Tool

Website Assistant Readiness Checker

Six checks you can run yourself in about ten minutes. They decide whether an assistant on your site would have anything worth answering from, which is a content question before it is a product question.

Start a public-content pilotSee the bot quality audit

YOU GIVE ITyour public sitereadcomputereturnSIX CHECKS, TEN MINUTESSitemap reachableHelp pages existDocs beyond marketingNot blocked to crawlersPublic enough for a pilotYOU GET BACKsix checks, ten minutes, before you buyRun by a person. Not an automated scan.
  • You run it yourselfTen minutes, no account
  • Content question firstNot a product pitch
  • Nothing is sent to usUnless you want a pilot

Do these yourself

The six checks

Is your sitemap reachable?

Try /sitemap.xml. If it 404s or lists ten URLs for a hundred page site, an assistant will index what it can find rather than what you meant, and the gaps will look like the assistant failing.

Fix the sitemap before training anything on the site.

Do you have help or FAQ pages?

Not a contact form. Pages that answer questions in prose. An assistant retrieves passages; if the passages do not exist, no configuration creates them.

Write the five questions your team answers most, as pages.

Is there depth beyond marketing copy?

Marketing pages describe outcomes. Buyers ask about limits, pricing edges, integrations and what happens when something breaks. If only the first exists, the assistant will be confident about the wrong half.

One docs page on limits is worth ten more feature pages.

Is there a contact or lead path?

When the assistant cannot answer, it needs somewhere to send the person. Without one it becomes a dead end that is worse than no assistant.

A contact page or form the assistant can point at.

Are your pages reachable by crawlers?

Check /robots.txt and any staging or password rules. Content behind a login cannot be a source for a public assistant.

Unblock the pages you want answered from, or accept they are out of scope.

Is the content public enough for a pilot?

A pilot runs on content a visitor could already read. If your best answers live in a private wiki or a support tool, they are out of scope today.

Decide which public pages carry the answers, before anyone configures anything.

HOW TO READ THE RESULT

Unmet rows do not block a pilot. They change what an assistant can answer. A site failing checks two and three will get an assistant that is polite and useless, and no amount of configuration fixes it, because the answers do not exist yet.

Honest scope

What this is

A CHECKLIST, NOT A SCANNER

Nothing on this page inspects your site. These are six things you check yourself, and they are the same six we check first when someone asks for a pilot. Publishing them means you can decide without talking to us, which is the point.

Next

If most of these pass

Start a public-content pilot

Public pages, public docs, approved files. Your data stays out of scope.

Or audit what you already run

If an assistant is already live, the teardown is the more useful step.

Read the pilot rules first

The trust page states what may enter a pilot and who should wait.

Readiness questions

No. It is a checklist you run yourself. Nothing here inspects anything.

That is common and not disqualifying. It tells you which content gap to close first, which is more useful than a score.

You need content that answers questions. Docs are the usual form. A thorough FAQ works as well.

Not for a pilot. If a visitor cannot read it, it is out of scope.

Fix that first. An assistant trained on what a crawler stumbled into will have gaps you cannot explain later.

No. The six checks apply to any retrieval assistant. If you use them and pick another tool, they were still the right checks.

Launch

Ready, or close to it?

A public-content pilot runs on the pages you already have, beside whatever you are using now.