Google AI Overviews Opt-Out: Exclude vs. Remove, Compared

Google has been testing new publisher controls that limit how sites show up in AI-generated search results since June 2026. MediaPost reported the rollout on July 21, and the mechanics matter more than the headline suggests. There is not one Google AI Overviews opt out — there are two, and they behave differently. The narrower option removes your content from AI Overviews and AI Mode summaries while your regular Search results stay untouched. The broader option removes you from AI search entirely. Per MediaPost's reporting, publishers cannot cleanly choose the narrow option without it affecting the broader one — the two settings are linked. Pick wrong and you either give up citation-driven influence you were getting for free, or you stay exposed to scraping you meant to shut off. Most publishers evaluating this decision right now are treating it as one switch. It is not, and the gap between the two tiers is wide enough to swing your entire AI search visibility one direction or the other.

The two tiers, explained

Start with what MediaPost actually reported, because the coverage is more specific about the tier boundaries than about the underlying mechanics. Google's test gives publishers two settings. The first excludes content from AI features specifically — AI Overviews, AI Mode — while leaving standard blue-link Search results untouched. The second is a full removal from AI search, which necessarily includes the first. MediaPost describes the two settings as linked rather than independent: a publisher cannot cleanly select 'exclude from AI features only' without it interacting with the broader AI-search removal setting. Google has not published granular documentation on exactly how that binding works — whether it is a hard technical dependency, a shared toggle, or something looser — and neither has MediaPost's sourcing. Treat that ambiguity as real. Do not assume you can quietly dial back AI Overviews exposure while keeping everything else about your AI-search presence exactly as it was, until Google documents the mechanism more precisely.

The practical read: tier one is a scalpel, tier two is a blunter instrument Google has not fully mapped out for publishers yet. If you only want to opt out of AI Overviews and AI Mode specifically, tier one is the theoretically correct choice. But because MediaPost's reporting flags the two settings as linked, verify with your own account team or a controlled test before assuming you are getting a clean, isolated exclusion. Discovering after the fact that your broader AI-search presence moved too is the expensive way to learn this.

There is also a sequencing question most publishers skip past. Before you touch either setting, establish a baseline. Pull your current AI Overviews and AI Mode citation rate, note which pages get pulled into summaries most often, and log what happens to referral traffic and branded search around those pages over the following month. Without that baseline, you cannot tell whether flipping tier one actually changed anything, or whether normal ranking volatility did. Publishers who skip this step tend to make the opt-out call on instinct — usually irritation at being scraped — rather than on what the setting is actually costing or saving them. An audit of where your content shows up across AI surfaces today is the cheapest insurance against reversing a decision six months in because the data was never collected in the first place.

Why this decision matters more than it looks

Here is what makes this a real decision rather than a checkbox. AI Overviews has more than 2.5 billion monthly active users. AI Mode has crossed 1 billion. That is not a niche surface you can treat as an afterthought — for a growing share of searchers, the AI Overview or the AI Mode conversation is the search experience, and opting out of it means opting out of how a large slice of Google's user base first encounters your brand. Matt Gilbert, CEO of Partnerize, has been arguing that publishers are already under-compensated for the influence they exert on AI-mediated buying decisions. Partnerize's HaloIndex research found publishers get 3.84x more AI-driven influence on purchase decisions than click-attribution models currently credit them for. In luxury fashion specifically, that multiplier climbs to 10.94x. For ecommerce brands especially, that gap is where the opt-out argument gets won or lost.

That gap is the real stakes of this decision. Opt out of AI features entirely and you do not just lose traffic you can already see in a dashboard — you lose a form of influence Partnerize's research says is not showing up in your attribution model at all. For any publisher whose products get discovered, compared, or recommended inside an AI answer before a click ever happens, that is revenue exposure most reporting setups are not built to catch, which is a separate problem from the opt-out decision itself but a closely related one worth solving in parallel.

Gilbert's argument, in plain terms, is that the industry is still measuring an AI-mediated web with a click-attribution ruler. A shopper who sees your product compared favorably inside an AI Overview, then buys two days later after a branded search, shows up in most dashboards as an unattributed direct visit — not as a result of your content doing the work. Partnerize built HaloIndex specifically to close that gap, and the 3.84x figure is their estimate of how much credit publishers are missing as a result. Whether or not you trust a single vendor's multiplier down to the decimal, the direction of the finding lines up with what anyone running attribution work across organic and AI channels has been seeing anecdotally for the past year: influence is real, it is happening earlier in the funnel than a click, and most reporting stacks are blind to it. That is exactly why walking away from AI visibility without first measuring it is the riskier of the two mistakes available here.

The case for opting out anyway

None of that means staying in is automatically correct. There is a real argument for taking the exclusion, and it starts with what AI Overviews has already done to click-through rates. Ahrefs analyzed roughly 300,000 keywords through Google Search Console — Ryan Law and Xibeijia Guan, February 2026 — and found that pages ranking #1 see a 58% lower click-through rate when an AI Overview appears on the results page, up from a 34.5% reduction Ahrefs measured in an earlier study from April 2025. The direction of that trend is the point: AI Overviews are suppressing organic clicks more over time, not less.

Sit with that for a second. A #1 ranking used to be the single most valuable position in Search. Ahrefs' data says that position is worth roughly half of what it used to be worth in click terms the moment an AI Overview appears next to it, and the trend line between the April 2025 study and the February 2026 study moved in the wrong direction for publishers in under a year. If that keeps compounding, the argument for opting out on principle gets stronger every quarter, independent of whatever citation value you think you are getting in return — because the clicks you would be protecting by staying in are shrinking regardless of your opt-out choice.

That reframes the whole decision. A publisher weighing tier one against tier two is not choosing between keeping clicks and losing clicks — Ahrefs' numbers say the clicks at the top of the page are already substantially gone once an AI Overview shows up. The real choice is whether your content is earning you anything for appearing in the summary that replaced those clicks. If Google's AI Overview cites you and that citation drives brand recall, comparison consideration, or a later branded search, tier two throws that away for nothing in return. If it does not — if you are purely getting summarized with no downstream benefit you can point to — tier two stops being a loss and starts being a reasonable trim. Understanding how citations actually form inside an AI Overview is worth doing before you flip either switch, especially since citations that used to anchor the top of an AI answer are increasingly showing up lower in the response — which changes how much a citation is actually worth to begin with.

The decision framework: who should pick which

Reduce it to one question: can you prove, with your own data, that AI-mediated visibility is producing something — brand-search lift, referral traffic, sales influence — that offsets what tier two would remove? If you can, stay in and instrument the measurement instead of opting out blind. If you cannot, and you have no realistic path to building that proof in the next two quarters, tier two is a defensible trim, not a mistake. Every publisher running this calculation lands in one of four positions.

Whichever bucket you land in, the fix is not a settings toggle — it is measurement most publishers do not currently have. Few teams can answer whether an AI Overview citation is worth anything, because their reporting stack was built for click attribution, not AI-mediated influence. That is the gap Partnerize's research is pointing at, and it is what turns this opt-out decision into a guess instead of a call. Before touching either tier, get a generative engine optimization program in place that tracks how your content performs inside AI Overviews and AI Mode specifically, and pair it with reporting and analytics work that ties AI-driven visibility back to pipeline instead of leaving it invisible. Do that first. Then decide which tier, if any, you actually need.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.