GSC AI Reports: Should You Block Google AI Responses?

Google's new Search Console AI performance report and the opt-out toggle that shipped beside it on June 3, 2026 hand publishers a control they have wanted for two years: a way to remove content from AI Overviews and AI Mode without losing a single position in regular Search. The question is no longer whether you can block Google AI responses. It is whether you should.

For most brands the answer is no — and the reasons are concrete, not ideological. But a defensible minority of publishers now have a clean mechanism to act on a decision that, until this week, came bundled with an organic-traffic penalty. The trade-offs finally separate.

This guide maps exactly what launched, untangles the three different AI controls that everyone keeps confusing, and gives you a decision matrix built on business model rather than blanket advice. It is the reporting-and-opt-out companion to our deeper look at crawler-level access controls like robots.txt and llms.txt — those govern what bots can fetch; this governs where your fetched content is allowed to appear.

On June 3, 2026, Google launched a dedicated Generative AI performance report inside Search Console, initially limited to a subset of UK website owners before a wider global rollout. The same release added an opt-out toggle that, when enabled, stops a site from appearing in AI Overviews, AI Mode, and AI Overviews within Discover — while leaving regular Google Search results and the Discover feed untouched.

Both changes trace back to a regulatory mandate. The UK's Competition and Markets Authority required these controls under the Digital Markets, Competition and Consumers Act 2024, giving Google roughly nine months — a deadline near March 2027 — to implement the full set of requirements, with "important parts" expected sooner. The CMA described the package as a world-first: the first binding obligation forcing a search engine to give publishers genuine opt-out control over AI-feature inclusion, distinct from blocking AI model training. This is the regulatory backdrop we covered when the UK CMA's intervention on AI Overviews opt-out first surfaced.

The new report tracks impressions and nothing else. You see impressions broken down by pages, countries, devices, and dates, at hourly through monthly granularity. There is no click data, no CTR, no average position, and no query-level breakdown. It covers AI Overviews and AI Mode specifically — Discover's AI features sit in a separate report, and Search Labs experiment data is excluded entirely.

That single design choice defines what the data can and cannot tell you. Impressions confirm where your content is being pulled into AI surfaces and at what relative volume across geographies and devices. They cannot tell you whether anyone clicked, what they searched, or how AI Overviews and AI Mode compare against each other. For click-through behavior and the traffic consequences of AI answers, you still need server logs and third-party tooling — the same gap we walk through in our guide to tracking AI Mode traffic in Google Search Console.

The right way to read impressions-only data is as a resonance signal, not a traffic metric. A page that earns heavy AI impressions is content the model finds worth citing. Cross-reference that against the organic clicks the same URL earns in the standard Performance report, and you get a genuinely useful picture: the pages working hardest in both worlds are your non-commodity assets, the ones least replaceable by a generated summary.

Most of the confusion around "blocking AI" comes from collapsing three separate controls into one. They are not interchangeable, and choosing the wrong one either fails to do what you intended or quietly costs you organic traffic. Here is the clean version.

Google-Extendedis a robots.txt token that blocks your content from training Google's AI models — but it does not block AI Overviews or AI Mode. Those features draw on content through the regular search index via Googlebot, not through Google-Extended crawls. Blocking Google-Extended is therefore insufficient to remove you from AI Overviews, a distinction that trips up most publishers.

nosnippet (and data-nosnippet / max-snippet) does remove content from AI Overviews and AI Mode — but it simultaneously strips your snippet from traditional search results, which carries a real organic CTR cost. A documented case study found nosnippet removes AI Overview link cards within hours to days; it also removes the very snippet that drives clicks on your standard listing. Google has confirmed these preview controls remain the primary technical mechanism for snippet control, but the organic tradeoff is the reason they were never a clean answer.

The new GSC opt-out toggle is the first control that blocks AI features with no organic-snippet penalty. It closes the gap the other two left open: AI features off, organic ranking and snippet intact. The table below is the comparison no single source had published cleanly before this week.

The reason the opt-out is finally a clean decision is that the penalty risk is gone — and not as a matter of Google's goodwill. Google confirmed the opt-out setting will not be used as a ranking signal for regular Search results. The CMA conduct requirement goes further: it explicitly bars Google from ranking opted-out sites lower or reducing their visibility because they exercised the control. The enforcement teeth are the CMA penalty regime, which can reach up to 10% of global annual turnover.

That said, opting out is not free of consequence — it simply moves the cost from rankings to reach. Flip the toggle and you forgo all AI impressions and any traffic AI surfaces would have sent. With AI Overviews reported by Google at over 2.5 billion monthly active users and AI Mode exceeding 1 billion monthly users, that is a large surface to step away from. Those user figures are vendor-stated; treat them as Google's own framing rather than independently audited counts.

The market reflects that ambivalence. A Search Engine Land poll of more than 350 respondents found 41.9% will not block Google from using their content for AI features, against 33.2% who plan to, with 24.9% still undecided. The plurality choosing to stay in is the tell: for most operators the inclusion upside still outweighs the control. The publishers leaning toward opt-out are concentrated where per-reader value is high and AI-visit conversion is low.

Blanket advice is the failure mode here. The right decision for a subscription news publisher — high value per reader, low conversion from an AI-summarized visit — is the wrong decision for a B2B SaaS blog that benefits from brand recognition every time it is cited. Segment by business model, not by traffic volume. The choice cards below map the most common cases, and the proprietary matrix that follows turns them into a single reference.

Read the matrix as a starting position, not a verdict. The decisive variable is almost always the "what would flip it" column — the specific, measurable condition that overrides the default. A subscription publisher that discovers AI Overviews are actually seeding trial signups should stay in despite the default; a B2B blog being consistently miscredited to a competitor has a real reason to reconsider. This is the same decision-matrix approach to technical SEO we apply to indexation choices: default by pattern, override by evidence.

Lost in the coverage of Google's launch is a concrete competitive gap. Bing Webmaster Tools launched its AI Performance report globally in public preview on February 10, 2026 — roughly four months ahead of Google's UK-limited rollout. And Bing's report is materially more granular: it surfaces citations, grounding queries, page-level citation activity, and citation trends, where Google's first release gives you impressions and nothing more.

The practical takeaway for operators is not to abandon Google — its AI surfaces are far larger — but to treat Bing Webmaster Tools as a free, complementary lens that already answers questions Google's report cannot. If you want to know which queries are grounding AI answers in your content, Bing tells you today. Google may add metrics over time, but the gap is real right now.

There is a forward-looking read here too. Google publicly signaled it intends to add metrics to the report over time, which means today's impressions-only view is a floor, not a ceiling. The operators who win the next twelve months are the ones building their measurement stack now — combining the GSC impressions signal, Bing's citation data, and server-side click tracking — rather than waiting for a single tool to become complete. It will not.

Here is the context absent from nearly every other write-up. The opt-out decision is landing in your lap during an active data blackout. Google's May 2026 core update — its second of the year — started May 21 and completed on June 2, 2026, the day before the GSC AI reports launched, after roughly twelve volatile days. Google itself recommends waiting until at least June 9, 2026 before drawing conclusions from post-update GSC data.

Now overlay the report's start date. The AI performance data appears to begin May 18, 2026 — which means every data point you currently have sits insidethe volatile core-update window. Deciding to opt out based on two weeks of update-distorted impressions is premature by Google's own guidance. If your AI impressions look alarming right now, that may be the core update talking, not a structural shift. For the recovery playbook on that specific update, see our breakdown of Google's May 2026 core update, which completed June 2.

For the vast majority of brands, the correct first move is to do nothing irreversible and start measuring. Treat the new report as a monitoring asset, not a trigger to opt out. Here is the order of operations we recommend to clients before June 17.

If, after that sequence, you genuinely belong in the opt-out minority, the toggle is a single setting and it is reversible — there is no structural cost to monitoring first and acting later. The one move we would actively caution against is reaching for nosnippet as a shortcut: it kills your organic snippet alongside the AI feature, and the new toggle now does the AI half without that collateral damage. Sequencing this decision correctly is exactly the kind of work our agentic SEO engagements are built around — and the traffic-impact side of it is covered in our analysis of CTR declines caused by AI Overviews.