Rankit Agent Ready

By Frédéric Jézégou

0+ active installationsVersion 1.0.0

Tell AI companies how they may use your content, serve them clean Markdown, and see which AI crawlers actually read your site.

Description

Decide what AI does with your content, then measure it.

AI assistants (ChatGPT, Claude, Perplexity, Gemini, Copilot, Mistral…) now read websites on behalf of their users, and AI companies train models on the web. Rankit Agent Ready gives you control over both, with the open standards these companies have committed to, and shows you what really happens on your site. Human visitors see no change.

1. Say no to AI training, in the format the AI companies agreed on

On October 30, 2026, OpenAI, Google, Microsoft and Anthropic said they will honor one shared robots.txt signal for AI training preferences. Rankit Agent Ready publishes it for you:

  • The shared __usage_ai-training preference, with path exceptions (for example: refuse training on your blog, allow it on your docs).
  • Microsoft’s per-page AI training and AI grounding preferences, as a meta tag and an X-Robots-Tag header, including on Markdown responses.
  • Content-Signal directives (search, ai-input, ai-train) and a per-bot robots.txt policy for 31 AI crawlers, with presets.
  • Optional server-level 403 for bots that ignore robots.txt.

Refusing training does not remove you from AI answers: AI search and assistant crawlers can still read and cite your pages.

2. Give AI agents a clean version of every page

  • A Markdown version of every post and page through standard HTTP content negotiation (Accept: text/markdown), .md URLs and ?format=markdown, with YAML front matter (title, author, dates, language, categories).
  • Works behind page caches thanks to static .md files and a marked .htaccess block (Apache, LiteSpeed).
  • Discovery: rel="alternate" links, a Markdown sitemap, a Markdown feed, an API catalog, /llms.txt and /llms-full.txt. If an SEO plugin already provides llms.txt, it keeps it.

3. Measure AI crawler visits

  • 30-day dashboard of AI crawler visits by bot and by page: which AI reads what, and whether it got Markdown or HTML.
  • CSV export, no IP address stored.
  • A readiness checklist and diagnostics (headers, .htaccess self-check, conflicts with cache and security plugins).

4. Connect AI agents

  • 10 MCP abilities through the WordPress Abilities API (WordPress 6.9+): readiness score, crawler statistics, bot policy… Write abilities require confirmation.
  • Open Knowledge Format (OKF) export of your published content, and an optional /okf/ endpoint (off by default).

Also: setup wizard, optional “Summarize with AI” widget, settings export and import, WPML and Polylang support. Everything sensitive is opt-in: no credit link on your site, no data sent anywhere unless you turn it on.

External services

Rankit Agent Ready works without any external service. It only contacts the services below, and only in the cases described. Full details: https://wp-agent-ready.com/privacy

Rankit Agent Ready observatory (opt-in, off by default)

Used to compare your AI crawler activity with other sites using the plugin. Nothing is sent unless an administrator turns on “Share anonymous statistics” in the Observatory tab.

  • When: once a week (WP-Cron), and when the administrator clicks “Send a test”.
  • Where: https://wp-agent-ready.com/api/contribute.php
  • What: the number of visits per AI bot and per resource type over the last 7 days; an anonymous random site identifier (generated once, unrelated to the site’s address or keys, only used to avoid duplicates); the domain extension (for example fr), the two-letter site language and the optional site category; the plugin and PHP versions in the User-Agent header. No page address, domain name, IP address, content or personal data.
  • Privacy policy: https://wp-agent-ready.com/privacy – Legal notice: https://wp-agent-ready.com/legal

isitagentready.com audit (Cloudflare)

Used to run Cloudflare’s public “agent ready” audit of your site.

  • When: only after an administrator has ticked the consent box shown next to the “Sync score” button (the consent is remembered), and only when that button is clicked (limited to one request per minute). Nothing is sent before that.
  • Where: https://isitagentready.com/api/scan
  • What: your home page URL and the list of checks to run. The User-Agent header also contains your home page URL and the plugin version.
  • Service provided by Cloudflare: terms https://www.cloudflare.com/website-terms/ – privacy https://www.cloudflare.com/privacypolicy/

“Summarize with AI” widget (optional)

The plugin itself sends nothing. When a visitor clicks one of the widget buttons, their browser opens the chosen AI assistant (ChatGPT, Claude, Perplexity, Gemini, Mistral) or Google in a new tab, with a prompt that contains the address of the current page. Each service applies its own terms and privacy policy: OpenAI https://openai.com/policies/ – Anthropic https://www.anthropic.com/legal – Perplexity https://www.perplexity.ai/hub/legal – Google https://policies.google.com/ – Mistral AI https://mistral.ai/terms

Schema identifiers

Some JSON files published by the plugin (for example /.well-known/agent-skills/index.json) declare a $schema URL such as https://schemas.agentskills.io/discovery/0.2.0/schema.json. This is only an identifier written in the file for the agents that read it: the plugin never contacts these addresses.

Third-party libraries

The plugin bundles these open-source libraries, unmodified except for a namespace prefix on the PHP library:

  • league/html-to-markdown 5.2.3 (MIT) – HTML to Markdown conversion. https://github.com/thephpleague/html-to-markdown
  • marked 18.2.0 (MIT) – Markdown rendering in the optional reader mode. https://github.com/markedjs/marked
  • DOMPurify 3.4.17 (Apache 2.0 / MPL 2.0) – sanitizes the reader-mode HTML. https://github.com/cure53/DOMPurify
  • Chart.js 4.6.1 (MIT) – charts on the plugin’s statistics screen. https://github.com/chartjs/Chart.js

Installation

  1. Install Rankit Agent Ready from Plugins → Add New, or upload the ZIP file through Plugins → Add New → Upload Plugin.
  2. Activate the plugin.
  3. The setup wizard opens. Answer the questions or keep the recommended values, then launch the configuration.
  4. If you use a page-cache plugin, open Rankit Agent Ready → Settings → Cache & Static, generate the .md files and update the .htaccess rules, then purge your cache.
  5. Check the result in Rankit Agent Ready → Diagnostic, or run curl -H "Accept: text/markdown" https://your-site.com/any-post/ from your computer.

FAQ

Will this change what my visitors see?

No. Human visitors keep receiving your normal HTML pages. Markdown is only served when a client explicitly asks for it (Accept header, .md URL, ?format=markdown) or, if you enable it, when a known AI crawler is detected.

Can Markdown pages create duplicate content in Google?

Markdown responses are sent with an X-Robots-Tag: noindex header by default, and each HTML page points to its Markdown version with a rel="alternate" link. You can allow Google to index the .md files if you prefer.

Does the plugin block AI bots?

Not unless you ask it to. By default AI crawlers are allowed. You choose which bots to disallow in robots.txt, and you can optionally add a server-level 403 rule for bots that ignore robots.txt.

Are Content-Signal and the AI-training preference legally binding?

No. They are declarations of intent written in robots.txt. Major AI companies have announced they will respect the shared __usage_ai-training signal, but compliance is up to each crawler. For hard blocking, use the per-bot policy or the 403 filter.

I use WP Fastest Cache / WP Rocket / LiteSpeed Cache. Will it work?

Yes. Page-cache plugins serve the same HTML to everyone, so the plugin generates static .md files and adds .htaccess rules that let Apache or LiteSpeed serve Markdown before the cache answers. The Cache & Static tab detects your cache plugin and lists the settings to add yourself (for example, the AI user agents to exclude from the page cache). The plugin never changes another plugin’s settings.

Does it work on Nginx?

Yes. On Nginx, Markdown is served by WordPress itself (Accept header, .md URLs, ?format=markdown). The .htaccess features (static .md files served directly by the web server, server-level 403 filtering) are for Apache and LiteSpeed only.

Does the plugin write files at the root of my site?

Only the .htaccess file, inside a clearly marked block, and only on Apache or LiteSpeed. llms.txt, llms-full.txt and the Markdown sitemap are served dynamically by WordPress. If your site has a physical robots.txt file, the plugin never edits it: it shows the lines to add so you can paste them yourself. Backups of .htaccess and the generated Markdown files are stored in wp-content/uploads/, in a folder protected from direct access.

Another plugin already creates my llms.txt. What happens?

Rankit Agent Ready detects the existing llms.txt, shows a warning and leaves it untouched. It never modifies another plugin’s settings and never deletes a file it did not create, including on uninstall.

What is the OKF export? Is my content made public?

OKF (Open Knowledge Format) packages your published posts and pages as Markdown files you can feed to an AI assistant, a RAG system or an archive. The ZIP is generated on your server, downloaded to your computer and immediately deleted from the server. It only contains content that is already public. The public /okf/ endpoint is disabled by default.

What do I need for the MCP abilities?

WordPress 6.9 or later (Abilities API). To connect a desktop client such as Claude Desktop, install the official MCP Adapter plugin. Write abilities always ask for confirmation in the MCP client before running.

Does the plugin send data to external servers?

Only in two cases, both described in the “External services” section: the anonymous observatory, which is off until you turn it on, and the isitagentready.com audit, which runs only when you click its button. AI crawler statistics are otherwise stored on your own site.

Is it compatible with WPML and Polylang?

Yes. Markdown versions, llms.txt and the Markdown sitemap follow your language setup.

What happens when I uninstall the plugin?

Deleting the plugin always removes the public files it generated (Markdown copies, caches) and its .htaccess block. Your settings, statistics and per-post choices are kept, so you can reinstall without losing them, unless you tick “Delete all plugin data on uninstall” in Configuration. Files created by other plugins are never touched.

Screenshots

Changelog

1.0.0

  • First public release on WordPress.org, under the name Rankit Agent Ready (formerly WP Agent Ready, distributed from wp-agent-ready.com).
  • Fully translatable interface (English and French included).
  • Opt-in only: no credit link on your site and no data sharing unless you turn them on.
  • Nothing is written at the root of your site except a marked block in .htaccess (Apache and LiteSpeed). llms.txt, llms-full.txt and the Markdown sitemap are served dynamically; a physical robots.txt is never edited.
  • llms.txt provided by another plugin (Yoast SEO, Rank Math, All in One SEO) is detected and reported, never modified.
  • Coming from WP Agent Ready? Rankit Agent Ready pauses while it is active, offers a one-click switch, and copies your settings and statistics.

The full version history is in changelog.txt.