Could an AI assistant actually use your website?
It’s not enough for AI to be allowed onto your site — the real question is whether it can DO anything once it’s there: find your pages, understand your forms, and complete an enquiry. This deep check reads your forms field by field the way an AI assistant would, then scores four things live. Most tools stop at the surface. This one tests whether your forms actually work for agents.
We test whether an agent can act, not just read
Reading your page is the easy part. Completing your enquiry form is where AI assistants succeed or give up — so that’s where we look hardest, measured live the moment you run the check.
We fetch your homepage and contact page
In one pass we pull your live pages the way an AI assistant would — the homepage plus whichever page your “contact”, “enquiry” or “get a quote” link points to, so we test the form that actually matters.
We read every form the way an agent would
Field by field: does each one have a name it can send data to, and a label so it knows what to type? Does the submit button have a readable name? Are the WebMCP annotations valid? This is the depth surface-level checks skip.
We check the front doors and the fallback
We probe for agent front doors — llms.txt, MCP and NLWeb endpoints — and confirm your content still reads without JavaScript, because many AI assistants never run it. Then we score all four, live.
The enquiry an agent couldn’t finish
Search sends the visitor; the agent completes the task
AI assistants don’t just read your page — they act on it. If they can’t operate your enquiry form, the task fails on your site, and the user is handed to a competitor whose form they can complete.
A form an agent can’t fill is a lead you never see
An input with no name can’t receive data. A field with no label leaves the agent guessing. A submit button with no readable name can’t be pressed. Each gap quietly turns an AI-driven enquiry into nothing.
Everyone else stops at the surface
Other readiness checks confirm a standard exists — a file is present, a tag is there. We go a level deeper and test whether the form is genuinely operable, field by field. We’ve not seen another free readiness check go this deep.
Fair questions
How is this different from an AI-visibility or robots.txt check?
Those answer whether an AI is allowed to read your site. This answers the harder question of whether an AI assistant could actually USE it — find its way around, fill your enquiry form, and complete the task. We inspect every form field for a name and a label, check the submit button can be identified and pressed, validate any WebMCP tool annotations, and confirm your content reads without JavaScript. It’s operability, not just permission.
What is WebMCP, and do I need it?
WebMCP is an emerging way for a website to hand an AI agent explicit, callable tools — for example “submit this enquiry” — instead of making the agent guess from your page. It’s currently exercised through an early Chrome origin trial, so almost no site has it yet. That means you do NOT need it to score well here, and its absence is never counted as a failure. We flag it as an early-mover opportunity: adding it now puts you ahead of nearly everyone.
Why do form labels and field names matter so much for AI?
An AI assistant operates a form through its structure, not its looks. The field’s name attribute is how data is actually sent, so an unnamed field can’t receive anything. The label is how the assistant knows what belongs in each box. And the submit button needs a readable name so it can be found and pressed. A form can look perfect to a human and still be impossible for an agent to complete if those three things are missing.
Will this change or store anything on my site?
No. It only reads a handful of your pages and files the same way an AI assistant would — your homepage, contact page, robots.txt, llms.txt and a few well-known endpoints. Nothing is changed, installed or stored on your site, and we make no more than about a dozen quick read-only requests.
Do I need to enter my email?
No. This check is deliberately open — run it as many times as you like. If you want the full picture across every AI-visibility category, that report is one click away.