Discussion starter: an AI search question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- AI Search
- Search Console
Search Console’s AI features opt-out reached every site on 31 August 2026, and Google says rankings won’t be affected. What evidence would you need to use it?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
I would need two kinds of evidence before using it, and neither is about rankings. First, evidence that your pages’ presence in AI Overviews, AI Mode and AI Overviews in Discover delivers less value (clicks, Discover traffic, brand exposure) than what opting out protects, such as visits the summaries currently absorb or a licensing position. Second, proof from a controlled comparison that traditional clicks hold up once you are out. Google’s promise that the control won’t affect rankings answers the smaller question: rankings can hold steady while traffic changes, because opting out removes your content from the AI answer, not the AI answer from the results page.
Access is no longer the obstacle, since every site now has the toggle. Measurement is. Search Console shows AI impressions but not AI clicks, so most sites can’t yet produce the first kind of evidence, and because the switch is site-level, a single site can’t split-test it. The practical position is to build the baseline before touching it, then treat any opt-out as an experiment with a stop rule agreed in advance, not a one-way decision made on the strength of the ranking statement.
What the control does, and what it doesn’t cover
The documented facts are limited. The Search generative AI control is a site-level toggle in Search Console that lets site owners “decide if they want their site to appear in and help ground responses in our generative AI Search features”, meaning AI Overviews, AI Mode and AI Overviews in Discover. Google first tested it with a subset of UK website owners, after a UK Competition and Markets Authority conduct requirement, and on 31 August 2026 it rolled out to all websites worldwide. Google says sites that opt out “will not receive traffic or impressions from our generative AI features”, and that the control “will not be used as a ranking signal for search results outside of these generative AI Search features”.
The UK requirement behind it adds a timeline. PPC Land reports that the substantive obligations come into legal force on 3 December 2026 and that page-level grounding controls, letting publishers block specific URLs rather than whole domains, follow on 3 March 2027. Those are UK regulatory dates, and Google’s worldwide announcement gives no date for page-level controls elsewhere. For now the control is all or nothing for a site. That matters: you can’t keep AI exposure for analysis pieces while withdrawing recipes or explainers.
Control What it affects Cost Search Console AI opt-out (site-level, all sites since 31 August 2026) AI Overviews, AI Mode and AI Overviews in Discover Loss of AI citations and any clicks or exposure they produce nosnippet,data-nosnippet,max-snippetGoogle’s existing way to limit use in AI features, shared with regular snippets Weaker or missing snippets in normal results Google-Extended Use for Gemini model training, not Search AI features None in Search GPTBot, OAI-SearchBot and other bot controls OpenAI and other platforms, which the Google toggle does not touch Depends on the platform Why “rankings won’t change” is not the evidence you need
Picture a newsagent’s front counter. Opting out takes your magazine off the counter display, but the display stays, filled with other titles. Your copy is still on the shelf in the same spot. Whether sales change depends on how many buyers ever walked past the counter to the shelf.
Technically: Google hasn’t said AI answers stop appearing when one source opts out. On most queries the likely result is the same AI Overview built from other sources, sitting above your unchanged blue link. So the real question is whether being cited in the AI answer produces clicks, brand recall or Discover traffic that being uncited below it doesn’t. That is a traffic and business question, and Google’s statement is silent on it.
Three indirect effects sit outside the ranking promise:
- Visibility that doesn’t depend on rankings. Ahrefs reported the share of AI Overview citations that also rank in the top 10 falling from 76% to 38% between July 2025 and March 2026, though a changed detection method means the two figures aren’t directly comparable. If some of your AI exposure comes from pages that rank outside the top 10, opting out removes visibility your blue links won’t replace.
- Discover. The toggle covers AI Overviews in Discover. For news publishers that rely on Discover, that effect needs measuring separately from web search.
- Brand effects. Google has expanded preferred sources to AI Mode and AI Overviews and says users are about twice as likely to click a preferred source. Because opted-out sites get no impressions from those AI features, a reader’s preference presumably stops surfacing you there too. Google hasn’t spelled out the interaction, so it is worth asking before you switch.
Publisher reaction has reflected this. During the UK test, Digiday’s headline framed it as a choice publishers can’t safely use, because the downside is measurable only after the fact.
The evidence I would want before recommending it
Question Evidence Where it comes from Pushes towards opting out if How exposed are we? AI impressions by page, country and device Search Console generative AI performance reports AI impressions concentrate on pages whose clicks have fallen Do AI appearances produce clicks? CTR trend on pages with high versus low AI exposure, for comparable queries Search Console page and query data, grouped by AI exposure High-exposure pages lose CTR faster than comparable low-exposure pages Is the content substituted or complemented? Content type of the most-cited pages Editorial review of top AI-impression pages Answers are fully contained in the summary (facts, definitions, quick how-tos) Does the toggle behave as promised? Blue-link positions and clicks after opt-out versus a control Controlled test, third-party rank tracking Positions and total clicks hold while AI impressions fall to zero Is it reversible? Time to reappear in AI features after opting back in Your own test (Google’s announcement gives no timing) Re-entry is fast and complete What is the commercial case? Licensing position, subscription conversion, revenue per session Commercial and finance teams Withholding content supports a licensing or subscription strategy The measurement gap is the hard part. Google’s generative AI performance reports show impressions only, with no clicks, CTR or queries, for AI Overviews, AI Mode and AI features in Discover. They rolled out to a subset of sites, UK first, in June 2026 and reached all eligible sites on 31 August 2026. AI Mode clicks are folded into Web totals in Search Console, and practitioners report they arrive in GA4 as google / organic. So you can see how often AI features show your pages, but not what those appearances are worth.
Watch for this:
Search Console’s Web totals have included AI Mode data since June 2025. If you opt out, Web impressions may drop simply because AI Mode impressions stop counting, not because rankings fell. Google doesn’t document how to subtract generative AI impressions from Web totals, so judge ranking neutrality with a fixed query set in a third-party rank tracker and with clicks, not raw impressions or average position.
How to test a site-level switch
The cleanest design uses more than one property. A publishing group with several comparable sites can opt out one and hold the others as controls. A single-site publisher can only compare before and after, which is much weaker because news cycles, seasonality and core updates all move traffic.
- Hypothesis: opting out cuts AI impressions to near zero without reducing blue-link positions, and changes total organic clicks by less than the site’s normal week-to-week range.
- Baseline: at least eight weeks of AI impressions, Web clicks, Discover clicks, branded query clicks and revenue per session, with the release date annotated in Search Console.
- Primary metric: total organic clicks (Web plus Discover), compared against the control property.
- Secondary metrics: positions for a fixed query set, branded search clicks, direct traffic, subscription starts.
- Segmentation: separate pages that had high AI exposure from those that had little. Almost all of the effect should sit in the first group.
- Confounders: core updates, major news events, seasonal peaks, and site changes during the test.
What would change the recommendation: if positions hold and total clicks on high-exposure pages hold or rise, the AI citations were worth little and opting out is a low-cost choice made on commercial grounds. If total clicks fall beyond normal variation while positions hold, the citations were delivering value that the ranking promise didn’t cover, and you should opt back in and record how long re-entry takes. If Discover drops sharply, that result alone may outweigh everything else for a news publisher.
Where I would land by site type
For brands, ecommerce, SaaS and local businesses, AI citations are exposure you would otherwise pay for, and I can’t see a traffic case for opting out on current evidence. For publishers whose content is mostly short factual answers, where AI summaries substitute for the visit, and who are negotiating licensing, the case is stronger, but it rests on commercial strategy as much as SEO data.
My recommendation for now: don’t opt out on the strength of the ranking statement. Start segmenting pages by AI impressions using the new report, track the CTR of high-exposure pages against comparable low-exposure pages, and decide in advance what result would justify switching. If that evidence points towards opting out, test it on the smallest comparable property first, and only extend it if total clicks, Discover and revenue hold for a full measurement period.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.