Can AI search skip your website? Google's opt-out control, explained

The question

Can I keep my site out of Google's AI Overviews and AI Mode?

The short answer

Yes. The Search generative AI control in Search Console excludes your links and content from AI Overviews, AI Mode, and Discover's AI features, without affecting regular Search ranking or AI training.

Source: Search generative AI control, Google, Generative AI performance report, Google, The Internet has a second audience, Cloudflare

TL;DR: Search Console now has a "Search generative AI control." Set it to Exclude and Google drops your links and content from AI Overviews, AI Mode, and AI features in Discover, worldwide. Google says it isn't a ranking signal for the rest of Search, and it doesn't affect AI training (Google points to Google-Extended for that). My default for businesses that sell something is to stay in. Check your own data before you decide.

A site owner wants to stop AI training. They find a Search Console setting with "generative AI" in the name and switch it to Exclude. That setting does something else.

That's Google's Search generative AI control. Before changing it, check what it covers.

What the control does

Google's Search Console help page describes it plainly. The control lives under Settings > Search generative AI in Search Console. Google says it rolled the control out "to all websites worldwide" as of 31 August 2026.

It covers three features today:

• AI Overviews
• AI Mode
• Generative AI features in Google Discover

Google says it expects to update that list over time. Bookmark the page.

There are three settings:

• Include (the default). Your content can appear in these features, as links and as grounding for AI answers.
• Exclude. Your links and content stop appearing, and content crawled from your site can't be used as input for these AI responses. You get no traffic or impressions from these features.
• Inherit control from parent. The default for sub-properties, such as a blog folder or a subdomain.

Google says a change generally takes a few days to apply. Content is excluded within one to two days after the control goes live, though some can take longer because of caching and propagation.

What it doesn't do

Here are the limits, per the same help page. This is where the "one click to block AI" posts get quiet:

• It isn't a ranking signal. Excluding your site doesn't push you down in regular Search.
• It doesn't affect AI training. To limit training of the models behind these features, Google points to Google-Extended.
• It doesn't fully remove you from Google's systems. Google says your content may still help power Search as a whole, for example in understanding the language of queries and pages.
• It doesn't remove you from Search. That's noindex.
• It doesn't make the topic disappear. "Content from other sites will still be available in those features, and it may appear similar to yours."
• It only covers Google's features listed above. It isn't a setting for other companies' AI assistants.

The parent and child trap

A property inherits from its closest parent that has manually stopped inheriting. If no parent has, it follows the top-level domain property. Any property can also set its own value. So the owner of a child property, like blog.example.com or https://example.com/shop/, can override the domain's setting for that property.

Nothing says "we have a governance process" like the blog subdomain quietly opting out. There's now a guard for this. The help page says owners of the top-level domain property "will see a notification" on the control page when a child property overrides the parent setting. If several teams or agencies have Search Console access, check that page.

Why this decision is bigger than it was

Cloudflare's 30 September 2026 post, "The Internet has a second audience" gives the context. It says that this year, for the first time, more than half of internet traffic wasn't human, and that daily requests from AI agents on its network grew by more than 1,700% over the past year. It also says some heavily crawled categories, like retail and computer software, saw human traffic fall "as much as 40% in less than one year." None of these figures measures what Google's setting would do to your site.

The same post says fewer than 1% of sites on Cloudflare block search crawlers, while 17% block training. My read: site owners treat being found and being trained on as separate questions.

Google's control lets you decide whether your links and content appear in its AI search features. Training is a separate choice, and wider agent traffic is a separate problem.

How to decide: four questions

These are my decision rules, not Google's.

1. How visible are you in these features today? Open the Generative AI performance report in Search Console. It shows how often links to your site appeared in AI Overviews and AI Mode, by page, device, country, and date. It's an impressions report. It doesn't show visits, leads, or sales, so it can't tell you on its own whether inclusion pays. If the report is missing, Google says your site usually hasn't had enough impressions in these features, and exclusion may be the reason. Google says it can help you estimate how changing the control may affect traffic. If those features show your links often, excluding gives that up.
2. What do you sell? If your page exists to bring in leads or sales, being a linked source in an AI answer is one more way to be seen. Giving that up needs a reason. If you sell the content itself, like subscriptions or paid research, ask whether an AI answer could give away what customers pay for. That's a real reason to consider excluding.
3. Will someone else fill the gap? Google says content from other sites may still appear and may look similar to yours. If competitors cover the same ground, excluding may simply leave them as the sources shown. If you're the only source of the material, my guess is that excluding matters more, though neither source measures that.
4. Are you trying to stop training instead? Then this is the wrong switch. Google points to Google-Extended for limiting training of these models.

What I'd do this week

• Open Settings > Search generative AI and confirm what each of your properties is set to.
• Check for the child-override notification if you have sub-properties.
• In the Generative AI performance report, pick a date range, write it down, and note impressions and the pages that get them. (Discover has a separate report.)
• Write down your decision (stay or exclude), why, when you'll review it, and what evidence would make you reverse it. Share it with whoever owns the site.
• If you exclude, log the date you changed it and allow for Google's stated delay. Pick comparable periods before and after. Track organic traffic and qualified leads separately, and note anything else that changed, like seasonality, campaigns, or site edits. Treat it as a rough test, not proof. A before-and-after comparison can't fully separate the setting's effect from everything else that changed.
• Separately, decide whether to limit training of the models behind these features. Google points to Google-Extended for that.

My default for a business that sells services or products is to stay included and make the pages worth citing. That's judgment, not a measured result. Your report and your business model can change it.

For why search features and browser agents need separate decisions, see Search is one job. A browser agent is another.