AI visibility
Being invisible to an assistant is a new way to lose a customer.
A growing share of "who should I call" now gets answered by something that read the web on the customer's behalf. Two things decide whether a business is in that answer, and most tools measure neither of them separately.
Two questions, not one
They fail independently. A site can be perfectly quotable and blocked at the door, or wide open and have nothing on it worth quoting, and the fix for each is nothing like the fix for the other.
Can it reach and read you?
Whether robots.txt admits GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended and the rest; whether there is structured data describing what the business is; whether any of the content is shaped as an answer to a question; and whether an llms.txt exists at all. A site that blocks the crawlers cannot be recommended however good it is, so that finding caps the category rather than averaging into it.
Would it cite you?
Whether the page says who wrote it and when, whether it cites anything itself, whether its structure is extractable as lists and tables rather than buried in prose, whether the business is a clearly defined entity, and how far a reader has to travel from the heading before reaching something quotable.
The uncomfortable part
What we will not claim to measure.
This is the fastest-moving corner of the category and it is full of numbers nobody can produce. Here is where the line is, and it is on the report as well as on this page.
- Not share of AI answers
- We do not measure how often an assistant mentions a business, because that cannot be measured from outside it. Anyone selling you that number is selling you a sample of prompts they chose.
- llms.txt is a weak signal
- It is cheap and we check for it, and we say plainly on the report that no major assistant has committed to reading one. Selling a client an llms.txt file as an AI strategy is selling them a deliverable, not an outcome.
- Crawler access is not consent
- We report what robots.txt says. Whether a business wants to be read by an assistant is a commercial decision, and some of them should say no.
- The score moves
- The signals behind this are two years old and still changing. Anything here that stops being true will be changed on the report first and argued about on the blog afterwards.
What an agency does with it
It is a finding they have never heard
Every business has been told about page speed. Almost none has been told that its robots.txt is turning away the crawler behind the assistant their customers ask. That is a conversation opener no competitor is using yet.
It is fixable in an afternoon
Most of what is wrong here is a robots.txt line, a schema block and a heading rewrite. Copy-ready snippets for all three are attached to the finding, so the work is scoped before the call.
It is a retainer, not a project
The signals move, which means the audit is worth re-running. Put the site on a monitor and the change alert is the reason for the next conversation.
See what an assistant can see
Point it at any address. The AI-visibility section is in the free audit, alongside everything else.