AI Overviews Opt-Out: Myth vs. Fact on What It Does

Google lets site owners pull their content out of AI Overviews, AI Mode, and Discover’s AI summaries. It does this without touching regular search rankings, through a Search Console control it has been expanding since June 2026. A fresh case study reported by Search Engine Roundtable on October 30, 2026, is the first real test of what flipping that switch actually buys you. The results are messier than the simple “opt out and get your clicks back” pitch suggests.

I have fielded questions about this setting since it rolled out. My honest answer has always been “it depends.” This piece separates the myths building up around the AI Overviews opt-out from what the evidence, thin as it still is, actually shows.

What the AI Overviews Opt-Out Control Actually Does

The control sits inside Search Console under Settings, in a section labeled Search generative AI. Site owners choose one of three states for a property. They can include it in AI Overviews, AI Mode, and Discover’s AI-generated answers (the default), exclude it from all three, or inherit the setting from a parent property.

Excluding a property stops Google from showing that content inside AI-written answers and stops it from being used to generate those answers. According to Google’s own documentation on AI features and your website, changes usually take effect within one to two days, though cached responses can take longer to clear.

Google began testing the toggle on June 3, 2026, starting with a subset of site owners before widening access over the following months. If you do not see the Search generative AI section in your Search Console settings yet, that is expected. It is still not universal five months in.

This is a separate decision from two other controls people often lump together with it. Noindex removes a page from Search entirely, including regular blue-link results. Google-Extended governs whether your content can be used to train Google’s AI models in the first place. The AI Overviews opt-out touches neither of those. It only governs whether your existing, indexed content can be pulled into an AI-written answer at the moment someone searches.

The reason this setting exists at all is the growing share of queries that now trigger an AI Overview before a single blue link appears. Publishers have spent most of 2026 asking whether that box helps or hurts their traffic, and Google built the opt-out partly in response to that pressure. Choosing to exclude a site is, in effect, a bet that losing AI visibility costs you less than the clicks you believe AI Overviews are absorbing. That bet is exactly what this case study tries, and mostly fails, to settle cleanly.

The Case Study Behind the Current Debate

SEO consultant Cyrus Shepard published a case study on LinkedIn describing what happened when he used the opt-out control on one site. He excluded it from AI Overviews and AI Mode while leaving everything else unchanged. Barry Schwartz covered the results for Search Engine Roundtable on October 30, 2026.

Before the test, AI impressions made up roughly 8% of the site’s total search impressions. Shepard estimated the click-through rate on those AI impressions at around 0.9%, working backward from the site’s overall CTR and impression counts rather than from a number Google reports directly.

Once the property was excluded, AI impressions for that site dropped to zero almost immediately, which is exactly what the control is supposed to do. What happened next is the part worth sitting with: regular organic clicks fell too, rather than recovering the way a clean “AI took my traffic back” story would predict.

Sitewide average position improved slightly during the test. But on the specific URLs that had carried the most AI visibility before the opt-out, position actually dropped, then recovered once AI visibility was restored at the end of the test. Shepard did not publish raw click counts, the number of URLs involved, or a confidence interval, and this remains a single-site, self-reported test rather than a controlled study.

One more wrinkle: AI Overviews kept appearing on the same search results pages throughout the test, generated from other sites’ content. Only Shepard’s test property was excluded from contributing to them. That matters because it means the experiment measured what happens when one site steps out of an AI answer that still exists for that query. It did not measure what happens when AI Overviews disappear altogether.

Myth vs. Fact: Does an AI Overviews Opt-Out Win Back Your Clicks?

Here is where the practitioner chatter and the actual data start to pull apart. I have grouped the most common claims I am hearing about the AI Overviews opt-out into myth and fact pairs, graded against what Google documents and what this case study showed.

Myth: Opting out hands your lost clicks straight back

This is the pitch I hear most often: exclude your site from AI Overviews, and the clicks AI was “stealing” flow back to your blue links. Shepard’s test does not support that. Regular organic clicks fell during the opt-out window rather than rising, even as AI impressions dropped to zero. One plausible read is that AI Overviews on that query still existed, built from competitors’ content. Searchers who clicked after reading the box were now reading a summary sourced elsewhere instead of this site’s.

Myth: AI Overviews are taking the bulk of your clicks, so this is a major lever

AI impressions accounted for about 8% of this site’s total impressions, with an estimated 0.9% click-through rate on those specific impressions. That is a small slice of overall visibility for this property, not the majority many site owners assume. Your own split will vary by query type and industry, and you can check it directly in Search Console’s AI performance reporting rather than guessing from anecdote.

This is the number I would want from any site before recommending an AI Overviews opt-out at all. A property where AI impressions sit at 8%, with sub-1% CTR on those impressions, has a different cost-benefit calculation. A site where AI Overviews dominate a large share of its top queries faces a much bigger decision. Pulling the opt-out lever on a small slice of traffic is unlikely to move your overall numbers much in either direction. That alone is useful to know before you spend time testing it.

Myth: Excluding your site changes your regular rankings

Google states plainly that this control is not a ranking or inclusion signal for standard Search results. The case study is broadly consistent with that: sitewide average position held steady or improved slightly. The one nuance is that URLs with heavy prior AI visibility saw their position dip during the opt-out and recover afterward. That looks more like measurement noise tied to how those specific queries get tracked, not an actual ranking penalty.

Myth: This opt-out also keeps Google’s AI from training on your content

These are two different switches. Google-Extended handles AI training opt-out. The Search generative AI control in Search Console only decides whether your pages can appear inside AI Overviews, AI Mode, and Discover’s AI summaries at query time. Flipping one does nothing to the other, and conflating them is an easy way to end up with the wrong setting toggled.

Myth: Every site already has access to this control

Google is still rolling this out in phases that started in June 2026. A site owner without the Search generative AI section in their settings has not necessarily done anything wrong. The option may simply not have reached that property yet.

Myth: One case study settles whether opting out helps or hurts

It does not, and treating it that way would be the biggest mistake here. This is one site, one short window, and a methodology built on estimation rather than Google-confirmed numbers. Greg Sterling, writing about related Near Media research, has raised a separate point worth weighing. AI Overviews may generate brand exposure that shows up as clicks later, outside the test window a short case study can capture. My own take is that the honest fact right now is “we do not fully know yet,” not a confident verdict in either direction.

Where This Leaves Site Owners Weighing an AI Overviews Opt-Out

If you are deciding whether to flip this switch, start with your own numbers instead of borrowing someone else’s. Check how large a share of your impressions are AI-driven before assuming the opt-out is worth the experiment at all. For a lot of sites, that share will be small enough that the decision barely moves the needle either way.

If you do test it, run it on a limited set of pages or a short window first, and watch AI impressions and plain organic clicks side by side instead of only one of the two. Shepard’s case study is useful precisely because it measured both and found they did not move the way a simple “AI stole my clicks back” story predicts.

Keep the two controls straight. If your goal is keeping content out of AI-written answers, use the Search generative AI setting. If your goal is blocking AI training on your content, that is a Google-Extended decision, covered separately from this one. Mixing them up is the most common mistake I see teams make when this topic comes up.

My practitioner’s take is that this control is worth testing carefully on a property where you can tolerate some noise, not worth flipping sitewide on the strength of one LinkedIn post. The data so far says “it’s complicated,” and complicated is a reasonable place to sit until more sites publish real before-and-after numbers. Our recent look at how often AI Overviews link out at all is a useful companion read if you are trying to size up how much upside an opt-out could realistically offer on your own site.

Agencies fielding client questions about this should set expectations before anyone flips the switch. A client who read one LinkedIn post and now wants every property excluded by Friday is a predictable conversation. The honest answer is that nobody, including Google, has published enough data yet to promise a specific outcome. Document the AI impression share you are starting from, pick a defined test window, and agree in advance on what counts as success before the opt-out goes live. That discipline matters more here than in most SEO experiments, because the one public test so far produced a result nobody was expecting.

FAQ: AI Overviews Opt-Out

What is the Google AI Overviews opt-out control?
It is a setting inside Search Console, under Settings in the Search generative AI section. It lets a site owner exclude a property from appearing in AI Overviews, AI Mode, and Discover’s AI-generated summaries.

How do I turn on the AI Overviews opt-out in Search Console?
Open Search Console, go to Settings, find the Search generative AI section, and choose Exclude for the property. Google says changes typically apply within one to two days, with some caching delay possible.

Does opting out of AI Overviews hurt my rankings?
Google states it is not a ranking or inclusion signal for standard Search results. A recent case study found sitewide position held steady or improved slightly, though URLs with heavy prior AI visibility saw temporary position movement.

Is the AI Overviews opt-out the same as blocking AI training?
No. Training opt-out runs through Google-Extended. The Search generative AI control only governs whether your content can appear inside AI-written answers at query time.

Will opting out get my lost clicks back?
Not reliably, based on the one public case study available so far. Regular organic clicks fell during that test rather than recovering, likely because AI Overviews kept appearing on the same queries using other sites’ content.

Is this control available to every website yet?
Not as of this case study. Google began testing it on June 3, 2026, and has been widening access to more site owners in phases since then.

Can I apply the AI Overviews opt-out to just part of my site?
Yes, at the property level. Search Console applies the setting per property, and a child property can inherit its parent’s choice or override it. That means you can exclude a subdomain or a section you have verified separately while leaving the rest of the site included.