RowanWeb Markdown for AI – .md Pages & llms.txt
| 开发者 |
ryabininvitaliy
freemius |
|---|---|
| 更新时间 | 2026年9月27日 14:05 |
| PHP版本: | 8.0 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPL-2.0-or-later |
| 版权网址: | 版权信息 |
详情介绍:
AI assistants and RAG systems increasingly fetch site content directly, but they have to work through your theme's HTML markup to do it — extra tokens, extra noise, and a real chance of misreading the actual content. This plugin gives them a clean alternative:
{permalink}.md— a markdown version of every public page, generated from the same rendered content your visitors see (post_content, blocks, shortcodes) and converted with a real HTML→Markdown parser, not a regex stripper. Headings, tables, links and images keep their structure./llms.txt— a dynamic index of public pages grouped by content type, with a short description for each. Every entry links both to the normal HTML page (so an AI citing your site points at a page a human can actually read) and to its.mdversion (for efficient re-fetching)./llms-full.txt(optional, generated on demand) — a single file with the full markdown content of your most recent posts per content type, for tools that want more than the short index. Generated with one click on the Diagnostics tab; after content changes, an admin notice reminds you to regenerate it.- Deep custom-field support — if Advanced Custom Fields is active, the plugin automatically finds and renders every ACF field used on a page (any type, any nesting — repeaters, flexible content, groups) without any per-site configuration.
- Privacy-safe by default — password-protected posts, drafts, and anything flagged noindex by your SEO plugin are never served as markdown. Per-post exclusion is also available from the editor sidebar.
- Plays well with other SEO plugins — if Yoast SEO or Rank Math already generates its own
llms.txt, this plugin automatically steps aside instead of racing for the same URL. - Pre-warmed, not lazy — markdown is rebuilt the moment you save a post, so the first crawler to arrive never waits on a cold cache. A bulk warm-up button covers existing content after activation.
- Bot activity log — see which AI crawlers actually visit your
.md/llms.txtURLs, how often, and which exact pages, right in the Activity tab. - Multilingual-aware (Polylang) — on a multilingual site,
llms.txtandllms-full.txtsplit into one section per language instead of mixing everything into a single list under one heading; custom-field labels are automatically translated to match each page's own language.
安装:
- Upload the
rowanweb-markdown-for-aifolder to/wp-content/plugins/, or install the ZIP via Plugins → Add New → Upload. - Activate the plugin.
- Go to Settings → RowanWeb Markdown for AI and pick which content types should be included (posts, pages, and any custom post types you use).
- That's it —
{any public URL}.mdand/llms.txtare live immediately.
屏幕截图:
常见问题:
How do I check that the markdown version works?
Open any page on your site and add .md to the end of the address (remove a trailing slash first, if there is one) — you should get plain markdown text instead of the normal page.
Why doesn't /llms.txt open, or it shows content I didn't expect?
If your SEO plugin (Yoast SEO or Rank Math) already has its own llms.txt generation enabled, this plugin intentionally stays silent to avoid a conflict — the Diagnostics tab shows which plugin is currently serving it.
Where does the list of known AI crawlers come from?
It's a snapshot of the community-maintained ai.robots.txt catalog (MIT License), bundled with the plugin and refreshed with each plugin release. Nothing is downloaded at runtime.
Does this affect my normal search engine SEO?
No. Google, Bing, and other conventional search engines are untouched — this plugin only adds AI/LLM-specific endpoints alongside your existing setup.
Does the plugin send any of my content anywhere?
No. Page content is only ever served in response to a normal HTTP request, exactly like any other page on your site. The plugin makes no outgoing connections on its own. The only exceptions are listed under "External services" below, and both are off until you turn them on: the DeepL translation of custom-field labels (only if you add your own DeepL API key) and the Freemius opt-in.
更新日志:
1.2.0
llms-full.txtis now stored in the uploads folder (wp-content/uploads/rowanweb-markdown-for-ai/) instead of the site root, and served at/llms-full.txtthrough WordPress. Requests to it now show up in the Activity tab.- The list of known AI crawlers is now bundled with the plugin; the free version no longer downloads it from GitHub.
- Detection of a physical
llms.txtleft by another plugin now also works when WordPress is installed in a subdirectory.
- Automatic background regeneration of
llms-full.txtis now part of Pro. The free plugin generates it with the button on the Diagnostics tab and shows a reminder notice when content has changed since the last generation. - Pro: the "llms-full.txt may be out of date" notice no longer appears, since the file is regenerated automatically.
- Renamed to "RowanWeb Markdown for AI" (new slug:
rowanweb-markdown-for-ai). - Settings-page CSS and JavaScript are now proper enqueued files instead of inline blocks.
- Stricter sanitization of request headers stored in the bot activity log.
- The readme now documents every external service the plugin connects to.
- Activating the plugin now redirects straight to its Settings page (skipped for bulk/network activation and WP-CLI, so it never hijacks an unrelated screen).
- Added the missing Russian translation for the 1.0.4 "View as Markdown" badge toggle strings.
- Added a Settings toggle to hide the floating "View as Markdown" badge for site visitors. AI/search crawlers are unaffected either way — they discover the
.mdversion through the page's<link rel="alternate">tag and robots.txt, not through this badge.
- Fixed the two scoped
X-Robots-Tagheaders on.mdpages overwriting each other (PHP'sheader()replaces a same-named header by default) — only thebingbotrule was ever actually being sent. Both now use$replace = falseso Google and Bing each get their own rule.
- The
.mdX-Robots-Tagheader is now scoped togooglebot:/bingbot:instead of a barenoindex, so Google/Bing search still drop the page while AI crawlers that build their own retrieval index (OAI-SearchBot, PerplexityBot, and similar) see no restriction.
.mdpages now sendX-Robots-Tag: noindex, followand a canonicalLinkheader pointing at the original HTML page, so the markdown twin doesn't end up duplicated in Google/Bing's search index while staying fully fetchable by AI crawlers.
- Initial public release:
.mdmarkdown pages, dynamicllms.txt, an optional full-contentllms-full.txtdump, deep ACF field rendering, privacy controls, bot activity logging, multilingual (Polylang) support, and a Pro tier (Cloudflare cache purge, higher llms-full.txt limits, configurable hit-log retention with CSV export, rate limiting, suspicious-activity detection, and deeper Activity-tab reporting).