AI visits and markdown copies: see which AI tools visit your posts
SEOBetter serves a plain markdown copy of every public post and page for AI crawlers, and counts how many AI crawlers and AI-referred visitors reach each post. No IP addresses, cookies or visitor profiles are stored, only daily totals.
Markdown copies
Every published, public post or page is also available as plain markdown at its own URL, with .md in place of the trailing slash (for example /my-post/ becomes /my-post.md). A query form, ?format=markdown, works at once on any permalink structure. AI crawlers get a clean, token-efficient copy of your content instead of having to parse your theme's HTML.
The markdown copy is marked noindex and carries a canonical link back to the normal URL, so Google indexes only the original page, never the markdown copy.
A post never gets a markdown copy when it is:
- a draft, private, or pending review
- password protected
- marked noindex (by SEOBetter or another SEO plugin)
- restricted by a membership plugin SEOBetter recognizes (Paid Memberships Pro, MemberPress, Restrict Content Pro, or WooCommerce Memberships)
- restricted by any other active membership plugin (markdown copies are off by default for those; a developer can re-enable them with the
seobetter_markdown_copy_eligiblefilter)
Turning it off. Under SEOBetter, then Settings, next to llms.txt, uncheck Serve markdown copies of posts for AI crawlers.
The pretty .md address needs one thing from you. The /my-post.md address works only after your site's permalinks have been re-saved once (WordPress has to register the new URL pattern). SEOBetter shows a notice in the admin area until that happens. The query form, /?p=123&format=markdown, works immediately, with no setup.
SEOBetter's own llms.txt file also tells AI crawlers that markdown copies exist.
AI visits
SEOBetter counts two things, every day, for every post:
- AI crawler visits: a request whose user agent matches one of 18 named AI bots (GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, Claude-User, Claude-SearchBot, anthropic-ai, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended, Bingbot, CCBot, meta-externalagent, Bytespider, Amazonbot, DuckAssistBot and MistralAI-User).
- Visitors referred from an AI tool: a human visitor whose browser arrived from chatgpt.com, chat.openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai, you.com, poe.com or meta.ai, detected from the browser's referrer or a matching
utm_sourcetag.
You see the counts in three places:
- the AI visits row in the GEO score box while you edit a post
- the AI visits column on the Posts and Pages screens
- an AI visits card on the SEOBetter dashboard, with the busiest posts and a breakdown by bot and by referral source
What is stored. Only a daily count per post, per AI bot or referral source. No IP address, user agent string, cookie or visitor id is ever stored with the counts. To avoid counting the same AI-referred visitor twice, rate limiting keeps a one-way hashed marker for one hour, never the IP address itself, in your site's own database, and then it expires.
How referred visitors are counted. A small script on your post pages checks whether the visit came from a known AI tool, and if so sends a single count to your own site's server. Nothing is sent to any external service.
How long counts are kept. The free plugin keeps 30 days of history. SEOBetter Pro keeps 12 months, with a 90-day and a 12-month view on the dashboard card.
Why a count can look low. If a full-page cache serves the page without WordPress running, an AI crawler's ordinary page view can be missed. Markdown-copy requests and AI-referral counts are not affected by page caching, so they always count.
The AI crawler access audit now checks 18 bots
The AI crawler access audit (SEOBetter, then AI Crawler Access) checks robots.txt, firewall rules and noindex signals against the same 18 bots AI visits tracks, up from the original 8. The audit's one-click robots.txt fix still allows only the original 8 bots (GPTBot, ChatGPT-User, Google-Extended, ClaudeBot, anthropic-ai, PerplexityBot, Bingbot and CCBot) - running it again does not widen what your robots.txt already allows.