Every enterprise SEO team we talk to has the same question queued up for their next leadership meeting: should we flip the Google AI Overviews opt-out? It is the wrong first question. The right one is whether you could tell the difference afterward — and right now, for almost everybody, the answer is no.
Here is the state of play. Google acknowledged in January 2026 that it was exploring opt-out options. In June the UK's Competition and Markets Authority imposed a conduct requirement on Google under its new digital markets powers, and Google began testing a toggle with UK properties inside the same week. From June 17 it started honoring the setting. By July the control had appeared on non-UK accounts, and it now sits under Settings in Search Console for anyone who goes looking for it.
That is a genuine shift. For two years the only honest advice on AI answers was that you could block the crawler and lose the search results with it, or stay in and take what you got. A control that removes you from AI Overviews, AI Mode, and Discover's AI features while leaving your blue links alone is the thing publishers have been asking for since the first AI Overview shipped. It just arrived without any of the instrumentation you would need to use it well.
What the Google AI Overviews opt-out actually controls
The confusion in most of the threads we read comes from treating four different controls as one. They are not interchangeable, they were built at different times for different reasons, and only one of them does what people mean when they say opt out. We covered the two-tier structure when the exclude-versus-remove split first landed, and the shape has held: the Search Console setting is the generative-features control, and everything else is either older or aimed at model training rather than search surfaces.
| CONTROL | WHAT IT ACTUALLY DOES | WHERE IT LIVES | RESOLUTION |
|---|---|---|---|
| Search Console generative AI setting | Removes content from AI Overviews, AI Mode, and Discover's AI features; Google states it is not a ranking or inclusion signal elsewhere in Search | Search Console → Settings | Property-wide until March 2027 |
| Google-Extended | Controls use of your content for training future Gemini models; does not affect Search inclusion or whether you appear in AI Overviews | robots.txt token | Path-level |
| nosnippet / max-snippet | Suppresses the text excerpt Google may show, including in some generative surfaces, at the cost of your regular result snippet too | Meta robots tag | Page-level |
| noindex / robots.txt disallow | Removes the page from Search entirely, generative surfaces included, because there is nothing left to surface | Meta tag or robots.txt | Page-level |
Read that table twice, because the fourth column is the whole argument. Google-Extended, nosnippet, and noindex all operate at the level of a path or a page. The one control that does the thing you want operates at the level of an entire property. A publisher who wants to keep its evergreen guides out of AI Overviews while leaving its news desk in cannot express that today. Neither can a SaaS company that would happily feed its documentation to AI Mode while keeping its pricing page out of a generated comparison.
The CMA wrote this timeline, not Google
It matters who set the dates, because it tells you what the dates mean. This is not a product roadmap Google published because the feature was ready. It is a compliance schedule. The CMA's conduct requirement carries obligations that run into 2027: page-level controls due by March 2027, and click-data sharing starting in December. Google shipped the property-level toggle first because it was the fastest thing to ship that satisfied the order.
A compliance schedule is a useful thing to plan against, because regulators publish deadlines and product teams publish intentions. If you are building a roadmap for the next three quarters, December and March are firmer dates than almost anything else on your calendar. It also means the current design is a floor, not a ceiling. Complaining that the toggle is too blunt is fair, and it is also temporary in a way that most Google behavior is not.
The Google AI Overviews opt-out ships without click data
This is the part that should stop you. To evaluate the switch you need to know what an AI Overview citation is worth in clicks. That number does not exist in your reporting today, and it does not exist in Google's public data either. Search Console has never separated clicks that came from inside a generated answer from clicks that came from a standard result. The click-data sharing obligation that starts in December is the first mechanism that will begin to close that gap — which is another way of saying the evidence arrives after the decision.
We have written before about the reporting gap that leaves AI-driven visits undercounted and misattributed, and this is the sharpest version of it yet. Opting out is a change with no clean before-and-after. Suppose you flip the setting and organic sessions fall four percent over the following month. Was that the opt-out? A core update? Seasonality? Your own publishing cadence? With no isolated citation-click metric on either side of the change, you cannot separate them, and you will end up arguing about it in a meeting instead of resolving it with a query.
Illustrative scenario — how a property-wide opt-out lands differently by content type. Worked example to show the trade-off shape, not measured data.
The point of that illustration is not the numbers, which are ours and hypothetical. It is that a single toggle applies one answer to four situations that want four different answers. Any team with a mixed catalogue is choosing the least-bad average, and the honest version of that decision names which pages are being sacrificed to protect which others.
What you give up in Top Stories
For news publishers there is a concrete cost hiding inside the abstract one. NewzDash data puts Top Stories inside AI Overviews on nearly one in six US trending news queries — 15.5%, rising to 17.46% in the UK. That placement is not a link buried in a source drawer. It is the news carousel, embedded in the generated answer, in the position where trending demand actually lands.
“Using Google's newer Search Console generative AI opt-out is different, and will likely remove publishers from Top Stories inside AI Overviews.”
That is John Shehata, CEO of NewzDash, and it is the most specific warning anyone has put on the record about what the toggle costs. Rahul Jain, CEO of Noble, put the general case more bluntly: opting out of Google's AI Overviews will hurt most publishers more than it helps. Neither claim is settled — nobody has the click data to settle it — but both are worth more than the confident takes circulating with no mechanism attached.
The counter-argument deserves a fair hearing, because it is not stupid. If generative answers are absorbing the demand your reporting creates and returning a fraction of the traffic, then staying in is subsidizing a competitor with your own content. That is the logic behind the wave of publishers who have blocked Google outright and published the referral numbers to justify it. The difference is that those teams measured first and blocked second. They have referral data. The opt-out has none.
How to decide before December
One caveat on the wait-and-see verdict: it is a default, not a law. If your content is being reproduced at length inside generated answers with no meaningful click return, and you have the referral data to show it, opting out is defensible today. Legal exposure changes the math too — a European court has already held Google directly liable for false statements about publishers in AI Overviews, and if your category carries that kind of risk, presence in the answer is a liability and not just a traffic question.
Do this next
Open Search Console and confirm the setting exists on your property, so you know what you are working with. Do not touch it. Then pull your top two hundred URLs by organic value and tag each one as generative-positive or generative-negative based on whether an AI answer that quotes it plausibly sends you a qualified visit. That segmentation, plus a frozen prompt set logged weekly, is what turns December's click data into a decision the same afternoon it lands instead of another quarter of debate.
The teams that will get this right are not the ones with the strongest opinion about AI Overviews. They are the ones who spent the eight months before page-level controls arrived building the measurement that makes the choice obvious. If you want help standing that up, our reporting and analytics work starts exactly here, and our generative engine optimization practice treats citation presence as a tracked metric rather than a vibe. For the background on how these controls were structured at launch, read our breakdown of the exclude-versus-remove split, the gap in AI search visibility reporting, and the referral data from publishers who blocked Google outright. Search Engine Journal's rundown of the current controls is the clearest public summary of the timeline.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.