A site owner removed his site from AI Mode and AI Overviews via Search Console and shared the data, Search Engine Roundtable reported in its October 30, 2026, recap. When the site was removed from the AI features, its AI impressions fell immediately to zero, as expected. AI impressions had made up about 8 percent of the site's total impressions before the opt-out.
The experiment proves the Search Console AI-features control works as documented: opting out removes the site from generative AI features, and the impressions go to zero rather than tapering. That sounds obvious, but it is the first public single-site confirmation many SEOs have seen, and it removes the uncertainty about whether the toggle is honoured promptly or only eventually.
The 8 percent figure gives the decision a scale. For this site, AI features represented about 8 percent of total impressions, so the opt-out cost was bounded and measurable. For sites where AI impressions are a much larger share, the same decision costs proportionally more, which is why the toggle should be tested with data rather than decided on principle alone.
The same coverage notes a counterpoint: a presence in AI Overviews seems to create brand lift that leads to later searches and clicks. The sample is one site, and AI Overviews still appeared on the results pages during the test, with only this site absent from them, so the brand-lift claim is suggestive rather than proven.
Even as a suggestion, it complicates the opt-out decision. Impressions that do not click today may still build familiarity that converts tomorrow through branded search or direct visits. An opt-out that saves content from AI summarisation may also remove the brand from the answer where future customers first encounter it.
Do not opt out site-wide on a hunch. Start by reading the generative AI report to learn what share of impressions come from AI features, then consider a limited test: opt out a section, or a set of pages, and compare branded search, direct traffic, and total impressions before and after. Keep the test long enough to see branded-search effects, not just the immediate impression drop.
Document the baseline before changing anything, because the immediate zero is dramatic and can mislead. The question is not whether AI impressions fall, since they will, but whether total business outcomes rise, fall, or stay flat once brand lift and citation loss are included. Only a measured test answers that.
The single-site opt-out story is useful but limited, because it has no control group: the site opted out and impressions fell to zero, which proves the toggle works but says nothing about what total traffic would have done otherwise. A stronger test would opt out half of a site sections, or stagger the opt-out across similar pages, so the opted-out pages can be compared against pages that stayed in.
That design is harder to run, but it is the only way to separate the toggle effect from seasonality, news cycles, and algorithm changes happening at the same time. Sites planning opt-out tests in November should build the control group first, even if it means a smaller test, because a small controlled result persuades a board while a large uncontrolled result only starts an argument.
The brand-lift hypothesis also deserves a direct test: compare branded search volume for opted-out sections against sections that stayed in, over several weeks, to see whether absence from AI answers depresses later branded queries. If it does, the opt-out costs more than the lost impressions suggest; if it does not, the opt-out is cheaper than feared. Either result turns a philosophical debate into a measured decision.
Sources: PPC Land, October 30, 2026, reporting Search Engine Roundtable.
Linkbaits helps you create content that earns citations and backlinks. Start with the free tools and build your first linkable asset this week.
Start free trial