內容簡介
LLMs.txt Curator 外掛讓您掌控網站對 AI 系統及其他機器的資訊傳遞。它自動發現並發布網站的公共內容，並提供編輯功能，讓您可以審核、描述、排序及標記重要內容。
【主要功能】
• 自動發現網站公共內容
• 支援 Markdown 頁面快照
• 提供內容檢查報告
• 允許機器只讀訪問
• 實時更新快照狀態
外掛標籤
開發者團隊
原文外掛簡介
LLMs.txt Curator gives you control over what your website tells AI systems and other machines about itself.
It finds the public content on your site, publishes it to llms.txt and llms-full.txt, and then lets you curate that representation: review what was found, write descriptions, arrange the order machines read, mark what matters most, and ask Curator why any individual page is or is not included.
Version 2.0 is a rebuild around that idea. Every page, custom URL and fact is a Resource with its own editorial state and a recorded origin, so Curator can always answer three questions: what am I publishing, where did it come from, and why.
New in 2.2
Markdown pages that match the page. With Markdown pages on, Curator takes a snapshot of each page in your llms.txt as a logged-out visitor sees it, in the background, and builds the Markdown from that. Content added by page builders and block templates, which is not in the stored post, is now included. Menus, headers, footers, comments, share buttons and “More posts” lists are left out.
A better converter. Tables come out as tables, code keeps its indentation, numbered lists keep their numbers and nested lists stay nested. The same converter writes llms-full.txt and every .md address, so they always agree.
See what machines receive. Open “Markdown” on any page in Your llms.txt to see its Markdown version laid out for reading, exactly as served, and how it was made: whether it came from a snapshot or the stored content, which part of the page was used, what was removed, and what was kept. “Take a snapshot now” refreshes it on the spot.
Checks for every page. A new Pages group in Checks reports pages built from stored content because a snapshot failed (and why), pages with very little text, unrendered shortcodes, leftover markup, repeated text, uneven tables and menus that leaked into the content. “Check what visitors receive” now also checks a sample of .md addresses for the right type, the noindex and canonical headers, and revalidation.
Never out of date. A snapshot is retired the moment its page, its approved comments or a synced pattern it uses changes, and the stored content is served until a new one is taken. When a post is unpublished, made private or given a password, every snapshot that links to it is retired too, so it leaves lists of posts straight away. Changes to your theme, templates, menus or plugins renew snapshots in the background.
Answer requests for Markdown at the page’s own address (optional, off). A request for the normal page that asks for text/markdown in its Accept header gets the Markdown version. These replies are never stored by shared caches, and “Check what visitors receive” tells you if a CDN mixes them up.
Read-only access for tools and agents. Through WordPress’s own Abilities API, signed-in tools can list what your llms.txt publishes, get a page’s Markdown version exactly as its .md address serves it, and ask what changed since they last looked. With WordPress’s MCP Adapter installed, AI clients such as Claude can use the same abilities. Nothing they do can change your curation, settings or what you publish, and they see only what you already publish. A new Agents group in Checks shows whether this is working.
New in 2.1
Notes for machines: your own answer to “What would a machine get wrong about your site?”, published as written.
Checks: Structure, Content, Delivery and Ownership. Observed facts, not a score.
Ownership: anything else that answers, or could answer, /llms.txt.
Bring in an existing llms.txt: staged for review, approved in one step, undoable.
Also: links to Markdown pages, the Optional section, Site Health, a block editor panel and the Abilities API.
What 2.0 gives you
Automatic discovery, human curation. Curator finds your public content and publishes it straight away, marked as found automatically rather than reviewed. You promote what you have checked, exclude what you do not want, and nothing waits on you to be represented.
Your llms.txt. Arrange sections and pages with exactly what machines receive shown beside you as you work, in readable form or as the Markdown that gets published. Drag if you like; every move is also a button.
Why? Open any resource and see the decisions behind it, separated into what Curator observed about the page and what you decided, ending with the exact line it contributes to the file.
Representation coverage. Not a score. Curator tells you whether it can account for every resource it found: represented, excluded by you, unavailable to visitors, or not yet checked. The goal is zero unexplained.
Unpublished changes. When your curation differs from the published file, Curator says so, shows what changed including the old and new wording, and publishes when you are ready.
Check what visitors receive. After publishing, Curator looks at your public file and tells you plainly whether visitors get the new version, or whether a cache or an old file is serving an older copy.
Markdown pages (optional). Switch on to offer each page in your llms.txt as Markdown at its address plus index.md, for example /about/index.md, following the llms.txt proposal. Off by default.
Suggested descriptions, honestly labelled. Curator can fill empty descriptions from what the page already publishes (schema, SEO meta, excerpt, Open Graph, page content) and tells you which one it used. A suggestion stays marked as a suggestion until you accept or edit it. Curator does not write text for you.
Site details. Describe the organisation behind the site once, and Curator publishes it above the first section. Where WordPress or your SEO plugin says something different, Curator shows you both and publishes your choice.
Facts and custom URLs. Publish a short factual statement that does not deserve its own page, or a URL Curator did not find on its own. Curator stores and represents a URL; it never fetches it.
Safe by default
Both files are generated as a logged-out visitor, not with the permissions of the administrator who publishes.
Drafts, private pages, password-protected pages and noindex pages are left out, and so are pages restricted by Paid Memberships Pro, MemberPress, Restrict Content Pro or WooCommerce Memberships. Any other membership or access plugin can mark content as restricted through the llmscu_is_restricted filter. Curator explains each exclusion.
An llms.txt this plugin did not create is never overwritten.
No AI ranking claims, no scores, no data sent anywhere. The only requests Curator makes are to your own site: to check what visitors receive, and, with Markdown pages on, to take snapshots of your own public pages in the background. Snapshots come from your server’s own address, so if your site shows extra content to requests from the server itself (an IP allow list, or a maintenance mode that lets the server through), leave Markdown pages off.
What llms.txt is, and what it isn’t (June 2026)
Being straight with you matters more than overselling. Google Search ignores llms.txt: it will not help or harm your Google rankings, and Google confirmed in June 2026 that it is fine to maintain the file for other systems that do read it. So this plugin is not an “AI ranking” tool.
Where the file genuinely earns its place: on-site navigation for agents already on your site, the surfaces that actually read it (such as Perplexity and coding agents), and passing Chrome’s Agentic Browsing audit. The honest, useful question is not “does this rank me” but “which AI bots are actually reading my file”, and this plugin answers it from your own server logs.
What makes this different
Most llms.txt plugins treat the file as a static output. LLMs.txt Curator treats it as a living asset:
Description coverage: Curator shows how many published resources have a description and which still need one. It is a count, not a score.
Suggested descriptions: fills empty descriptions from what the page already publishes (schema, SEO meta, excerpt, Open Graph, page content), names the source it used, and leaves each one marked as a suggestion until you accept or edit it. Never overwrites what you wrote.
Change detection: the header tells you when your curation differs from the published file, and Publish brings it up to date.
Per-Page Title Override: set a different title for each page specifically for AI consumption, without touching your on-site SEO.
Safety checks: Curator will not publish a file with missing or duplicated pages or broken formatting. Includes an on-demand format check with plain-English results.
AI Crawler Analytics: see which bots visited in the last 7 days, with a visual bar chart. OpenAI, Anthropic, Perplexity, Google: know who is actually reading your file.
Core features
Visual composition: arrange sections and pages in Your llms.txt, with the resulting file shown beside you; each section becomes an ## heading per the spec
llms-full.txt generation: the companion file with full Markdown content for each curated page
Five SEO plugin integrations: Rank Math, Yoast SEO, All in One SEO, SEOPress, The SEO Framework
Schema-aware descriptions: uses structured data before falling back through the description chain
When to publish: Instant (about 30 seconds after a change), Daily, Weekly, or only when you click Publish
WooCommerce support: SKU, price, stock, categories, dimensions; respects product visibility
WordPress Multisite: network-activate across all sites; each site manages its own independent llms.txt; a network screen to publish each site or all of them
Pre-built templates: Business, E-commerce, SaaS, Blog, Local Business
Show the whole file: read the exact file before you publish, in readable form or as Markdown
Markdown pages: optional per-page Markdown for everything in your llms.txt, built from a snapshot of the public page, with excluded and non-public pages returning not found
Import / export: move your configuration between sites as JSON
WP-CLI: wp llms-txt regenerate, wp llms-txt status, wp llms-txt crawler-log
REST API: POST /wp-json/llms-txt/v1/regenerate, GET /wp-json/llms-txt/v1/status, plus the llms-txt/v2 resource endpoints the workspace itself uses
Atomic file writes: temp file → rename; no half-written files served to bots
ETag/304 caching: proper HTTP headers for CDN revalidation
Subdirectory / Bedrock support: correctly finds the site root on non-standard installs
Robots.txt reference: automatically adds a spec-compliant comment
Description Suggestions in detail
When pages lack descriptions, AI systems get less context. The suggestion engine fills the gap automatically:
Schema markup (Rank Math, custom _schema_json)
SEO plugin meta description
WordPress excerpt
Open Graph description (_og_description / og_description)
First 160 characters of post content
Already-set descriptions are never touched. Pages that can’t be filled automatically are listed for manual review.
Description coverage
Curator shows how many published resources have a description, and Review lists the ones that still need one. None of this is published in the file. Up to 2.0.0 the file ended with a coverage report; 2.1.0 removed it, because the llms.txt proposal makes every ## section a list of links, and the reference parser published with the proposal could not read a file that ended any other way.
AI Crawler Analytics
Track 12 known bots: GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, Meta-ExternalAgent, Bytespider, CCBot, Cohere, DeepSeek, Amazonbot.
The 7-day analytics card shows a visual bar chart of recent activity. All-time totals are kept separately. IP addresses are anonymised before storage: last octet zeroed for IPv4, last 80 bits for IPv6. No data leaves your server.
Safety checks and the format check
Before publishing, the safety checks look for:
Pages missing or unpublished
Duplicate URLs across sections
Members-only pages (always left out)
Noindex pages (left out of both files by default)
Items whose content type is not ticked
Canonical mismatches
Password-protected pages (always left out)
Thin content (< 100 words, no meta, no excerpt)
File size > 50 KB
Too many pages (> 80)
When safety checks are on (Settings), errors stop the file being published and the results are shown immediately. Warnings are shown but don’t block. Check the file’s format in Settings runs the same checks on demand. Everything is explained in plain English.
When to publish
Instant: about 30 seconds after any page is published or trashed (default)
Daily: once a day, via a WP-Cron recurring event
Weekly: once a week
Manual only: only when you click Publish
How to get started
Install and activate. Curator finds your public pages and publishes them straight away.
In Your llms.txt, arrange your sections and remove anything you do not want. Aim for the pages that best represent your site.
In Review, decide on anything Curator flags, such as page-builder templates carried over from an earlier setup.
Click Suggest descriptions from your pages, then accept or edit each suggestion.
Click Publish, then Check what visitors receive to confirm the public file is current.
Optional, in Settings: llms-full.txt, Markdown pages, and the AI crawler log.
WooCommerce integration
When WooCommerce is active, the plugin automatically includes SKU, price, and stock status in llms.txt descriptions, and full product details in llms-full.txt. Products with “hidden” visibility are excluded, and you can optionally exclude out-of-stock products.
Developer hooks
llmscu_capability filter: override the required capability (default: manage_options)
llmscu_post_limit filter: scanner post limit per type (default: 500)
llmscu_full_word_limit filter: per-page word cap in llms-full.txt (default: unlimited)
llmscu_full_max_bytes filter: size ceiling for llms-full.txt in bytes (default: 5 MB, minimum 256 KB)
llmscu_max_sections / llmscu_max_entries filters: limits for saved or imported settings (defaults: 200 sections, 5,000 entries)
llmscu_is_restricted filter: mark a post as members-only so it is left out of both files (function ( bool $restricted, int $post_id ))
llmscu_backup_dir filter: where replaced files are backed up (default: uploads/llms-txt-curator-backups; point it outside the web root if your host allows)
llmscu_regenerated action: fires after each successful regeneration with the content string
