Your help center tells search engines and AI tools what they may read and where to find it, with no setup on your part. Here is what every public help center publishes, what the signals mean, and how to turn it all off.
Plans: All plans
What your help center publishes
These addresses work on your help center's domain, including your custom domain if you have one.
Address | What it's for |
|---|---|
| Lists your home page, categories and published articles, with links between language versions, so search engines find every page. |
| Tells crawlers what they may read, links your sitemap, and states your content signals. |
| A Markdown index of your categories and articles for AI assistants, in your default language, with the address of your public MCP server. |
| Your AI usage preferences in plain text: AI assistants may read your pages and cite them, and training models on them is not permitted. |
| Points AI tools to your public MCP server, when it is on. |
| Your public MCP server, which AI assistants use to search and read your articles. See Let readers use your help center in their AI assistant. |
A Markdown version of every page
AI tools read Markdown more reliably than a styled web page. Your home page, your list of categories and every category and article page return clean Markdown at the same URL when a tool asks for it with the Accept: text/markdown header:
curl -H "Accept: text/markdown" https://help.example.com/content/your-articleReaders can also download an article as Markdown from its Export menu, when your article page has one. Developers find every address, header and file format in llms.txt, ai.txt and Markdown pages on the developer portal.
What the content signals mean
Your robots.txt includes this line:
Content-Signal: search=yes, ai-input=yes, ai-train=nosearch=yes: search engines may index your pages and show them in results.
ai-input=yes: AI assistants may read your pages to answer questions and cite them.
ai-train=no: your content may not be used to train AI models.
Like robots.txt itself, these signals are instructions that well-behaved crawlers follow, not a lock.
Only what anyone can read
The sitemap, llms.txt and the MCP server list only published content that a visitor without an account can read. Drafts, private articles and categories, unreleased translations and team notes stay out.
Search results pages and "not found" pages tell search engines not to index them.
Private and password-protected help centers are closed to crawlers, because a visitor has to sign in or enter the password first. Help centers limited to specific IP addresses are closed to crawlers from other addresses too.
Turn it off
In the left menu, click the gear icon, then Settings.
In the General card, switch off Search engine indexing.
Click Save changes.
Your robots.txt then asks every crawler to stay out and sets all three signals to no. llms.txt stops answering, ai.txt states that your help center has opted out, and the public MCP server turns off. Search engines can take a while to drop pages they already indexed.
The Hide from search engines (no-index) switch in the template editor works differently: it adds a noindex tag to your pages, but it doesn't change llms.txt or the public MCP server. See Hide your help center from search engines.
Comments
Be the first to comment.