Skip to main content

AI permissions

Your site tells AI systems and search engines what they are allowed to use its content for. The setting is delivered on the Content-Signal line of the robots.txt file — a standard proposal by the AI Preferences working group that major players (OpenAI, Anthropic, Google, Common Crawl) have committed to honoring.

The settings are site-specific and live in the ai_signals block of the site-config.json file. The easiest way to change them is to ask Kamu: "Block my content from being used for AI training."

The three permissions​

  • search — whether the content may appear in search results (traditional search engines and AI-based search engines like Perplexity).
  • ai_input — whether an AI assistant (e.g. ChatGPT, Claude) may fetch information from your page when a user asks it something. This is how customers find you through AI assistants today.
  • ai_train — whether your content may end up in AI models' training data, i.e. become part of the "memory" of next-generation models. Different from ai_input — training happens once in advance, usage happens every time someone asks.

Each value is either yes or no. By default everything is yes, because a marketing site's core job is to be found.

When to choose no?​

Most businesses have no reason to restrict any of these, but there are situations where a restriction is justified:

  • ai_train: no — if your content is unique expert writing you do not want absorbed into general AI models. A user can still find you via an AI assistant, but your content is not stored in the model's weights.
  • ai_input: no — if you do not want AI assistants to fetch information from your page at all. This is a strong restriction and weakens your visibility in AI searches.
  • search: no — effectively hides your site from search engines. A very rare choice for a public marketing site.

Changing the settings​

Ask Kamu to edit the settings for you:

  • "Allow everything: search engines, AI use, and training." (default)
  • "Block training use but allow everything else."
  • "Block AI searches entirely."

After the next deployment the change shows up in your site's robots.txt, for example like this:

User-agent: *
Content-Signal: search=yes, ai-input=yes, ai-train=no
Allow: /

Notes​

  • Content-Signal is a declaration mechanism, not a technical barrier. It relies on AI operators reading and respecting the file — exactly how traditional robots.txt has always worked.
  • The settings only apply going forward. If your content has already ended up in some model's training data, ai_train: no does not remove it from there.
  • You can check your site's current state at isitagentready.com by entering your site's address.