Google AI Overviews CTR ~0.9%: What Cyrus Shepard’s Opt-Out Test Means for Publishers (2026)

Publisher takeaway: A small late-October 2026 case study from SEO consultant Cyrus Shepard (Search Engine Roundtable spelling; Search Engine Land sometimes writes “Shepherd” for the same LinkedIn test) estimates that Google Search generative AI impressions — AI Overviews and AI Mode — clicked through at roughly ~0.9% CTR. More important for publishers: when the site was opted out via Search Console, regular organic clicks also fell. Shepard frames this as a possible “Search SERP Halo Effect”: earning an AI citation may lift clicks elsewhere on the SERP, not only from the AI answer itself. Treat Exclude as a high-stakes experiment, not a casual toggle.

Industry coverage from Search Engine Land (Oct 29) and Search Engine Roundtable (Oct 30) walks through the LinkedIn notes. This article is news + implications — not another how-to for the switch. For the Include / Exclude control itself, use our Search Console generative AI control guide.

What Cyrus Shepard actually tested

Shepard used Google’s Search generative AI controls in Search Console to block a site from appearing in AI Overviews and AI Mode, then compared impressions, clicks, and CTR before, during, and after. Critical nuance from Barry Schwartz’s clarification with Shepard:

• AI Overviews still appeared on the SERP before, during, and after the experiment.
• The site was cited in those Overviews before and after — but not during the opt-out window.
• AI impressions went to 0 immediately when generative AI visibility was disabled.
• ~0.9% is an estimated CTR on AI impressions, derived by looking at overall site CTR and impressions and subtracting known AI impressions during the experiment — not a pixel-perfect per-impression AI click log.

In other words: this is not a clean A/B of “SERP with AI Overview vs SERP without.” It is “my brand is in the Overview vs my brand is missing from an Overview that still exists.” That distinction matters when you interpret the organic-click drop.

The key numbers publishers are quoting

• ~0.9% estimated CTR on AI impressions (before/after analysis of overall clicks).
• AI impressions ≈ 8% of total impressions sitewide during the measured window.
• Regular organic clicks also dropped while AI features were disabled — the surprising finding.
• Average position rose overall (Shepard cited new posts + brand search), but URLs with the most AI visibility saw a drop, then a rise when AI visibility was restored.
• Overall impressions dropped during the opt-out period as AI impressions disappeared.

Neither SEL nor SER published the absolute traffic of the site, the exact test dates, or the size of the organic click decline. Shepard himself called it a small case study. Barry Schwartz and Danny Goodwin both flag limited sample size — useful directional evidence, not a universal rate card.

The “Search SERP Halo Effect” — why Exclude may cost more than AI clicks

Shepard’s framing: the more places your brand appears on a results page, the more likely a user clicks at least one of them. An AI citation or mention can act as brand proof that makes the classic blue link further down feel safer to click — even when the user never taps the AI answer. Greg Sterling (Near Media), commenting on the SER thread, said their research often sees Overviews create brand lift or awareness that shows up in later searches and clicks, not only in the first session.

If that mechanism is real even some of the time, then measuring AI value only via direct AI referral sessions understates the benefit of citations. It also raises the cost of casually choosing Exclude in Search Console: you do not merely lose AI referrals — you may lose halo-driven organic clicks while competitors remain cited in the Overview that still sits above the fold.

This sits alongside product changes that make citations more clickable — for example Google’s more prominent AI Mode inline links and AI Overview citation card summaries — and Google’s broader generative AI SEO guidance. Visibility in the answer unit is increasingly a brand-surface problem, not only a referral-CTR problem.

Include vs Exclude after this study

Choice	What you keep / lose	Risk after Shepard’s test
Include (default)	Eligible for AI Overviews / AI Mode citations + generative AI performance reporting	Still compete for citations; AI CTR may be modest (~0.9% in this sample) but halo + brand may help organic
Exclude	AI impressions → 0; site not cited in Search generative AI features	Organic clicks may fall even though Overviews still appear for users — you simply are not in them
Exclude temporarily (test)	Short measurement window if you annotate carefully	Only with a clear baseline, URL-level AI visibility list, and a restore plan — not a set-and-forget policy

Step-by-step for the control lives in our generative AI control walkthrough. Default recommendation for most publishers remains Include unless legal, licensing, or a deliberate negotiation strategy requires otherwise. Separately, Google’s invite-only AI contribution pilot pays for significant contribution during generation — not for citations alone — so Exclusion also removes you from that product surface if you were ever invited.

How to measure this without burning your own traffic

You do not need to opt out to start measuring:

• Search Console → Generative AI performance report: track AI impressions (and related visibility) for Overviews / AI Mode. Shepard’s sitewide share (~8% of impressions) is a useful benchmark to compare against — your niche may be higher or lower.
• URL-level cuts: note which URLs get the most AI impressions. In Shepard’s data, those URLs were the ones that wobbled when visibility flipped.
• GA4 AI referrals: Google still counts many AI answer clicks as Organic Search. Use our GA4 AI referral tracking guide so you are not blind to the direct click path — while remembering halo effects will not show up as a neat “AI channel.”
• Brand / assisted search: if Overviews create awareness, look for lifts in branded queries and return visits around high-AI-impression pages — directional, not causal.
• Do not rewrite only for CTR hacks: Google’s systems still reward helpful, non-commodity pages. Citation work remains retrieval + clarity — see getting cited in AI search and our GEO overview.

Practical publisher checklist (Oct 2026)

1. Leave the generative AI control on Include unless you have a written reason to Exclude.
2. Do not opt out casually to “force users to the blue link” — this study suggests you may lose both AI and some organic clicks.
3. Measure AI impressions weekly in Search Console; flag URLs where AI share of impressions is high.
4. Treat citations as brand lift + referral opportunity, not only as a 0.9%-class CTR channel.
5. If you must test Exclude: pick a short window, export generative AI + classic Search performance first, restore Include on a calendar reminder, and document that Overviews likely still appeared without you.
6. Keep shipping citation-worthy pages: clear answers, original data, updated dates — the same work Google’s generative AI SEO docs already describe.

---

FAQ: Google AI Overview CTR and the opt-out halo effect

Is ~0.9% CTR from AI Overviews / AI Mode proven industry-wide?

No. It is an estimate from one small Cyrus Shepard case study covered by Search Engine Land and Search Engine Roundtable in late October 2026. Useful as a directional benchmark; not a universal CTR.

Did turning off AI visibility remove AI Overviews from Google?

No. Shepard clarified that AI Overviews still appeared on the SERP during the experiment. The site simply was not cited in them while Exclude was active.

Why would regular organic clicks drop if AI impressions go to zero?

Shepard’s hypothesis is a Search SERP Halo Effect: appearing in the AI answer may increase trust and clicks on classic organic results (or later searches). Without the citation, some of that lift disappears even though the Overview unit remains.

Should publishers Exclude their site from generative AI in Search Console?

For most sites, no — keep Include. Exclusion zeros AI impressions and, in this small test, coincided with lower organic clicks. Use Exclude only with a deliberate policy (legal, licensing, or a carefully instrumented test). See our control guide for the exact setting.

How should I track AI value in analytics?

Combine Search Console’s generative AI performance report (impressions / visibility) with GA4 techniques that surface AI referrals inside Organic Search. Expect brand and halo effects to be under-counted if you only look at direct AI click sessions.

---

Bottom line: Shepard’s small opt-out study gives publishers a rare before/after signal — roughly 0.9% estimated CTR on AI impressions, AI share around 8% of impressions on that site, and a warning that Exclude can hurt classic organic clicks via a possible SERP halo. Keep measuring AI impressions, default to Include, and invest in earning citations as brand surface area — not only as a thin referral CTR line.