AI visibility scan

The AI visibility scan sits on the Search & AI tab of the Overview, below the AI assistants card. That card shows what AI took from your site and what it sent back. The scan looks for what could be getting in the way.

Run it

Administrators see a “Page to scan” box, filled with your home page. Enter a path such as /about/ or a full address on your own site, then click “Run the scan”. Addresses on other sites are refused. Other users see the last result, or a note that an administrator can run the scan.

Your server fetches the page, robots.txt, llms.txt and your sitemap from your public address, logged out, the way a crawler does. It also requests the page as GPTBot, ClaudeBot and PerplexityBot. A scan takes up to half a minute. If your site is slow enough that the scan runs out of time, the checks that didn’t run are marked “Not run” and left out of the score.

Only one scan runs at a time. If you start a second within a few minutes, you get a note to wait. The last result is saved and shown with how long ago it ran and which address was scanned.

The scan’s own requests carry a one-time token. They’re left out of the crawler figures, the missing pages report and your pageviews.

The score

The result is a score out of 100 and a percentage for each of four groups. Each check passes, warns or fails, and some count for more than others. A pass earns full points, a warning half and a fail none. A score of 80 or more is shown as good, 50 to 79 as middling and below 50 as poor.

Failed and warning checks come first in each group, and each one says what to change.

AI crawler access

• Whether robots.txt lets AI search crawlers in: ChatGPT search and browsing, Perplexity, Claude search, Google and Bing. Blocking these keeps you out of live AI answers, so it fails.
• Whether robots.txt lets AI training crawlers in: GPTBot, ClaudeBot, Google-Extended, Applebot-Extended and Common Crawl. Blocking them is a legitimate choice, so it only warns.
• Whether the page is allowed in search results. A noindex tag, a “noai” tag or WordPress’s “Discourage search engines” setting counts against it.
• Whether your server answers GPTBot, ClaudeBot and PerplexityBot. A firewall, CDN or security plugin can refuse them even when robots.txt allows them. Hosts that verify real crawlers by IP address may treat the scan’s test requests differently from real ones.

Content AI can read

• Text in the raw HTML. A page that builds its content with JavaScript looks nearly empty to most AI crawlers.
• One clear main heading, and content split into sections under descriptive H2 headings.
• Answers to the questions people ask, such as an FAQ with FAQPage markup.
• Alt text on images.
• A language declared on the page.

Structured data

• Business identity markup: Organization or LocalBusiness JSON-LD, and sameAs links to your other profiles.
• Contact details as tel: and mailto: links or in your markup.
• Open Graph tags for title, description and image.
• A canonical URL.

Discovery files

• An XML sitemap, found through robots.txt or at the usual addresses.
• An llms.txt file. It’s optional: a plain-text file at /llms.txt that says what your site is and links to your most useful pages.

Pro

Vireo Pro runs this scan every week on your home page and most viewed pages, charts the score over 26 weeks, ranks the fixes by the traffic they affect, and sends an email, Slack message or webhook when a crawler access check starts failing or recorded AI crawler requests drop off.