Google has introduced a new control mechanism that allows website owners to exclude their content from appearing in the company’s generative AI search features. Initially rolled out in the United Kingdom and limited to a selection of websites, the feature is currently in a testing phase, giving publishers and SEO professionals a first look at how content visibility can be managed within Google’s evolving AI-driven ecosystem.
This development, known as the Search Generative AI Control, represents a significant shift in how publishers can interact with Google’s AI-powered search results, including AI Overviews, the AI Mode, and AI features within Google Discover. For the first time, website owners have a straightforward, granular option to decide whether their content is used in these generative experiences, separate from their standard search rankings.
How the Search Generative AI Control Works
The new setting offers two distinct choices: content can either be included, with links, in Google’s generative AI features, or it can be completely excluded. This binary decision applies uniformly across all of Google’s generative search surfaces, from the concise AI Overviews that appear at the top of query results to the more conversational Google AI Mode and the curated content in Google Discover.
A screenshot shared by a user on Twitter illustrates the new toggle within Google Search Console, showing a clear “On” or “Off” switch for generative AI content usage. The option is designed to be simple for site owners to understand, but the implications for search traffic and brand visibility are profound.
Critically, Google has confirmed that this setting has no impact on a website’s performance in standard, non-generative search results. A publisher who opts out of AI features will still see their pages rank and be displayed in the traditional list of blue links. This separation is a key concession from Google, acknowledging that a publisher’s preference for AI-generated summaries should not penalize their organic search presence.
What It Does Not Control: AI Model Training
There is a crucial distinction that all website owners must understand: the Search Generative AI Control only governs the display of content in Google’s generative search features. It does not prevent Google from using that content to train its underlying large language models. A site can be blocked from appearing in an AI Overview while still being ingested as training data for the models that power that same overview.
For publishers who wish to prevent their content from being used for model training, a separate mechanism is required. The existing Google-Extended directive in the robots.txtcodecodecodecode file remains the appropriate tool for this purpose. However, Google-Extended is specific to certain systems, such as Bard (now Gemini) and Vertex AI, and does not automatically extend to all the training pipelines that feed Google’s search-based generative AI.
This layered approach to control means site owners must carefully consider their strategy. A publisher may choose to block display via the new control for brand or accuracy reasons, but must separately manage the training aspect if they have concerns about intellectual property or data usage.
Inheritance and Hierarchy: A Complex Configuration
The settings for the Search Generative AI Control operate on a hierarchical inheritance model, which can become convoluted for large domains with multiple sub-sections or subdomains. This is one of the more technically demanding aspects of the new system.
By default, a property (such as a URL path) inherits its setting from its nearest parent property that has had its manual configuration stopped. If no parent property has been manually adjusted, the setting cascades all the way up to the top-level domain. For example, a page at https://example.com/business/catscodecodecodecode will inherit its setting from example.comcodecodecodecode, unless an owner has manually configured a setting for https://example.com/business/codecodecodecode that breaks the chain of inheritance.
This means that a single change at the root domain can propagate to every sub-folder and sub-page, unless specific overrides are created. For large enterprise sites with diverse content strategies—where some sections are designed for AI visibility and others are not—this inheritance system requires careful planning and testing. Google has published a diagram of this inheritance hierarchy to help users visualize the flow of settings.
Testing Phase and Go-Live Date
Publishers who have been granted access to the Search Generative AI Control are currently in a testing window. The settings made during this period will not immediately affect search results. Google has indicated that the changes will take effect starting on June 17th. This grace period is intended to give site owners time to experiment with different configurations, understand the inheritance logic, and assess the potential impact on their traffic and content strategy before the controls go live.
After this date, any changes to the setting may take several days to propagate fully. Google attributes this delay to caching mechanisms and the time required to distribute the updated instructions across its various systems.
Strategic Implications for Publishers and SEOs
The introduction of the Search Generative AI Control marks a pivotal moment in the relationship between publishers and search engines. It signals a move toward a more modular and controllable search experience, where content can be curated for different types of results.
For publishers, the immediate question is strategic. Excluding content from AI Overviews could protect against the risk of low-quality or incorrect summarization that misrepresents the original content. It also prevents generative features from potentially cannibalizing click-through rates, as users may find the information they need directly in the overview without visiting the source site. On the other hand, inclusion in AI features offers visibility in a new and rapidly growing search surface, which could drive brand awareness and authority, even if the direct click-through metrics are lower.
For SEO professionals, this adds a new dimension to search management. Beyond traditional indexing, crawling, and ranking optimization, they must now consider a “generative search strategy.” This includes deciding which sections of a site are optimized for AI consumption, how to structure content for accurate summarization, and how to balance the trade-off between visibility and click protection.
The Broader Context of Generative AI in Search
This launch comes at a time when Google is aggressively integrating generative AI into its core search product, facing pressure from competitors like Microsoft’s Bing Copilot and emerging AI-native search tools like Perplexity. The ability for publishers to opt out of these features was a long-standing demand from the publishing industry, which feared a loss of traffic and control over their intellectual property.
Google’s implementation of this control, while not complete in terms of training data exclusion, represents a step toward addressing those concerns. It provides a formal, official channel for content exclusion, which is a more reliable method than the ad hoc attempts that some publishers had been using, such as noindexcodecodecodecode tags or blocking specific user-agent strings.
The complexity of the inheritance system, however, suggests that the tool is still in its infancy. It is likely that Google will refine this interface, potentially adding more granular controls for individual AI features (e.g., disabling for AI Overviews but enabling for AI Mode) or providing clearer visual feedback on how inheritance is being applied.
What Content Owners Should Do Now
For those with access to the test, the immediate priority is to audit their Google Search Console property hierarchy. Understanding the parent-child relationships between all subdomains and folders is a prerequisite for effective configuration.
Publishers should create a content audit, labeling each section of their site as “AI-friendly” or “AI-restricted,” based on factors such as: the uniqueness of the content, the likelihood of accurate AI summarization, the dependence on direct referral traffic, and the potential for brand harm from misrepresentation. This audit should then be mapped to the inheritance structure to ensure that the desired settings cascade correctly without unexpected overrides.
It is also advisable to review any existing robots.txtcodecodecodecode directives related to Google-Extended to ensure there is no conflict or confusion between the two systems. While they serve different purposes (display versus training), a coordinated strategy is necessary for comprehensive control.
The coming weeks will be a critical observation period for the entire search industry. As the systems go live on June 17th, the first real-world data on usage, performance, and publisher behavior will begin to emerge, shaping how generative AI and traditional search coexist in the long term.
