Google flags subdomains that override your AI Overviews setting

Google now notifies owners of a top-level domain property in Search Console when one of their child properties overrides the parent setting for AI controls. Search Engine Roundtable caught the line on October 6. It was added to Google’s help document for the Search generative AI control, the setting that decides whether a site’s content can appear in AI Overviews, AI Mode and the generative AI features in Google Discover.

The setting takes one of three values. Include is the default for a top-level property and lets content show up as links and help ground AI responses. Exclude keeps the content out of those features, including out of their links. The third option, Inherit from parent, is the default for any property that has a parent: Google’s document says a property follows its closest parent that has stopped inheriting. Owners of a child property can leave that path and set their own value, and that is the override the new notification reports.

So a domain property’s value is the site’s value only until someone with owner access on a child property changes it. Take a made-up setup: a retailer holds the domain property for its site, an agency holds a URL-prefix property for the shop subdomain, and a support vendor holds one for the help center directory. Either outside owner can set Exclude on their property, and until now the retailer looking at Include on the domain property had no way to see it.

The new notification closes part of that gap, but only for one kind of property. Google documents it for owners of the top-level domain property, and the help page says nothing about accounts that have none. The likelier reading is that an organization running on URL-prefix properties alone, with no domain property ever created, has no page where an override surfaces at all.

Getting this wrong is costly in one direction. AI Overviews now appear for most big brand names, so an Exclude sitting on a subdomain nobody audited quietly removes that subdomain from the surface where brand queries get answered. Reversing it is not instant either. Google puts exclusion at one to two days after the control goes live, with some content taking longer because of caching and propagation. It gives no figure for how quickly content comes back, so a wrong Exclude takes days to undo rather than being a toggle.

The control is also narrower than its name suggests. Google’s document says outright that it does not affect AI training and points to the Google-Extended robots.txt token for that. It says nothing about other vendors’ bots such as GPTBot, ClaudeBot or PerplexityBot, and nothing about Googlebot’s ordinary crawling. Those need separate handling, and reaching for robots.txt to do it can reach further than intended: a robots.txt rewrite meant to stop AI bots took a site out of Google.

What to do

• Open the Search generative AI control page on every property, including each subdomain and subdirectory property, and record the value. A child inherits by default, but any owner of that property can change it, so read it there rather than assuming it matches the domain.
• Create the domain property if the account only has URL-prefix properties. It is the only place Google documents the override notification.
• Review owner access on each child property. Google’s document gives the choice to owners of the child property, which makes agency and vendor owner access part of this audit.
• After changing a setting, wait out the documented one to two days plus caching before deciding it had no effect, and put a recheck on the calendar rather than judging it the same day.