For most brands, the answer to "should you opt out of AI Overviews" is no, and the first public test of Google's opt-out switch makes a decent case for staying in. Cyrus Shepard turned the switch on for one site, his AI impressions went to zero right away, and his regular organic clicks dropped along with them, which isn't what you'd expect from the way Google describes the control.
It's a small test on one site, so I'm not going to pretend it settles anything. It's the first real data I've seen on a question marketing teams have been going back and forth on since June, and it lines up with a couple of other things we've learned this year about what happens when you pull your content out of AI systems.
The short version
Google's Search Console opt-out takes your site out of AI Overviews, AI Mode and the AI summaries in Discover. Google says it won't affect your regular rankings.
In the first public test, AI impressions had an estimated CTR of about 0.9%, and organic clicks still went down while the site was opted out.
Among major publishers blocking Claude's search crawler, 81% got almost no Claude citations, per Muck Rack.
For a brand, staying in and fixing what the AI says about you is the better bet. Publishers who sell the click have a harder call to make.
What Google's opt-out switch does
Google announced the control on June 3. It came out of the UK, where the Competition and Markets Authority required Google to give sites a way out of its AI features (Digiday). It started with a subset of UK site owners, and Google updated its post on August 31 to say the controls are now available to all websites worldwide (Google). It's a toggle in Search Console, and when you flip it your content stops being shown as a link or used to ground answers in AI Overviews, AI Mode and AI Overviews in Discover.
Google's wording on the tradeoff is pretty blunt. "Sites that opt out will not receive traffic or impressions from our generative AI features," the post says, and the control "will not be used as a ranking signal for search results outside of these generative AI Search features" (Google). So on paper you give up the AI traffic and keep everything else.
Google's Search Console AI control
What the opt-out switch turns off
AI Overviews
Your pages can be cited as links and used to ground the answer.
Can cite youAI Mode
Your pages can show up as sources in the conversational answer.
Can cite youAI Overviews in Discover
Your content can feed the AI summaries in Discover.
Can cite youRegular blue links
Google says the control isn't a ranking signal outside its AI features.
Ranking unchangedSources: Google, June 3, 2026 (updated Aug. 31); Cyrus Shepard's test via Search Engine Roundtable, Oct. 30, 2026.
The publishers Digiday talked to in June didn't love it. Paul Bannister at Raptive called it "nonsensical," and Phil Andraos at Reuters said, "There's not enough data" (Digiday). Their complaint was that the new AI performance reports in Search Console show impressions and no clicks (Search Engine Roundtable), so anyone deciding whether to leave would be deciding blind. That's still mostly true, which is a big part of why Shepard's test got passed around so fast last week.
What happened when Cyrus Shepard opted out
Shepard used the control to pull one site out of AI Overviews and AI Mode, then compared impressions, clicks and CTR before, during and after the opt-out (Search Engine Roundtable). He called it a small case study, and Barry Schwartz said the same when he wrote it up. AI Overviews kept showing up on the results pages for those searches; Shepard's site just stopped being cited in them.
AI impressions dropped to zero as soon as the opt-out kicked in, which is what you'd expect, and they'd been about 8% of the site's total impressions before that. By backing the known AI impressions out of the overall numbers, Shepard estimated those AI impressions had a CTR of roughly 0.9%. The part nobody was expecting was that regular organic clicks dropped too while the AI features were off, and on the URLs with the most AI visibility, average position went down during the test and came back when AI visibility returned.
Shepard's explanation is something he calls a SERP halo effect, where showing up in more places on the page makes a click more likely anywhere on it. Greg Sterling added that Near Media's research has seen AI Overviews lift later brand searches and clicks, though not consistently.
I'd treat the halo idea as a hypothesis for now. It's one site and one test window, and Shepard himself pointed out that a couple of new posts and some brand search were moving his sitewide numbers at the same time. Still, the 0.9% figure sits right next to what Pew found in 2025, when users clicked a link inside the AI summary on about 1% of visits (Search Engine Roundtable). The citation inside the AI Overview is a lousy click source on its own, and Shepard's test hints that it might still be doing something for the rest of your listing.
What the 2025 and 2026 data says
What happens when you pull out of AI
Why clicks alone won't answer it for a brand
If you only look at clicks, AI search looks rough for anybody who publishes on the web. A preregistered field experiment from researchers at Penn and Northeastern, with 1,100 real Google users, found that hiding AI features raised the external click-through rate by 8.8 percentage points, and sending every search into AI Mode cut it by 18.8 points (arXiv, numbers via TechWyse). Pew found people clicked a traditional result on 8% of visits that had an AI summary, against 15% when there wasn't one. I went through what that does to traffic in The Click That Didn't Happen, and I don't see any of it reversing.
A brand is in a different spot than a publisher, though. A recipe site makes its money on the pageview, while most brands make theirs when somebody buys the thing, and the AI answer is often where a buyer forms an opinion about that thing before clicking anything. In my AI Verdict Study, about four in five classifiable citations pointed to sites other than the brand's own, and on the first unbranded question a brand's own site supplied 4% of citations (Where AI Engines Go to Decide About Your Brand). My read is that opting out takes your own pages out of the set Google can quote while the review sites, Reddit threads and competitor comparison pages stay in, so Google keeps describing you with your side of the story missing.
That matters most on searches for your own name. I wrote about what AI Overviews tell people who look you up, and those are the answers I'd least want to leave to everybody else's pages.
What happens when you block the crawler instead
The Search Console switch only covers Google. For ChatGPT, Claude and Perplexity, the lever is robots.txt, and Muck Rack's new Generative Pulse report has a number on what pulling that lever does. "81% of major publishers blocking Claude's search crawler received almost no Claude citations," compared with 12% of the publishers that allowed it (Muck Rack via GlobeNewswire), which is a big gap to come from a few lines in a robots.txt file.
The same report found journalism's share of AI citations fell from 27% in May to 22%, and owned media rose from 13.7% to 15.2%. ChatGPT's journalism share went from 28.1% in January to 22% in September. My guess is some of that is publishers walking away on purpose and some is the engines changing what they like, and I can't tell you the split. Either way, somebody's content fills the space, and brand-owned pages are part of what's filling it.
I covered the crawler side in more detail in Should You Let AI Read Your Site for Free?, including which bots are for training and which ones decide whether you get cited. The advice there still holds; judge each bot by its job, and keep the search bots in.
Who should think about opting out
There are a few cases where I'd at least run the test. Number one, you're a publisher whose content is the product and the AI answer gives the whole thing away, like a recipe, a how-to with one answer, or a single statistic. If people get what they came for in the Overview and Shepard's halo doesn't show up for you, the math might favor leaving. Two, you're licensing or charging for content, and keeping it out of AI answers protects that deal. Then last, you have a legal or compliance reason to control where your exact wording shows up.
For a brand selling a product or a service, I'd urge you not to flip the switch. If you're tempted anyway, run it the way Shepard did, with a clean before period, a set test window, and somebody watching organic clicks and branded search along with AI impressions. His test had an after period where his AI visibility came back, so it isn't a one-way door, though I don't know how fast it comes back for a bigger site.
Six-question self-check
Should you opt out?
Answer for your own site. Nothing you click here leaves your browser.
What I'd do with this
Number one, open the AI performance report in Search Console and look at your AI impressions as a share of the total. Shepard's were about 8%. If yours are a lot higher, you have more riding on this either way.
Two, before anybody on your team talks about opting out, find out what the AI answers say about you. Run your top buyer questions in AI Mode, ChatGPT, Gemini and Perplexity a few times each and write down which sites get cited. The free Quick Check walks you through that with three buyer prompts and nine checks.
Three, check robots.txt for the search crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) and make sure nobody blocked them by accident while going after the training bots.
Then last, if your own pages aren't getting cited, fix the pages before you blame the feature. I wrote about the click side of that in How to Get Clicks From AI Overviews and AI Mode, and most of my book, Explainable, is about why the engines pick some brands' sources and pass over others.
If you've flipped the switch, or tested it on part of a site, I'd love to see what you found, especially if it doesn't match Shepard's. There isn't much data on this yet, and every honest test that gets shared makes the call a little easier for the rest of us.
Jarred Smith is the author of Explainable: Why AI Recommends Some Brands & Ignores Others, an Amazon bestseller on AEO, GEO, and SEO. He's a marketing leader with nearly 20 years of experience across healthcare, public media, retail, and environmental services. Find him at jarredsmith.com.