In a 28-day opt-out experiment published November 5, 2026, SEO Cyrus Shepard of Zyppy found that Google's AI Overviews and AI Mode accounted for roughly 23% of his site's search clicks — and that those clicks converted at a higher rate than its regular organic results.
2026-11-11 · by Moe Ameen
Most of what we know about Google's AI search traffic comes from aggregate studies, and almost all of it is bad news for publishers. On November 5, 2026, SEO Cyrus Shepard of Zyppy published a single-site counter-test: he used Google's Search Console opt-out toggle to remove his own site from AI Overviews and AI Mode, measured what disappeared, then opted back in and measured what returned. The opt-out ran roughly September 1 to September 29, 2026, paired with a matched opt-in window — a 28-day interrupted time-series, read through Search Console's standard performance data and its newer generative AI features report.
The toggle acted almost immediately. Within about an hour of flipping it off, the site stopped appearing in AI Overviews and AI Mode, and its daily AI-feature impressions — several hundred a day beforehand — fell by more than 99%, effectively to zero. Impressions came back nearly as fast after he opted in. In the example query Shepard screenshotted, the site kept its #1 position in regular organic results after opting out; only the citation links inside the AI answer went away. That's consistent with Google's stated behavior: the AI-features toggle is not a ranking signal for ordinary Search, so opting out costs AI-feature visibility without touching your blue-link rankings.
Then the numbers. On Shepard's site, AI features were about 10% of total search impressions but a larger share of clicks — roughly 21% directly, and about 23% once he added a modest "halo" effect: appearing inside an AI Overview lifted the click-through rate on the site's regular organic listing beneath it by a couple of percentage points. He estimated the click-through rate on AI-search appearances at around 0.9%, higher than the site's overall rate near 0.4% — so the AI traffic was not just volume, it converted at a better rate than the average listing.
Shepard is careful about the scope. This is one site, measured once; AI visibility varies enormously by niche, query mix, and how often a site gets cited, so other publishers may see far less (or more). The opt-out covers only AI Overviews and AI Mode — it does not govern AI training, which is a separate control. His conclusion for his own site was plain: AI search sent meaningful traffic, so he will keep opting in. Before treating the opt-out as a way to "protect" traffic, check what those features actually send you.
The takeaway from this experiment is not "flip a toggle." It is that being present and citable inside AI answers can be worth real, higher-intent traffic — and the way to earn that is to have on-brand content living across the surfaces those answers pull from. Google's AI Overviews and AI Mode increasingly quote social posts, short video, and cleanly structured articles, not just one ranking page. Hand-feeding every one of those surfaces is exactly the work that does not scale for a single creator or a small team. That footprint is what [Kompozy](/) is built to produce.
Drop one source — a post, a transcript, a product update — into Kompozy's Quick Ingest and it fans into a brand-exact [Carousel](/glossary/hyperframes) breakdown, captioned [Persona Shorts](/glossary/persona-shorts) where a face-locked avatar delivers the point on camera, a Blog Article, an Email Newsletter, and native Text Posts, all held to one voice by a [Persona Brief](/glossary/persona-brief) and published across the eight social platforms plus blog and email on [Autopilot](/glossary/autopilot) behind a per-post review gate. That is the citable footprint an answer engine retrieves from. Then close the loop the way Shepard did: after Kompozy publishes, read Search Console's generative AI features report to see which pieces are getting retrieved, and make more of what works. Build the footprint, then measure it — see our guide to [Google AI Overviews and web traffic](/guides/google-ai-overviews-and-web-traffic) for the full measurement picture.
In the example query the experiment showed, no — the site kept its #1 organic position after opting out, consistent with Google's statement that the AI-features toggle is not a ranking signal for ordinary Search. What disappeared was the AI-feature traffic and the citation links inside the answers — AI impressions fell by more than 99% within about an hour.
On Shepard's site, AI Overviews and AI Mode were about 10% of total search impressions and roughly 21% of clicks — about 23% once a ~2% halo effect on regular organic clicks is included. The click-through rate on AI-search appearances (~0.9%) was higher than the site's overall rate (~0.4%). It is one site, and the share varies widely by niche and how often a site is cited.
No. The Search Console toggle governs whether your site appears in and helps ground AI Overviews and AI Mode only. AI training is a separate control, handled through robots.txt directives and Google-Extended. Opting out of AI search does not opt you out of training, and vice versa.
Google Search Console's generative AI features report shows AI impressions and which pages appear in AI responses; compare it against your total clicks to estimate the share. As this test showed, a short opt-out followed by an opt-in reveals the delta directly, because AI-feature impressions change within about an hour of flipping the toggle.