Google has quietly completed one of the most important control changes for publishers since generative search began: a website can now opt out of Google’s AI search experiences without removing itself from traditional Google Search.
The official Search Console documentation for the Search generative AI control says the setting is now available to all websites worldwide following a rollout completed on August 31, 2026. It covers AI Overviews, AI Mode and generative AI features in Google Discover.
When a publisher chooses exclusion, Google says the site’s links and content will no longer be visible to users in those generative features, including as links or as material used to ground AI responses. The site will also stop receiving impressions and traffic from the affected AI experiences. Crucially, Google separates that decision from ordinary Search inclusion: publishers do not need to use noindex simply because they want to stay out of generative AI Search.
This creates a boundary publishers have been asking for since AI-generated answers began occupying more of Google’s discovery surface. Traditional search visibility, generative search visibility and AI training are now documented as separate control problems.
The switch covers AI Overviews, AI Mode and generative Discover
Google’s documentation currently lists three product areas governed by the Search generative AI control: AI Overviews, AI Mode and generative AI features in Google Discover. The company also says it expects to update the list over time as Google Search evolves.
The setting is managed in Search Console under Settings and then Search generative AI. Inclusion is the default for properties unless another setting is inherited from a parent property.
When inclusion is enabled, Google says content can appear as links and can help ground AI responses in these features. That participation can generate impressions and traffic for the site.
Exclusion reverses those permissions for the covered Search experiences. The site’s content is no longer shown to users there as links or used to help ground the generated responses, and the property no longer receives impressions or traffic from those surfaces.
This is not a noindex switch
The most consequential part of the new architecture is what the control does not do. Google explicitly distinguishes generative AI Search exclusion from removing a page from Google Search altogether.
If a publisher wants content completely absent from Google Search, Google still directs the site owner to use noindex. The Search generative AI control is narrower: it governs participation in the specified AI-powered Search and Discover features.
That means a publisher can preserve conventional organic discoverability while declining to participate in AI Overviews, AI Mode and generative Discover.
For news organizations, specialist publishers and sites whose business models depend heavily on direct visits, that is a fundamentally different decision from disappearing from Google’s index.
Google has separated AI Search participation from AI training
The documentation draws another line that is equally important. Google says the Search generative AI control does not affect AI training.
Publishers that want to restrict model-training uses need to use Google-Extended, the separate control Google provides for specified AI uses. Google therefore now documents three distinct layers: conventional Search inclusion, generative Search participation and AI training.
NetContentSEO recently examined this distinction when Cloudflare introduced a way to express AI-training preferences without blocking Googlebot. The central warning in that analysis remains relevant: Google-Extended should not be mistaken for an AI Overviews or AI Mode removal mechanism.
The new Search Console control makes that separation much more operational. A publisher can make one decision about Google Search AI visibility and a different decision about training.
Exclusion removes more than visible citations
The wording of Google’s documentation goes beyond hiding a publisher link. When a site is excluded, its content is not visible to users in the affected generative features “including showing up as links and helping to ground AI responses.”
That grounding language matters because AI visibility has at least two layers. A source can influence the generated answer internally, and it can be exposed externally as a citation or link. Google’s control addresses both forms of participation for the covered Search features.
This is therefore broader than a cosmetic request to hide attribution. It is a request not to use the site’s content as part of the grounding process for those generative Search experiences.
Google also states the commercial consequence clearly: an excluded site will not receive traffic or impressions from those features.
The change generally takes one or two days to propagate
Search Console controls do not rewrite Google’s systems instantly. Google says that after a site changes the setting, affected content is generally excluded within one to two days.
Some material can take longer because of caching and propagation across Google systems. Publishers should therefore avoid testing the switch and expecting every AI citation or appearance to vanish immediately.
The same operational reality matters when reversing an exclusion. Large sites should plan for propagation rather than treating the control as a real-time traffic lever that can be toggled hour by hour.
For editorial organizations, that suggests the decision should be governed as a publishing policy rather than an ad-hoc Search Console experiment.
Parent properties can push the policy down to child properties
Google has also built inheritance into the control. By default, a property can inherit its Search generative AI setting from the nearest parent property that has explicitly changed the control.
Changes at the parent level can therefore flow automatically to child properties. A child property can override the inherited setting and configure its own choice manually.
This is especially significant for organizations with complex Search Console structures, multiple subdomains or segmented properties. AI participation can be governed centrally without requiring every property manager to make the same change individually.
At the same time, the override mechanism allows organizations to treat parts of a site differently where their publishing economics or content policies differ.
Google now gives publishers a measurement layer before they opt out
The control arrives alongside dedicated reporting for generative AI visibility. In its June 3 Search Central announcement, Google introduced Search Generative AI performance reports for Search Console, with dedicated reporting for Search and Discover. Google later updated the announcement to say those insights had rolled out to all websites worldwide by August 31.
Google explicitly recommends using the generative AI performance report to understand how content performs in these features before deciding how a control change might affect traffic.
That pairing of measurement and control is important. Publishers no longer have to make the decision entirely in the dark. They can evaluate AI-generated impressions and associated performance before deciding whether exclusion is economically sensible.
The reporting still has interpretation challenges. As NetContentSEO recently covered, Google has acknowledged that position measurement in increasingly complex AI Search interfaces is not an absolute physical metric. But dedicated reporting still gives publishers a first-party baseline for the visibility they would surrender by opting out.
Publishers can now price the AI Search trade-off more explicitly
Before a dedicated control existed, the debate around generative Search often collapsed into a binary question: allow Google to crawl or risk damaging conventional Search visibility.
The Search Console setting makes the trade-off more granular. A publisher can ask whether appearances in AI Overviews, AI Mode and generative Discover produce enough traffic, audience development, subscriptions, conversions or brand value to justify allowing its content to participate in those experiences.
If the answer is no, exclusion no longer requires the drastic step of removing the content from traditional Search with noindex.
That does not make the decision easy. A site that opts out gives up both the potential traffic and the potential brand exposure from the covered generative features. The new control creates agency, not a universally correct strategy.
AI Search visibility has become a separate distribution channel
The existence of a dedicated opt-out formalizes something SEO teams have increasingly observed in practice: generative Search is becoming its own distribution layer.
A publisher can rank conventionally while appearing differently in AI Overviews. It can be retrieved and linked in AI Mode through conversational query fan-out. It can surface in generative Discover experiences that do not begin with an explicit search query at all.
Google’s control groups these experiences under one generative AI participation decision while leaving conventional Search intact.
For reporting and strategy, that means “Google visibility” can no longer be treated as a single state. Traditional results and generative experiences can now be measured and controlled separately.
The control also exposes the real cost of opting out
Publisher control is meaningful only when the consequence is clear, and Google is unusually explicit here. If a site excludes itself, it will not receive impressions or traffic from the covered generative AI features.
That means the decision is not equivalent to asking Google to stop summarizing content while continuing to show all of the same publisher links. The opt-out removes the site from the relevant AI discovery opportunity.
For publishers that already receive meaningful AI referrals, the cost could be measurable. For sites receiving negligible traffic despite substantial AI visibility, the calculation may look different.
The new performance reports make that cost easier to quantify over time, particularly if organizations record a baseline before changing the setting.
Discover makes the control broader than query-based search
The inclusion of generative AI features in Google Discover is strategically important because Discover is not driven only by explicit user queries. It is a personalized recommendation environment.
Google has been experimenting with AI-assisted Discover experiences that insert generated topic exploration between a feed item and the publisher. A single control spanning Search and Discover therefore reaches across both active information seeking and algorithmic recommendation.
For publishers with substantial Discover audiences, the decision should not be evaluated solely through AI Overview or AI Mode traffic. The affected distribution footprint can extend into personalized feeds.
This makes the Search Console reporting split between Search and Discover especially useful when estimating potential impact.
Google can expand the list later
Google says it expects to update the list of covered features over time as it develops Search. That sentence prevents the current three-item list from being treated as permanent.
As generative features spread into additional Search surfaces, the same control may govern more experiences. Publishers will therefore need to monitor the documentation rather than assuming that a decision made today always maps to the same product scope.
This is particularly important for organizations with formal AI-use policies. A centrally approved setting can change practical meaning if Google later adds another high-traffic generative surface to the list.
The policy should therefore include periodic review of both the setting and the features it covers.
Technical SEO teams now need an AI participation policy
The new control is simple in the interface but significant in governance. Someone inside the organization needs authority to decide whether a site participates in Google’s generative Search ecosystem.
That decision should involve more than SEO. Editorial leadership may care about source use and attribution, commercial teams may care about referral value, legal teams may care about content-use policy and audience teams may care about brand discovery.
Search Console administrators should not toggle exclusion casually simply because the option exists. The setting can remove an entire category of impressions and traffic within days.
For large organizations, documenting who owns the decision is now part of technical SEO governance.
Google has finally decomposed the old crawl-or-disappear dilemma
The most important consequence of the rollout is architectural. Publishers no longer need to treat Google’s conventional index, generative Search features and AI training as one indivisible permission.
noindex remains the tool for keeping content out of Google Search entirely. Google-Extended addresses separate AI-training and related uses covered by that token. The Search generative AI control governs participation in AI Overviews, AI Mode and generative Discover.
Those are different choices with different consequences.
For publishers, that separation is more important than the existence of another toggle in Search Console. It means AI Search visibility has finally become a permission that can be evaluated on its own economics. A site can keep competing for traditional organic traffic while deciding that Google’s generated answers should not link to, ground themselves in or generate impressions from its content.
The era when opting out of AI Search appeared inseparable from opting out of Search itself is over. Google has now documented the boundary—and publishers can decide which side of it they want to be on.