AI Readiness checks

Before anything else: may AI search engines read your site? S6S reads your robots.txt for the three AI search crawlers that feed the engines it checks: OAI-SearchBot (ChatGPT), PerplexityBot and Googlebot (Google Search, which AI Overviews and AI Mode draw on). The rules are checked for your homepage and for the other pages of your site S6S found in your menu and sitemap.

• If any of the three is blocked from your homepage, your grade is held at D or lower until it's fixed, and unblocking it is the first task in your Fix Plan. The block adds no points of its own, so it can never raise your grade or give you one you didn't have.
• If one of them may read your homepage but is blocked from other pages, the row says which pages and shows “Limited” instead of a check mark. It doesn't change the grade.
• Google-Extended is not only a training control: Google uses it for whether the Gemini app may use your pages to ground its answers. Blocking it doesn't affect Google Search or AI Overviews. The row shows it as “Limited”; it doesn't change the grade.
• AI training crawlers (GPTBot, ClaudeBot and others) can be blocked without any effect on the grade.
• robots.txt is only part of access: a firewall rule aimed at a crawler can't be seen from outside, and an allowed page is not proof that the page is in an engine's index.
• robots.txt is read the way crawlers read it (RFC 9309): the group naming the crawler wins, the longest matching rule wins, and Allow wins a tie.
• No robots.txt at all means every crawler is allowed. A robots.txt S6S couldn't read just then (a server error, rate limit or timeout) is “not measured”, never “blocked”.
• S6S fetches your site as S6S-Auditor/1.0. If your firewall turns that request away, the Report says it was S6S's request that was refused, not an AI crawler.

Each Run Check reads your site again: the homepage and up to 10 of your pages. A page S6S couldn't read counts as unknown, never as a fail. When the homepage itself can't be read, the Report says what happened (for example “your site didn't let us read it (HTTP 403)” or “your site didn't answer”), and the website checks stay unknown until a later Run Check can read it.

Ready for AI agents

Shown as “N of M in place” (checks S6S couldn't read are left out), for reference only: no AI search engine has said these change its answers, so they are never graded and never create a Fix Plan task.

• A page that doesn't exist answers with a real “not found” (404 or 410).
• Pages can be served as Markdown for agents.
• robots.txt declares Content Signals in every group.
• robots.txt names at least one AI crawler.
• An /llms.txt file is served as text.
• The homepage registers WebMCP tools.
• An MCP server card is published under /.well-known/.

Sites your AI answers cite

From each check's answers, S6S lists which sites each engine cited for which question: directories, review sites, roundups, forums, videos, reference sites, your own site and competitors' sites. After the check it reads the most-cited pages (up to 10) and asks three things: are you named on the page, is what it says about you flagged as wrong, and is the page current (dated within about 18 months)?

A cited page is a listing when it names two or more of the businesses the answers named, reads as a list, or its address or title names your kind of business and your town (“…/almere/aankoopmakelaar/”), even when the site blocks the read. A page that answers with a “please wait, are you human?” screen is not read, so your presence there is unknown, never “not on the page”. A widely cited listing that couldn't be read gets a Fix Plan task to check it.

Reputation counts whether you are named on the cited pages S6S read. Each page read counts for the question-and-engine pairs that cited that exact page (since 11 Oct 2026; before, for every pair that cited any page on the site). A pair counts once however many answers cited the page, with or without a text highlight in the link, or the address the page redirected to when S6S read it. Other variants of an address (http or https, with or without www) count as separate pages unless the read shows they are the same, and the site's pages that weren't read count for nothing either way. Being on those pages doesn't guarantee a mention. S6S does not count backlinks.

Under each site you see the pages the answers cited, the buyer questions and engines that cited each one, and whether that exact page was read. The Report says how many cited pages were read and how many question-and-engine pairs those reads cover; a page that wasn't read is unknown, never “not on the page”. On a page that was read, it says whether your name sits in the page's main content or only in its navigation or footer. That says where the name is, not what the page says about you. Pages marked up as articles, trade bodies and research pages are labelled, with where the label came from (the site's address, or the page's own markup or name) and the words around your name; press-release copies are marked as not independent. A label describes the page; it doesn't show the page is independent, relevant or credible (a page marked up as an article can be sponsored, or written by a business that sells in your market). None of this is graded.

Your pages per buying situation

For each buying situation your questions cover, S6S looks through the pages it read on your site for that situation's own words and quotes the passage that matched. A match makes a page a candidate; it doesn't show the page answers the question. For the place, price and booking the questions ask about, it says what your pages state: a place is shown as mentioned where a page names it (your business name doesn't count), which shows the page names the place, not that you offer the service there; prices or opening times are shown as stated, because S6S can't tell whether they fit what a buyer wants. When no page matches, that is about the pages S6S read (up to 10, and only the start of a long page), not proof that your site has no page for it. Not graded.

For local businesses

• Your Google rating and number of reviews. They count once, here, and nowhere else.
• Google Business Profile checks: found, claimed, main category, opening hours, website, phone number matching your site, recent reviews, photos, booking link, owner replies and description. The number of reviews is listed there too, but graded only once, as the review count above.
• Reserve with Google: if you book through a Reserve with Google partner (Salonized, Treatwell, Fresha, Booksy, Setmore), the check is whether booking works inside Google itself. It does when your profile's booking button is Google's own booking page, or your booking tool's link that Google's booking connection put there. It replaces the booking-link check.
• Your hours on booking sites: your opening hours on your own Treatwell, Fresha or Booksy page, compared day by day with your Google profile. Shown for reference, not graded; a mismatch becomes a Fix Plan task.

Google & Maps presence (on request)

For local businesses, a separate grid shows where you appear in Google Search (page 1) and Maps (top 3) from 9 points around your Google listing, 2 km apart, for 1 to 6 searches. It runs only when you ask for it, after a free quote, costs 5 credits (charged only if at least one point answers), and can run 3 times per brand per day. It is a search ranking, not an AI answer: it feeds Reputation in AI Readiness and never the AI Visibility score.

You can upload your own numbers: the generative AI performance export from Google Search Console and the AI performance exports from Bing Webmaster Tools. S6S stores a summary, not the files. They appear in their own “Your own data” section and never change your AI Visibility score or your readiness grade.

• Every Run Check refreshes all of the above.
• The free “Check readiness” button on the AI Readiness card re-checks site access, your homepage and your reputation sources without asking any AI engine.
• The website read, page speed and agent checks run only with Run Check. “Check readiness” keeps their last results until the next Run Check reads the site again.