Skip to content

The UpSearch search briefing

Google AI Search Opt-Out vs Google-Extended: What to Check

Understand Google AI Search opt-out, Google-Extended and property inheritance, then assess the visibility trade-offs before changing your website settings.

By UpSearch · · 5 min read

A decision about AI training should not accidentally become a decision to disappear from AI search. Before changing any setting, separate three objectives: controlling training use, controlling AI answer inclusion, and controlling whether pages appear in search at all.

Google's Search generative AI control became available worldwide on 31 August 2026. Its announcement for website owners makes the commercial trade-off clear: excluding a site removes a potential source of AI-feature visibility and traffic.

This is a decision guide for website owners, not a recommendation that every publisher should choose the same setting.

Reviewed 8 September 2026.

Which control does what?

Google's control documentation distinguishes Search generative AI inclusion from AI training. The Search control covers AI Overviews, AI Mode and generative Discover features. Google-Extended covers training permission and grounding in Gemini Apps and Grounding with Google Search on Vertex AI. It is not just a training switch: see Google's crawler documentation. The Search control is not used as a ranking or inclusion signal for other Search areas, and does not override Merchant Center or Ads choices.

Think about the intended outcome before selecting a technical mechanism. A publisher protecting licensed material may have a different objective from a local business trying to reach customers. Both need a deliberate choice, not a generic instruction to block AI.

How to check the Search setting

In the relevant Search Console property, open Settings, then Search generative AI. The options are inclusion, exclusion or inheritance from a parent property. Inspect the effective setting, not just the top-level domain: a child property can override inheritance. Google explains the scope and propagation delays in its settings guide.

We recommend recording the property name, effective choice, inherited source where relevant, date and responsible person before changing anything. This gives your team a useful audit trail if visibility changes later.

Do not make a site-wide change solely because one member of the team dislikes a particular answer. Identify the actual content, commercial concern and intended scope first.

Decide what your business is trying to protect or gain

Our suggested decision framework starts with the business model.

For a service business, ask whether clear explanations of your services help prospects decide to contact you. For a subscription publisher, ask which material is publicly available, what readers pay for and where the site's value genuinely depends on a visit. For an ecommerce business, consider how product information supports discovery and how sales are measured.

These are planning questions rather than claims about what Google will do for a particular site. A decision can be sensible even when the available traffic evidence is incomplete, provided the uncertainty is explicit.

Write a short rationale such as: "We are keeping public service pages eligible because they answer pre-sale questions; we will review visibility and qualified enquiries monthly." A specific rationale is much easier to revisit than "AI is good" or "AI is bad".

Keep an access inventory, not one AI toggle

Our proposed inventory has a separate row for each system and purpose. Record search inclusion, training permission, ordinary indexing and any infrastructure restrictions independently. Add an owner and a last-reviewed date.

For comparison, OpenAI's publisher guidance distinguishes OAI-SearchBot, used for search discovery, from GPTBot, associated with training. A choice affecting one provider should not be treated as proof of the setting for another.

Ask your developer to document the intended policy and the implemented configuration side by side. That is more useful than a report saying only "AI bots blocked", because it tells you whether the technical change matches the commercial decision.

A controlled review before and after a change

First, save the current settings and your chosen baseline. Second, document the exact scope of the change. Third, choose a review period that is appropriate for the amount of data your site receives. Avoid changing content, navigation and eligibility simultaneously when you are trying to understand one effect.

During the review, compare comparable periods and keep business outcomes separate from visibility. For example, a campaign or a seasonal promotion could alter enquiries independently of the setting you changed. Put those events in the same notes rather than attributing every movement to AI.

Define a reversal condition before making a discretionary change. That might be a commercial threshold, a revised licensing decision or new evidence about the pages involved. A reversible experiment with a written rationale is easier to manage than an undocumented permanent switch.

What exclusion cannot promise

Excluding your own content is not a guarantee that your business will never be discussed. Other websites can still provide information about it. Google also separates this control from removing a site from Search entirely; its documentation explains that distinction.

The practical lesson is to define success narrowly. "Our own pages are excluded from this feature" is a testable policy objective. "No AI system will ever mention us" is not the same objective and should not be promised to a client.

Where to go next

Use the UpSearch AI Visibility Toolkit as a starting point for broader visibility work, and record configuration decisions alongside your technical review. Check the relevant provider's own settings directly rather than assuming a tool can observe every private control.

Our Generative AI reporting guide explains how to interpret the resulting exposure data. The important principle is simple: match the control to the outcome you actually intend.

FAQ

Should every business opt out of AI Search?

No single choice fits every business model. We recommend making the decision around your content, acquisition goals and commercial constraints, then documenting the reason and reviewing the evidence.

Should I change several AI settings together?

Prefer a clearly scoped change when possible. Separate decisions are easier to audit and make it less likely that an unrelated configuration change will be blamed for a visibility shift.

What should my team record?

Keep the property or system, intended purpose, effective setting, scope, change date, owner and review criteria together. That creates a useful record for both marketing and development.