AI CONTENT RIGHTS / ANALYSIS

AI training preferences: check your site before the shared signal arrives

OpenAI says it will honor a common robots.txt training-preference signal in the coming months. Creators can use the announcement to audit today’s rules without mistaking a draft for a live safeguard.

WASSAI editorial desk ·

Conceptual generated photograph of an independent creator reviewing a website on a laptop at a sunlit desk, with no readable interface text.
Original AI-generated conceptual photograph. The blurred screen is illustrative and does not show a real robots.txt file or service interface.

OpenAI said on 30 October 2026 that it plans to honor a shared signal in a website’s robots.txt file for foundational generative-AI training preferences. The appeal for an independent creator is fewer separate rules to maintain across participating providers. But the timing matters: OpenAI says implementation is still months away, and the linked proposal is an individual Internet-Draft, not an adopted IETF standard. If you publish tutorials, photography, designs or a portfolio on your own domain, this is a moment to clarify your policy, not to assume a new switch already protects your work. [1] [2]

Write down the choice before touching settings

List the pages and assets you control: articles, images, downloadable files and public project pages. Decide which material you want available for AI training and which, if any, you would prefer excluded. A creator may choose different treatment for a public press kit and an archive of original work. Record the decision and the person responsible for it, so a later technical change reflects an actual publishing choice rather than a copied template.

Now read the live robots.txt file on your domain and document its current rules. OpenAI says its existing crawler controls will continue to apply and be respected when it adds the common signal. The proposal uses ordinary robots.txt groups to express a training preference for paths; it does not introduce a new directive or change the file’s basic grammar. Keep a copy of the current file and test any future edit against the pages you intend to affect. [1] [2]

Keep training and discovery separate

A training preference describes a use of content; it is not the same question as whether a crawler can visit a page or whether a reader can discover your site. The draft explicitly treats the proposed policy token separately from crawler-specific access rules. Make a small matrix for your site: which sections should remain public, which should be discoverable, and what training preference you wish to express. If visibility matters to your business, review the relevant provider documentation before changing any crawler rule. [2]

Do not present a robots.txt preference as a lock, a retroactive deletion request or a licence agreement. The draft describes a machine-readable expression of preference and notes that it is work in progress. It also says private agreements may override the mechanism. Keep contractual and platform-specific rights questions separate from this file, and use the controls available to you today if you need an immediate provider-specific choice. [2] [1]

Revisit the implementation when it ships

OpenAI promises further implementation details. When those arrive, compare the final wording with the draft, identify which providers actually honor it, and test the served robots.txt file rather than only a local copy. Check a sample path that should be included and one that should be excluded. Until then, the useful output from this announcement is a documented content policy and an accurate inventory of existing controls, not a claim that a common opt-out is already working everywhere. [1] [2]

Sources

  1. OpenAI: Making it easier for websites to express AI training preferences · Published 2026-10-30; checked 31 October 2026.
  2. IETF Internet-Draft: Machine-Readable Training Preference Signal · Published 2026-10-30; checked 31 October 2026.