Check whether AI can read your website

Why this is the first thing worth checking

When an AI assistant answers a local question, it frequently quotes the business’s own website. That only happens if it can reach the site, parse it, and find a passage worth lifting. If a robots rule blocks the crawler, or the page builds its entire content in the browser and arrives as an empty shell, then everything else you might do — the reviews, the profile, the local press — is competing without the one source you fully control.

This is also the cheapest category of problem to fix. Most findings are a line in a file or a paragraph moved to the top of a page. None of them require anyone’s permission.

What it fetches, and what it looks at

For one website the engine requests the handful of surfaces that decide the outcome, with a short timeout, a redirect limit and a size cap, under its own declared user agent:

• /robots.txt — the file that decides whether AI crawlers are allowed in at all
• /sitemap.xml — what you are telling crawlers exists, and how fresh you claim it is
• /llms.txt and /llms-full.txt — the emerging convention for describing a site to a language model
• the homepage HTML, exactly as served, before any JavaScript runs

That last clause matters more than it sounds. The engine reads what a non-rendering crawler reads. If your pricing table or your service copy only exists after a script runs, it is invisible here for the same reason it is invisible to the crawler — and you find that out from us rather than from a missed answer.

The five families of checks

How the score works

The score starts at 100 and subtracts by severity, so a number is always traceable to a specific list of findings rather than to a weighting no one can inspect:

Findings are sorted by severity, so the thing actually costing you answers is at the top rather than buried under a list of cosmetic nits. Each one carries a plain-language explanation of why it matters and one concrete action.

Every stored result is stamped with two version numbers: the scoring model it was graded under, and the taxonomy of finding codes that existed at the time. They move for different reasons and both carry a change history, so a score from March is never silently compared against a score from August across a rule change. Your history view marks each past evaluation as comparable to the current model or not.

Find, then fix — and the fix is never the paywall

Detection is entirely deterministic. The rules produce the findings; no AI is in the loop, which is why the same site scored twice in an hour gives the same answer and why a finding can be argued with.

So is the repair. Every finding renders into one of two things, both built from measurements already taken:

• a paste-ready snippet — the markup or config line, with your own measured values already filled in; or
• a written instruction — the same change phrased as a prompt for whichever assistant or developer you already work with.

Because no model is called and no credit is spent, every fix for every finding can be produced at once. Paying cite0 buys the implementation and the ongoing measurement, never access to the answer. If you want to take the list and do it yourself, the list is yours.

Web Readiness compared, feature by feature

The two suites most local businesses already pay for, set against cite0 one capability at a time. Read from each vendor’s own documentation on 25 August 2026; sources at the foot of the page.

cite0 vs Semrush

Semrush crawls at enormous scale and covers AI-crawler robots rules well. cite0 adds the local half on top of that: it tells you that your service pages are missing from your own sitemap relative to the services on your Google profile, that your phone number disagrees with your listing, or that your pricing section arrives as an empty container to a crawler that does not run scripts. Those three are the findings that most often decide whether an assistant can describe a local business at all — and on Boost they are fixed for you rather than added to a list.

cite0 vs Ahrefs

Ahrefs runs one of the best crawlers in the industry and its backlink data has no equal. The gap it leaves for a local business is that its site audit was built for ranking pages, so it checks nothing about AI crawler access specifically — and its AI tracking starts at five prompts, which is roughly one question per service you sell.

What this audit does that a ranking audit cannot

• It reads what a non-rendering crawler reads. The engine evaluates the HTML as served, before any script runs, and flags content that only exists after JavaScript. Most AI crawlers do not execute JavaScript, so a page that looks complete in your browser can arrive empty to them — and no other audit in the tables above will tell you.
• It knows what your business actually sells. Sitemap coverage is checked against the service items on your Google Business Profile, not against a generic template. A missing page is only a finding if it is a page you needed.
• It checks the details AI hesitates over. Whether your title carries your brand, whether your service areas appear anywhere on the page, and whether your name, address and phone match your listings. Inconsistency here is one of the most common reasons an assistant hedges instead of naming you.
• Every score is traceable and every fix is free. Four severity bands, published. Two version stamps, so a score from March is never falsely compared with one from August. Every finding renders into a paste-ready snippet or a written instruction, deterministically, all at once, with no credit spent.
• It is measured, not asserted. On Boost we make the changes, and Radar tracks every week whether AI started quoting you afterwards. The audit and the outcome live in the same subscription.

What the score covers

Web Readiness measures whether an AI crawler can reach, parse and quote your site. That is the half of AI visibility you fully control, and it is the half that must be true before anything else can work. Whether AI then chooses you is decided alongside it — by reviews, listings, local coverage and who else is being cited — which is what Radar and Profile Optimization are for. Buying Boost gets you all three, which is the point.

What does a good Web Readiness score look like?

The number is a way of finding blockers, not a target to optimise. What matters is the severity of what is open, not the total: a site with no critical and no high findings has nothing structurally stopping an AI crawler, whatever the number beside it. Because the scale is published — critical −18, high −8, medium −4, low −2 — you can read any score backwards into exactly which findings produced it, which is the useful direction. And a 100 with an empty pricing section is worth less than an 88 with your prices in plain text, because the score measures what a crawler can read, not whether what it read was worth quoting.

What is the difference between Web Readiness and Content Optimization?

Web Readiness audits your live website and how AI crawlers reach it — the files, the markup, the structure, whether the door is open. Content Optimization takes a single piece of text you paste in, published or not, scores how quotable it is, and rewrites it into answer shape. One is about the site; the other is about the words. You can run Content Optimization on a page that has never been published, and you cannot run Web Readiness on anything that is not live.

Do I actually need an llms.txt file?

We check for it and we weight it near zero, which is the useful answer. llms.txt is a proposed convention for describing a site to a language model; it costs almost nothing to add, so we flag its absence and hand you the file ready to paste. What we will not do is let it inflate your score, because the evidence that it drives citation does not exist yet. That weighting is published like every other, so you can see exactly what a low-severity finding is worth here — which is the point of publishing weights at all. If a vendor tells you an llms.txt file will get you into ChatGPT, ask them what it is worth in their score and whether they will show you.