What we check, and why it matters to a machine

An assistant answering “where should I go tonight” has seconds and no browser. Everything below is something it either finds on your site or does without. The audit reports each one as passed, failed, or — honestly — not testable.

Two small machines, drawn as children. Everything below is that same gesture written down for one: here is who we are, here is what we have, please come in.
Group one

Can they get in at all?

Most sites that are invisible to assistants are not hiding on purpose. They are turning the right visitors away by accident.

robots.txt

The file every crawler reads first. We check whether you have one, and then which of the 166 known AI crawlers it lets in — because the ones that answer questions and the ones that collect training data are different bots, and most sites block them as a single group without meaning to.

bot walls

Security challenges and “prove you are human” pages. A person clicks through them. An assistant sees an empty page and moves to whoever is next on its list.

https, redirects

Whether the address resolves cleanly and securely. Assistants weigh this the way search engines have for a decade.

Group two

Is there anything there when they arrive?

This is where beautiful sites fail hardest. The page a visitor sees and the page a crawler receives can be two different documents.

rendered text

Many modern sites build their content with JavaScript after the page arrives. Crawlers do not wait. We measure how much of your text is present in the raw page — on two of the restaurants we scanned this month, the answer was under two hundred characters.

plain-text facts

Whether your phone number, address, opening hours and prices exist as readable text rather than inside an image or a booking widget. An assistant cannot read a picture of your hours.

content depth

How much you actually say about what you do — services described, questions answered, a menu with words on it. This is the score that decides whether the rest is worth fixing.

Group three

Do they understand what they read?

This is the layer we install. It is invisible on the page and it is the difference between being read and being understood.

business data

A short block of structured facts, written in the format assistants and search engines both parse: what kind of business this is, where, when it opens, what it charges, where the menu or service list lives. We check for seven fields and it is common to find none.

the right label

Declaring “a business” is not the same as declaring “a dental clinic”. The specific label is what decides whether you are considered when someone asks for your exact kind of thing.

llms.txt

A short summary of the business written for assistants rather than for visitors — a new and still optional convention, cheap to add, and almost nobody has one yet.

sitemap, titles

Whether your pages can be found at all, and whether each one says what it is. Old ground, still frequently broken.

Group four

And then: do they recommend you?

The first three groups are about your site. This one is about the answer a customer actually gets.

the questions

Twenty-four questions per assistant, written for your category and your town — best, nearest, open late, cheapest, good with children, and three that name you directly to see whether you are known at all.

who else

Every other business named in those answers, counted. This is usually the part owners read twice.

the evidence

The assistant's own words around each mention, quoted in the report, so nothing rests on our summary of it. Where a mention is ambiguous — a business with your name in another state — we say so instead of counting it.

Two things we cannot promise

Assistants change their answers on their own, week to week, and no one controls what they say. We can make your business legible to them, and we can measure the difference month after month. We cannot guarantee a place in an answer, and anyone who does is guessing.

One audit is one reading. A single run tells you where you stand today; the pattern only appears once there are several, which is the whole reason the monthly watch exists.

Send your address and your town. We will run all of it and send you the report.

[email protected]