Google has published Search Console documentation for a control that lets website owners keep their pages out of AI Overviews, AI Mode, and generative AI features in Google Discover, formalizing a mechanism the company began testing in June 2026 under regulatory pressure. The control is rolling out to a subset of website owners, and the help documentation confirms the technical rules that govern how it works, how long changes take, and what it does not cover.

The control appears in Search Console under Settings, in a section labeled Search generative AI. It presents website owners with a binary choice for their property: include the site's links and content in Search generative AI features, or exclude them. The first option is described as the default for all properties. The second prevents a site's content from being visible to users inside those features, meaning it will not be linked to and will not be used to ground AI responses.

What the control does and where it lives

The published documentation describes three configurable states rather than two. Beyond the include and exclude choices, a property can be set to inherit its control value from a parent property. Inheritance is the default whenever a property has a parent, and any change made at the parent level automatically applies to the child unless the child's owner overrides it.

According to Google, the exclude setting stops a site from appearing in the listed generative AI surfaces. Content crawled from an excluded site will not be eligible as an input to generate an AI response or a preview in those features. Content from other sites remains available in the same features, and Google's documentation notes that such content may appear similar to material from the excluded site.

The interface itself carries plain-language framing. In the Search Console settings screen, the include option is labeled with the instruction to maximize a site's visibility by appearing in Search generative AI features, including showing up as links and helping to ground AI responses. The exclude option states that a site will not be visible to users in those features and will not receive traffic or impressions from them.

Three features currently fall under the control: AI Overviews, the summaries that sit above conventional search results; AI Mode, Google's conversational search interface; and generative AI features in Google Discover. Google states in the documentation that it expects to update the list over time as it develops Google Search, leaving the scope open to expansion.

The scope of what the control does not touch

The documentation is specific about the boundaries of the setting, and those boundaries matter as much as the function itself. According to Google, the control affects only whether content can appear in certain Search generative AI features. It is not used as a ranking or inclusion signal that affects other parts of Search. Excluding a site from AI features, in other words, does not by itself change how that site ranks in conventional blue-link results.

Several separate mechanisms remain untouched by the toggle. The control does not override a publisher's other choices to participate in services such as Merchant Center and Google Ads. A site's content may still be used to power Google Search as a whole, including helping Google's systems understand the language of user queries and the language on pages across the web.

Most consequentially, the control does not affect AI training. To limit training of the models that generate responses in Search generative AI features, Google directs website owners to Google-Extended, a standalone robots.txt token that manages whether crawled content may be used for training future generations of Gemini models and for grounding in Gemini Apps and Vertex AI. To block content from appearing in Google Search entirely, the documentation points to the separate noindex directive.

This layering means a website owner who wants content withheld across Google's AI surfaces must operate more than one control. The Search generative AI setting handles product-level visibility in Overviews, AI Mode, and Discover. Google-Extended handles model training and Gemini grounding. Noindex handles removal from Search altogether. Each addresses a different stage of how content moves from a crawl to a generated answer, and no single switch covers all of them.

Timing, propagation, and caching

The documentation sets expectations for how quickly a change takes effect. Once an owner changes the control in Search Console, it generally takes a few days for a site's content to be excluded from the applicable features. Google states that content will be excluded within 1 to 2 days after the control goes live, though some content may take longer to disappear because of caching and propagation across Google systems.

That interval is a practical detail for anyone weighing the decision. A site that switches to exclude will not vanish from AI features instantly, and residual appearances during the propagation window are anticipated rather than anomalous. The delay reflects the distributed architecture through which Google stores and serves generative responses.

How inheritance works across properties

A substantial portion of the documentation addresses property inheritance, a structural feature that determines how the control cascades through the tree of properties an owner may have registered. By default, a property inherits its Search generative AI control from its closest parent that has changed its own control to stop inheriting. When none of a property's parents have manually configured their control, the property inherits the setting of its top-level domain property.

The documentation illustrates the logic with a worked example. A child property such as the URL-prefix property for a business subdirectory inherits the control of its parent domain property by default. That default holds unless the owner of a closer parent property manually configures the control to stop inheriting, at which point the closer parent's setting governs.

Google's documentation defines the vocabulary precisely. A parent property is a higher-level property with a child property nested below it, and it may be a top-level domain property, a subdomain property, or a URL-prefix property. A child property is a lower-level URL-prefix path property, or a domain property that is not the top-level, nested under another property. Many properties, the documentation notes, can be both a parent and a child of other properties at the same time.

The practical import is that owners managing large or segmented sites through multiple registered properties inherit a single decision by default across the whole structure, then override selectively. To specify a control for a given child property, an owner logs in to that property in Search Console and updates its configuration on the Search generative AI control page. The setting changes all pages in a property, including its child properties, unless those children are individually overridden.

Measuring the effect before deciding

Google links the control to a measurement tool intended to help owners understand the trade-off. To measure how content performs in Search generative AI features, the documentation directs owners to the Generative AI performance report, which it describes as a way to get an idea of how changing the control may affect traffic to a site.

That report is the productized answer to a data gap that persisted across the industry for more than a year. Google finally gave Search Console its own generative AI visibility reports in June 2026, exposing five categories of data: impressions, pages, countries, devices, and dates. The reports do not create a separate data silo. Generative AI impression data is included in the overall performance report while also being surfaced through a dedicated view for generative AI features.

The absence of exactly this kind of reporting had previously produced confusion. A fabricated screenshot of a Search Console AI Overviews filter circulated in September 2025 before Google's John Mueller debunked the fake announcement, an episode that captured how badly publishers wanted the measurement Google has now shipped. When AI Mode clicks were first folded into Search Console totals in June 2025, the data merged with the Web Search bucket, leaving no way to isolate AI Mode performance.

The regulatory backdrop

The control did not originate as a voluntary product decision. Google reacted to a UK order with the Search Console AI opt-out toggle on June 3, 2026, the same day the UK Competition and Markets Authority imposed its first binding conduct requirement on Google under the Digital Markets, Competition and Consumers Act 2024.

The CMA's requirement forced Google to give publishers AI opt-out rights, covering AI Overviews, AI Mode, and Gemini training. The substantive obligations of that Publisher Conduct Requirement come into legal force on December 3, 2026. The scope reaches beyond the search features named in the Search Console documentation to include the Gemini AI Assistant and the Vertex AI API, and it requires that publisher controls be effective, clearly explained, and technically enforced.

A limitation built into the current control traces directly to that regulatory record. The version documented for Search Console operates at the domain level. It tells Google's systems not to return an entire site's content within the covered features, rather than blocking individual URLs or subdirectories. The CMA's final decision recorded that Google told the regulator implementing page-level grounding controls would require additional engineering work. The regulator accepted that position and set a nine-month implementation period specifically for page-level controls, placing their arrival on March 3, 2027.

Google executives had signaled the engineering difficulty months earlier. Google described letting publishers skip AI Overviews as a huge engineering challenge at the FT Strategies News in the Digital Age conference on February 11, 2026, where a senior executive framed the problem as blocking content from AI Overviews and AI Mode while leaving pages indexed and ranked normally in conventional results.

Why the decision carries weight

The stakes attached to the include-or-exclude choice come from the scale of the features and the documented erosion of referral traffic they are associated with. AI Overviews count more than 2.5 billion monthly active users, and AI Mode crossed one billion monthly active users by May 2026, according to figures cited at Google I/O 2026. A property that excludes itself forgoes any traffic and impressions from those surfaces while retaining its standing in conventional search.

The traffic case for keeping content in AI features has grown harder to make. Ahrefs research found that AI Overviews correlate with a 58% reduction in click-through rates for pages ranked at position one, nearly double the 34.5% figure documented in earlier analysis of 300,000 searches. A later randomized field experiment produced causal evidence: AI Overviews cut publisher clicks by 39.8% in the first randomized study, involving 1,065 desktop Chrome users, and raised zero-click searches by 34.5%, with clicks on sponsored results staying flat.

Those numbers frame the exclude option as something close to a false choice for many publishers. Withdrawing from AI features removes a site from surfaces where Google increasingly expects users to spend their time, yet remaining opted in means participating in a system that measurably diverts clicks away from the site producing the content. The individual publisher consequences are concrete: one Berlin expat guide lost 70% of its traffic to AI Overviews, its founder reported in May 2026.

The default position compounds the tension. Because every property begins in the include state, the burden falls on owners to notice the control, understand its layered scope, and act. Smaller operators without teams dedicated to monitoring Google product launches are the least equipped to exercise a right that requires attention to exercise. The European Commission opened a probe into Google's AI content practices in December 2025 over precisely these dynamics, and independent publishers have argued that an opt-out framework maximizes the content pool available to Google's features by capturing everyone except those with the resources to respond.

For marketers and site owners, the documented control changes the mechanics of a decision that was previously theoretical. The switch now exists, its propagation timing is stated, its inheritance rules are defined, and its boundaries with training and indexing are explicit. What the documentation does not resolve is the underlying arithmetic of whether appearing in a feature that summarizes content is worth more than the clicks that feature absorbs, and that calculation remains specific to each property's data in the Generative AI performance report.

Timeline

  • May 2024 - AI Overviews launch in the United States after debuting as the Search Generative Experience at Google I/O in May 2023
  • June 17, 2025 - AI Mode clicks, impressions, and position data begin counting toward Search Console totals, merged into the Web Search bucket
  • September 15, 2025 - John Mueller debunks a fabricated screenshot claiming Search Console had added an AI Overviews performance filter
  • December 2025 - The European Commission opens a probe into Google's AI content practices
  • February 11, 2026 - A senior Google executive describes a clean AI Overviews opt-out as a huge engineering project at a London publishing conference
  • June 3, 2026 - The UK Competition and Markets Authority imposes its first binding Publisher Conduct Requirement on Google; Google begins testing the Search Console opt-out toggle for a subset of owners
  • June 3, 2026 - Google launches dedicated Generative AI performance reports in Search Console, exposing impressions, pages, countries, devices, and dates
  • December 3, 2026 - Substantive obligations of the CMA Publisher Conduct Requirement come into legal force
  • March 3, 2027 - Page-level grounding controls scheduled to arrive after the CMA's nine-month implementation period
  • Present - Google publishes Search Console Help documentation for the Search generative AI control and rolls it out to a subset of website owners

Summary

Who: Google, through documentation and a rollout managed in Google Search Console, affecting website owners and publishers whose content is eligible to appear in Google's generative AI features. The control's origins lie in a binding requirement from the UK Competition and Markets Authority.

What: A Search generative AI control that lets a website owner include or exclude a property's content from AI Overviews, AI Mode, and generative AI features in Google Discover, with a third option to inherit the setting from a parent property. The control operates at the domain level, takes effect within 1 to 2 days subject to caching, does not affect conventional search ranking, and does not stop model training, which remains governed by Google-Extended.

When: The underlying toggle began testing on June 3, 2026, under a CMA conduct requirement whose substantive obligations take legal force on December 3, 2026. Page-level controls are scheduled for March 3, 2027. Google has now published Search Console Help documentation for the control and is rolling it out to a subset of website owners.

Where: In Google Search Console under Settings, in the Search generative AI section, applied per property across a structure of parent and child properties.

Why: AI Overviews exceed 2.5 billion monthly active users and AI Mode exceeds one billion, while research documents click-through reductions of 39.8% to 58% for affected pages. The control gives owners a formal mechanism to withdraw from those surfaces, though the default include state and the layered scope place the burden of action on individual owners.