and we could only measure 54, because two blocked us and two never answered. The index names those four rather than quietly dropping them.
If you'd rather have someone go through your own site with this list, book a strategic fit call and we'll tell you which of the four signals you're missing before you spend anything.
