Managing Google Search AI Inclusion at Page and Folder Level

Google now documents a Search Console control for inclusion in specified generative AI features on Google Search

Direct answer

Google now documents a Search Console control for inclusion in specified generative AI features on Google Search. It can be configured through properties, with child properties inheriting an applicable parent setting unless they have a manual choice. A publisher should map the affected URLs, inspect inheritance, and record the effective decision before changing a folder or narrower scope. This control is separate from training restrictions, ordinary Search removal, and another platform’s crawler policy. Treat an inclusion choice as eligibility rather than a promise of exposure. Verify configuration and subsequent observations separately, allowing for Google’s stated propagation and caching delays.

A new control changes an old operating assumption

For much of the recent GEO discussion, publishers were told that Google Search participation and its AI features could only be managed through existing crawl and preview controls. That advice needs a date. Google’s current Search generative AI control help page documents a separate Search Console choice and states that worldwide rollout occurred on August 31, 2026.

An operational guide written before that change should not be reused as if it describes today’s entire control set. This is a good example of why a GEO knowledge base needs versioned facts rather than timeless instructions about platform behavior.

The control covers AI Overviews, AI Mode, and specified generative AI features in Discover. It offers inclusion, exclusion, and inheritance through Search Console properties. Exclusion prevents the site’s links and content from appearing or grounding responses in the applicable features. It does not function as a general removal instruction for every Google service.

Those are product boundaries. The rest of this article proposes a change-management procedure for applying them to a real site; it does not claim to have inspected or changed any particular site’s Search Console configuration.

Start with the decision rather than the switch

A site may contain public product documentation, licensed research, a help center, regional offers, and historical material. The right inclusion decision can differ across those sections. A broad setting is easier to configure but can affect a larger inventory than the person making the change intended.

Write the objective in terms of content and surfaces. For example: keep public product instructions eligible in the applicable Google AI features while excluding a defined collection of licensed reports. This statement identifies the desired boundary. “Block AI” does not identify whether the concern is Search presentation, model training, user-initiated access, or confidential content.

Confidentiality is a separate system requirement. A publicly accessible report should not be treated as confidential merely because one search feature is excluded. If information requires authentication or authorization, enforce that in the application rather than treating Search Console as an access-control system.

Property hierarchy determines the affected scope

The control is configured through Search Console properties, not a universal new HTML tag attached to every page. Google’s documentation describes top-level domain, subdomain, and URL-prefix relationships. A child can inherit from the closest parent with a manually configured choice or set its own choice.

In practice, folder or narrower treatment requires identifying an appropriate property scope and verifying which URLs belong to it. A URL-prefix scope is not a synonym for a content-management category. A category can span several hosts or paths, while a path can contain more than one editorial category.

Consider a fictional site with a public catalog under https://example.com/products/ and licensed reports under https://example.com/reports/. If the relevant properties are configured and managed as documented, the reports scope can have a manual exclusion while the wider property remains included. A nested property may also have its own manual choice.

The useful mental model is a hierarchy of effective settings. The value displayed for a broad property is not sufficient evidence that every descendant follows it. Conversely, a manually included descendant should not be assumed to follow a later exclusion at its parent without inspecting its inheritance state.

Draw the scope map before making a change

List the actual URLs that matter, including representative pages at boundaries. Identify hostname, protocol, path, redirects, and alternate versions. Search Console URL-prefix relationships distinguish these elements; a setting for one prefix should not be assumed to cover a differently formed URL.

Record the property that controls each example and any closer manually configured parent. The same property can be a child in one relationship and a parent in another. A diagram can help a large site, but a short table is usually enough for a small one.

This is an editorial operations table rather than Google’s mandated reporting format. Its purpose is to make a site-wide change understandable to someone who did not configure it.

Keep Search inclusion separate from other controls

Google says the new choice is not a ranking or inclusion signal for other parts of Search and does not override separate participation in Merchant Center or Google Ads. It also distinguishes the choice from training restrictions and complete Search removal. These boundaries belong in the change record.

Google’s crawler documentation describes Google-Extended as a control token for specified model training and grounding uses outside ordinary Search inclusion. Its policy should be reviewed for the actual use being governed, rather than treated as another spelling of the new Search control.

Robots meta and snippet directives address other presentation and indexing decisions. An inclusion choice does not turn an otherwise inaccessible or ineligible page into a guaranteed source. The Google AI optimization guide remains relevant to ordinary technical and content requirements.

Another platform requires another review. OpenAI’s crawler overview distinguishes search, training, and user-initiated fetching. A change in Google Search Console should not be described in a client report as a verified change to those OpenAI policies.

Use a staged configuration procedure

The following procedure is proposed for an authorized site owner. It is not an account-specific instruction based on a live inspection.

1. Inventory the URLs and content collections whose treatment needs to change.
2. Identify the applicable Search Console properties and document their current effective choices.
3. Check for manually configured descendants, especially those managed by another team.
4. Translate the content policy into include, exclude, or inherit choices at the necessary scopes.
5. Record the reason, affected examples, responsible owner, and previous values.
6. Apply the choice through the documented Search Console control using an appropriately authorized owner account.
7. Reopen the relevant properties to verify the saved choice and inheritance state.
8. Observe eligible surfaces and reporting after propagation, keeping timestamps and limitations with the findings.

On a complex site, begin with a well-understood scope rather than changing a broad property whose descendants have not been inventoried. That is a practical way to limit unintended impact. It is not a claim that Google requires a staged rollout or offers a testing sandbox for the setting.

Record the exact previous configuration. Restoring inheritance is different from manually selecting the same effective value. Both can look included today, but they behave differently when the parent changes later. A rollback record should restore the intended relationship as well as the visible value.

Verify three different things

First verify configuration: the intended property shows the intended choice, with the expected inheritance or override. This can be checked without waiting for an AI answer to change.

Second verify scope: the inventory and property relationships cover the intended URLs. A successful save says little about a page that falls outside the selected scope. Redirecting or alternate-host pages deserve particular attention because a human may think of them as one document while the configured prefixes differ.

Third verify subsequent observations: record any displayed links, sampled responses, and available report data. Google’s help page describes a delay of a few days for changes and additional caching or propagation. That is not a promise that every observed surface will change at the instant the configuration is saved.

A page not appearing in one sampled answer does not prove exclusion succeeded. It may simply not have been selected. A falling impression count does not identify configuration as the cause without considering demand and other changes. Verification should combine the saved state with scoped observations rather than treating absence as a definitive execution trace.

Exclusion does not erase the surrounding web

Google’s documentation notes that content from other sites may still appear, including content similar to the excluded site’s material. A publisher can control the applicable treatment of its own site through this setting. It cannot use that choice to rewrite the rights or configuration of every secondary publication.

This makes source lineage relevant. If another site quotes an announcement or hosts a licensed copy, its treatment may need separate contractual or editorial attention. The presence of a similar fact in an AI answer does not by itself establish that the excluded site was used contrary to its setting.

When investigating a suspected issue, save the answer, its linked sources, the effective property choice, and the timing. Distinguish direct references to the controlled URL from independent or derived material elsewhere. Avoid assigning a mechanism that the available evidence cannot establish.

Measure participation without promising visibility

An include choice opens eligibility within the control’s scope. It does not guarantee impressions, citations, favorable description, or traffic. Those outcomes require separate evidence.

Google’s generative AI performance report provides an official source for supported impression data. Check its current availability and reporting scope for the property. Use it alongside a change log, not as a complete causal test of the inclusion policy.

For a content team or an agency such as Xindar, a sensible handoff includes the scope map, inherited and manual choices, reasons for exceptions, verification records, and report definitions. That makes the policy maintainable when a new product section or regional site is added. No customer implementation or visibility result is represented here.

Frequently asked questions

Can this control exclude one folder while keeping another included?

Google documents manual choices for child properties. The operational requirement is to use the appropriate property scopes and inspect their relationships. Do not assume that a CMS category automatically has an equivalent Search Console boundary.

Does a child setting always follow a parent change?

No. A child may have a manual choice. Check the effective state and whether it inherits. The hierarchy, rather than the folder name alone, determines which configuration applies.

Is a drop to zero impressions proof that the setting worked?

It is an observation, not sufficient proof by itself. Configuration, scope, propagation, report availability, and ordinary selection or demand can all matter. Keep the saved setting and the observation separate in the audit.

Should every public page be included?

That is a publisher policy decision. Public discoverability, licensing, and the intended use of a collection can differ. Decide by content scope and relevant surfaces, then apply the control through documented boundaries.

Source and method note

Official pages were retrieved on September 15, 2026. The control description reflects the current help documentation, including its stated August 31 rollout. Property examples and the change-management workflow are editorial proposals. No Search Console account was inspected, no setting was changed, and no site-specific effect was measured. Review the current documentation before implementation because covered features and operational details can evolve.