Choose separate controls for AI search, training and private content

Match the control to the publishing decision

This is a decision map, not a configuration to paste into a website. The provider-specific sources and a worked handoff below explain what to check.

Google has a separate generative Search control

Google documents a Search Console setting under Settings → Search generative AI for AI Overviews, AI Mode and generative features in Discover. Its Help page states that the control reached websites worldwide on August 31, 2026. The inclusion choice permits participation; it does not promise that a page will appear.

The exclusion choice covers links and content in the listed features. Google says this choice is not a ranking or inclusion signal for other parts of Search, does not govern training, and does not override separate participation in services such as Merchant Center. Content from other sites can still appear.

Check the exact property before making a change. A child property can inherit a parent's setting or have its own configuration. That makes a domain-wide decision different from a decision for a specific section. Changes take time to propagate. These are documented behaviors, not a verification of the controls in your account. Google's Search generative AI control

Use page and passage controls for a narrower Google preview decision

Google's nosnippet rule prevents text snippets and direct use of the page's content as input to AI Overviews and AI Mode. max-snippet limits the amount of text, with exceptions for separately permitted uses. data-nosnippet excludes designated text from snippets. It belongs on a valid span, div or section; the attribute's presence matters, even if its value is written as false.

For example, this invented passage illustrates the markup, not a tested result:

<section data-nosnippet>
  <p>Seasonal terms for existing customers appear here.</p>
</section>

The passage stays visible on the page. Structured data has separate treatment, so inspect it as well. A preview restriction does not remove the information from other public copies or establish a universal AI-use restriction. Google's preview specifications

If the intended result is removing the whole page from Google Search, use the appropriate noindex implementation instead. Google must be allowed to fetch the page to see that instruction. Blocking crawling at the same time can prevent it from reading the rule. Non-HTML resources can use an X-Robots-Tag response header. Google's noindex guidance

Keep training choices separate from search choices

OpenAI documents OAI-SearchBot for ChatGPT search and GPTBot for content that may be used to train its foundation models. Their settings are independent. Declining GPTBot does not require declining the search crawler. OpenAI also distinguishes ChatGPT-User, which can visit a page following a user's request; robots.txt rules may not apply to those user-initiated actions. A training preference therefore should not be described as preventing every visit from ChatGPT. OpenAI's crawler documentation

Google's Google-Extended documentation covers specified Gemini training and grounding uses. Grounding means providing content to a model when it answers. Google says the token does not affect inclusion or ranking in Google Search, and it has no separate HTTP request user-agent string to look for in server logs. Review its current product scope rather than treating “Google” as one undifferentiated bot. Google's crawler and control-token documentation

Put private material behind access controls

If a customer document should be available only to authorized people, protect access to the document itself. Google explains that robots.txt relies on crawlers respecting instructions and that a disallowed URL can still be indexed when other pages link to it. Password protection is a different mechanism. Google's robots.txt limitations

Likewise, noindex changes search eligibility while leaving the URL available to people who can reach it. Decide whether the requirement is less public exposure or restricted access before selecting a technical rule. Google's content-control guide

Give the site owner an outcome and a verification plan

Consider an invented publisher that wants public research discoverable, wants to decline a documented training use and has a private client appendix. The handoff should contain separate requests: review the relevant training preference; preserve the intended search settings; put the appendix behind authorized access. A single “block AI” ticket leaves all three outcomes ambiguous.

For each request, record the exact URL or property, intended audience, chosen control, current setting, responsible owner and evidence needed to verify the result. Preserve the previous configuration so an unintended change can be reversed deliberately.

For Google page-level controls, inspect the HTML or headers Google received and allow for recrawling. A saved configuration proves the instruction was set; it does not prove when search systems processed it. Google's AI-feature guidance describes checking the fetched HTML and allowing time for the change. Google's preview troubleshooting

Afterward, distinguish implementation checks from observed results. Use the citation and crawler audit for access evidence. Keep training preferences, public search participation and private access as separate decisions throughout the review.