| 开发者 |
navarroido
benwpco llmagnet |
|---|---|
| 更新时间 | 2026年9月30日 04:10 |
| PHP版本: | 7.4 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPLv2 or later |
| 版权网址: | 版权信息 |
/llms.txt and markdown exports so ChatGPT, Claude, Perplexity, and other AI crawlers can understand your WordPress site. It logs all 25 known AI bots locally, scores pages for AI readiness, and offers optional Agent-Ready controls (markdown endpoints, well-known discovery files, robots.txt rules, IndexNow) — all off by default.
Connect GA4 and Search Console from Connections for read-only AI referral metrics. An opt-in MCP read co-pilot can query visibility data; the only write tools are regenerate llms.txt and recalculate the visibility score. Works alongside Rank Math and Yoast.
On LLMagnet admin screens only, an optional news/promo feed can be fetched from llmagnet-wp-news.vercel.app after an administrator approves it. It does not run for site visitors. Details: External services.
/llms.txt and /llms-full.txt are served by WordPress. Markdown for each public post is served at that post's permalink with a .md suffix. A separate plugin, LLMagnet Pro, is sold at llmagnet.com and adds schema, scheduled email reports, catalog discovery, and image export. This plugin does not sell those features or lock any of its own.
Demo: TasteWP
https://www.youtube.com/watch?v=XvE-6SnEX04
/wp-content/plugins/ or install via Plugins → Add Newhttps://your-site.com/llms.txtllms.txt is an emerging standard (like robots.txt for search engines) that helps AI models understand your site structure and content.
No. Documents are built when requested and cached. There is no extra work on the public page for visitors.
ChatGPT, Claude, Perplexity, Gemini, Grok, Bing AI, Mistral, DeepSeek, Llama, and more — 25 bots in total. All of them are detected, logged, and reported by name.
All of them. The Pages screen lists every page that received AI bot traffic in the selected window.
Yes. On WordPress 6.9+ LLMagnet registers its analytics and llms.txt tools as WordPress abilities (under the llmagnet category). Install the official MCP Adapter plugin to let AI assistants like Claude securely query your AI SEO data using an Application Password. On older WordPress versions the plugin works normally; abilities are simply unavailable.
No. Core features work on any WordPress site. WooCommerce commerce analytics and UCP catalog discovery are part of the separate LLMagnet Pro add-on and also need WooCommerce to be active.
It scores published content for AI crawler readiness (headings, meta description, length, image ALT coverage, and related checks). WooCommerce product visibility scores are part of the separate LLMagnet Pro add-on.
LLMagnet ships bundled translations for the plugin admin dashboard (menus, settings screens, React admin UI, and related notices). Supported locales: Hebrew (he_IL), Arabic (ar), and Spanish (Spain) (es_ES). Set your site language under Settings → General → Site Language; WordPress loads the matching translation files automatically. Brand names (LLMagnet) and technical terms (llms.txt, MCP, JSON-LD, product names, and similar) are kept in English. Additional languages can be contributed via the WordPress.org translation project or by adding custom .mo and .json files to the plugin's languages/ folder.
Yes. llms.txt, llms-full.txt, and markdown are served by WordPress for each site. The plugin does not write those files into a shared site root.
LLMagnet serves /llms.txt and /llms-full.txt through WordPress. Markdown is served at {permalink}.md. A physical llms.txt already in the site root, including one written by another plugin, is left in place and is what the web server serves. Older versions of this plugin remove files they created, once, during upgrade.
src/) and build tooling now ship inside the plugin.md endpoint and the MCP content tools already refused protected content, but llms.txt, llms-full.txt and the /llms-docs/ markdown files did not — so a protected post's full body could be written to a publicly readable file. All export queries now exclude password-protected content, the writers refuse it individually, and protecting or unpublishing a post immediately deletes the markdown file it had already exported instead of leaving it until the next daily run.md URL is injected as an encoding / MediaObject field into Yoast SEO's own Article/WebPage schema piece, or into Rank Math's graph, rather than LLMagnet standing down and contributing nothing. Falls back to LLMagnet's own graph, or a minimal standalone block, when neither is present/.well-known/agent-skills/index.json with a $schema identifier and a SHA-256 digest per skill, alongside a real SKILL.md artifact that teaches agents how to read the site as markdown. The digest and the served file are computed from the same source, so they cannot drift apart. The previous /.well-known/agent-skills path keeps working and returns the same index/llms.txt and /llms-full.txt are now served by WordPress when the physical files are missing, so they work on hosts where the web root is not writable. An existing file always takes precedenceContent-Signal directive in robots.txt was emitted after a blank line, which ends a record — so it belonged to no user-agent group and parsers could ignore it. It is now emitted inside each AI crawler groupLink response headers advertise /.well-known/api-catalog and the Agent Skills index, both of which were already being served but never announcedmarkdown_url