Why a page can be invisible to AI answers
To use a page, an AI tool must fetch and read it. A robots.txt rule written years ago may block the bots that AI search uses. A page built entirely in JavaScript may arrive as an empty shell. A product page with no headings and a few lines of text gives an answer little to quote. These problems do not appear in a visibility chart, yet each can exclude your page from an answer’s sources.
Page checks look for these problems one URL at a time. SearchSeal fetches the page HTML, your robots.txt and your llms.txt, then runs six checks. No AI model is involved, so the same page gives the same result each time.
The six checks
Checks pass, need a fix or cannot be judged. A pass earns the check's full weight, a fix earns half and the score adds up to 100.
- AI crawlers (35 points). Whether robots.txt blocks the search bots that AI tools use from reaching this page. If robots.txt is unreadable, the check stays unknown and is excluded from the score.
- Schema markup (20 points). Whether the page says who you are with Organization or LocalBusiness markup, and whether question and answer pairs on the page are marked up as an FAQ.
- Page structure (20 points). Exactly one H1, at least three headings and at least 150 words of text in the HTML that a crawler receives.
- Title and description (10 points). A title of 60 characters or fewer and a meta description of 160 or fewer.
- Social card (10 points). An og:image tag, so the page shows a picture when shared.
- llms.txt (5 points). Whether your site publishes an llms.txt file at its root.
From a failed check to a fix
Every check that fails becomes a fix in the fix queue, tied to that page and marked high, medium or low impact. Schema fixes give you a JSON-LD snippet to paste. Structure and title fixes show the counts behind them, such as the number of H1 headings or the length of your title. An llms.txt fix covers the whole site once, rather than each page.
Crawler blocks work slightly differently. SearchSeal already checks robots.txt for your site every day, so a blocked page links to that site-wide fix instead of opening a second one. A recheck closes the fix automatically if the problem is gone. If it fails again, the fix reopens. Mark a fix done by hand and it asks for a recheck to confirm.
Run checks when you need them
Run page checks when you need them. Start one from the Actions page, from the REST API or from an MCP client. Checks are limited to your brand's own domain and its subdomains, so you cannot use them on a rival's site.
Each plan includes a monthly allowance of page checks: 200 on Starter, 500 on Pro, 1,000 on Max. The allowance resets on the first of each month in UTC. Failed checks do not count, and neither does the automatic check SearchSeal runs on a live page when you mark an article as published. Repeat a check within a minute to get the earlier result without using another check.
What a page check does not cover
A page check reads your server’s HTML. It does not run JavaScript, so content that appears only after scripts load will not be counted. That matches how many crawlers read a page, but it can make a script-heavy page score lower than it looks in a browser.
It does not measure page speed, mobile layout, broken links or the quality of your writing. A high score means the page is readable and well marked up. It does not mean an AI tool will cite it. Clear technical blockers with the score, then use saved answers and their source links to decide what to write.
If you want to try a single URL before signing up, the free AI readiness check runs the same six checks without an account.