OpenAI backs a shared AI-training signal: what website owners should know

What OpenAI committed to on October 30

OpenAI says it will honor a common robots.txt signal for preferences about foundational generative AI training. The company aims to implement it in the coming months and promises a further update with implementation details. The October 30, 2026 announcement does not provide a specific activation date.

The effort is voluntary and intended to let operators communicate one preference to several participating providers. OpenAI says its existing crawler controls will continue to apply. For a publisher planning a change this week, the announcement is therefore a roadmap commitment rather than evidence that the shared signal is already active at OpenAI.

Public record: OpenAI announcement — October 30, 2026

The linked proposal is a draft, with a narrow purpose

The linked document, Machine-Readable Training Preference Signal, was updated October 30. The IETF Datatracker labels it an individual Internet-Draft, not an IETF-endorsed standard. It remains work in progress.

The proposal uses the policy token __usage_ai-training in a robots.txt User-Agent group, with familiar Allow and Disallow path rules. It separates content-usage preferences from traffic-management rules. The token itself is not a crawler-behavior directive.

That distinction matters when interpreting the word “block.” A declared preference and a server refusing a request are different mechanisms. The draft also allows for provider-specific interpretations and private agreements; a single line should not be presented as a universal guarantee.

Public record: Training preference signal — individual Internet-Draft, October 30, 2026

Search and training already have separate OpenAI controls

OpenAI’s crawler documentation distinguishes OAI-SearchBot, used for ChatGPT search discovery, from GPTBot, used to crawl material that may support generative foundation-model training. It explicitly allows a site to permit the search bot while disallowing GPTBot.

The same documentation says ChatGPT-User handles certain user-triggered visits rather than automatic web crawling. Robots.txt rules may not apply to those requests. A training preference therefore should not be described as a blanket switch for every ChatGPT interaction.

For search changes, OpenAI says its systems can take about 24 hours to adjust after a robots.txt update. Allowing a crawler helps make discovery possible; it does not promise a ranking, a citation or a stream of referral traffic.

Public record: OpenAI crawler documentation — checked November 1, 2026

A useful next step is an inventory, not a sweeping replacement

Our recommendation for publishers is to record three decisions separately: which automated crawlers may fetch pages, whether content may be used for training, and which discovery channels the business wants to support. Give each decision an owner before changing a production file.

Then compare the live robots.txt file, hosting controls and relevant provider documentation. Save the previous configuration and review the exact paths affected by a proposed edit. Watch for the implementation update before assuming the new shared preference has replaced maintenance of existing controls.

The potential benefit is less repetitive configuration across providers. The test will be whether operators can understand who honors the signal, when it takes effect and how it interacts with the controls they already use.