What is this test measuring?

Three things, in this order. That an assistant can find your site. That it can understand what you do and who you serve. And that it decides to trust you enough to name you when someone asks.

The order matters because they depend on each other: if it cannot find you, how well your content is written is beside the point. Which is why a low score is almost never fixed by publishing more.

The most useful test you can run today

Open ChatGPT or Perplexity in a fresh conversation and ask something a customer who does not know your name would ask: “which companies help implement a CRM in Bogotá?” or “I need to automate processes at my company, who do I talk to?”

Note which brands come up and in what order. Then ask directly about your company and see whether the description is right. With those two answers you already know whether your problem is that they cannot find you or that they cannot understand you — and those are different problems.

What you can check yourself, with no technical help

  • The first screen. Show it to someone outside the business for five seconds and ask what you do. If they cannot say, that is the first fix.
  • The kind of content. Do your pages answer customer questions or talk about your company? Assistants cite the first kind.
  • The dates. Can you see when each page was published? It is a cheap trust signal plenty of people leave out.
  • The consistency. Are your name, your description, and your city identical on your site, your social profiles, and the directories you appear in? If not, an assistant may think you are two different companies.

What to ask of whoever maintains your site

Three checks you cannot run yourself, but can ask for. You do not need to understand the detail: you need to know they are handled.

  • That the content is in the HTML. Many AI crawlers do not run the programs that assemble the page in the browser. If your text is produced that way, to them the site is empty. It is the most serious failure and the most common.
  • That AI crawlers are allowed in. A file on the site may be blocking them without anyone having decided to.
  • That there is structured data. It is the spec sheet that declares who you are, so the assistant does not have to guess.

And if the score is high but they still do not name you?

Then the problem is not technical but one of coverage: your site can be read and is understood, but it does not answer the questions people ask. The fix is writing content that does answer them, with concrete data, and keeping it up.

It is the slowest of the three, and the one people most often try to skip. That diagnosis sets the order of our work, because investing in content when the problem was access is spending on something nobody will read.