Why won't I be able to signal to AI/LLMs to sod off?

[TheJessieKirk]

I cannot fathom why it apparently isn't included in this proposed standard to signal AI crawlers to sod off?

Honestly, it would just be the case of adding a "no, go away" signal and implementing it via the HTTP header and/or robots.txt file.

That way everyone is happy, right?

[TimidRobot]

The IETF AI Preferences working group is already planning on a "sod off" option. CC signals is meant to provide content holders with a way of allowing machine use with reciprocity.

Our README gives the following example:

User-Agent: *
Content-Usage: train-ai=n;exceptions=cc-cr
Allow: /

To communicate "sod off", CC signals isn't needed. The following example will deny permission:

User-Agent: *
Content-Usage: train-ai=n, ai-usage=n
Allow: /

The working group is still very much in process and these examples are not final. For more information on the IETF AI Preferences working group, see: AI Preferences (aipref).

[updated 2025-07-28 to reflect changes in IETF AIPREF drafts]

[TheJessieKirk]

Okay dokey then. Thanks!

[benjamindaass]

Sorry for an ignorant question but I am just trying to understand the README file. Am I correct in saying that the option can be added to a robots.txt file OR the HTML head element? Or does it need to be both? Secondly, with respect to the HTML head element, would I need to add those lines to the head element of every page on my site? Or would adding it to the index page suffice?

[TimidRobot]

@benjamindaass

The initial two attachment methods are robots.txt and the HTTP Content-Usage header. Either (or) can be specified. For robots.txt the preference/signal applies to a path; for the HTTP header, it applies to a URL.

As far as HTML goes, the http-equiv attribute of the meta tag does not yet define a content-usage value. It is my understand that this would need to happen before CC signals could be used within a web page (versus the headers returned by the web server). I expect there will be additional work on adding file formats (like HTML) after the IETF AI Preference Vocabulary is finalized.

[ccjocelyn]

@TheJessieKirk since this discussion addresses technical implementation questions, any objections to us moving it to the technical implementation section?

[TheJessieKirk]

None at all.