In August 2025, Google’s own John Mueller warned the SEO industry about something specific. Speaking about the growing pile of new acronyms, GEO, AIO, AEO, he warned that pushing new terminology too aggressively was often a signal of low quality tactics rather than genuine innovation. A month later, in September 2025, when a fake screenshot claimed Google had built a dedicated AI Overview filter inside Search Console, Mueller shut that down too, calling the screenshot fake and confirming no such feature existed.
Eleven months later, on 31 August 2026, Google built exactly that feature, for real, and rolled it out to every website on earth.
I want to walk through what actually changed, why the timing and the backstory matter more than the feature itself, and exactly how to check this on your own site.
The rumour that turned out to be a preview of the future
The story is worth telling properly, because it says something about how much pent up demand existed for this.
On 15 September 2025, a user holding Platinum Product Expert status in Google’s own Search Console help forums, Adrian Coutin, posted that Google had integrated data for AI-generated search features directly into Search Console’s Performance reports. The post included a fabricated screenshot showing a new Search Appearance filter specifically for AI Overview data. It spread fast, because it was exactly what site owners had been asking for.
Mueller responded on LinkedIn within hours. “There’s no filter like this in Search Console, and the screenshot itself is fake,” he wrote. Coutin deleted his original post and apologised for the error.
For the following year, Google’s official position stayed consistent, AI Overview and AI Mode data was folded into the standard performance numbers, with no way to separate it out, and no committed timeline to change that. Multiple SEO guides published through late 2025 and into early 2026 describe workarounds involving third party tools precisely because Google offered nothing native.
Then, in June 2026, Google quietly began testing a real version of exactly the feature Coutin had faked, on a limited subset of properties. On 31 August, that testing phase ended, and the feature went live everywhere.
What actually shipped
Google confirmed the global rollout on its own Search Central blog, describing the new reports as dedicated views of impressions specifically within generative AI features on Search, covering AI Overviews and AI Mode, as well as the generative AI features that now appear inside Discover.
Search Engine Journal’s coverage of the global launch adds a detail worth knowing before you go looking for it. Not every property has access yet, and sites with a low volume of AI impressions may not show meaningful data in the report even where it is available. If you check and see very little, that is not necessarily a sign the report is broken. It may simply reflect where your content currently sits in AI systems.
The report breaks impressions down by page, by country, and by date. What it does not include, and this is the detail every source I checked agrees on without exception, is clicks, click through rate, or the specific queries that triggered an appearance. You can see that a page from your site showed up inside an AI generated answer. You cannot see what someone typed to get that answer, and you cannot see whether they clicked through to you afterward.
The report is best understood as a record of where you showed up, not a measure of what that appearance actually earned you in traffic or business.
Google also shipped an opt out control alongside the reporting, letting site owners remove themselves from these AI features entirely. Google has stated that using this toggle does not affect a site’s ranking anywhere else in standard Search results, which matters, because it means the choice to opt out is a genuine preference rather than a trade-off against traditional visibility.
Why the backstory matters more than the button
I keep returning to the Mueller timeline because it tells you something the feature announcement alone does not.
In August 2025, Google’s own search advocate was publicly wary of the entire AI visibility conversation, treating the proliferation of GEO and AEO terminology as a likely signal of low quality hype rather than a legitimate new discipline. A month later, Google was actively denying it had built any tooling to support that conversation.
Less than a year on, Google built the tooling anyway, at a global scale, unprompted by any specific competitive pressure that forced its hand. That sequence is a stronger signal than the launch blog post itself. It suggests the internal case for treating AI visibility as a distinct, measurable category became undeniable even to a company that was publicly skeptical of the framing not long before.
For anyone running SEO or GEO work for a client, that timeline is useful ammunition. The scepticism you may have encountered from more traditional stakeholders, the idea that AI visibility is a marketing buzzword rather than something worth the budget, gets harder to sustain once the platform itself has built a dedicated reporting surface for exactly that.
The data that explains why Google finally built this
Two independent data points, unrelated to this launch, explain the pressure Google was likely responding to.
Semrush’s AI Visibility Index, tracking brand mentions across Google’s AI Mode over several thousand prompts, found a 4 percent drop in mention share between August and October last year, while the diversity of sources being cited across the same period rose 13 percent. Visibility was getting harder to hold onto and more competitive at the same time, with no native way for an affected business to even confirm it was happening.
BrightEdge’s industry wide data adds the commercial pressure point. Search impressions are up 49 percent year over year, while click through rate on those impressions is down 30 percent over the same period. Businesses have been watching their visibility numbers climb while their actual traffic stalls or falls, with no first party data source able to explain why. That is precisely the gap this report is built to close, even if only partially.
How to actually find and use this on your own site
Coverage of exactly where this sits inside the Search Console interface is not fully consistent across sources I checked, some describe it as a dedicated left navigation item, others as a filter within the existing Performance report. Rather than send you hunting for one specific click path that may not match what you see, the reliable approach is this. Open your Performance report inside Search Console for the property in question, and look for a Generative AI or AI features option, either as a separate section or as a filter on your existing search appearance data. If it is not visible yet, that is consistent with Google’s own rollout note that access is expanding property by property rather than appearing identically for everyone on day one.
Once you find it, compare the AI impression volume against your total Web impressions for the same period. If AI generated impressions represent a meaningful share of your total visibility, and your click through rate has been falling without a ranking drop to explain it, that gap is very likely hiding inside exactly this data. This is the same pattern the BrightEdge numbers describe industry wide, positions can hold while clicks fall, because an increasing share of answers are being delivered before anyone reaches your listing.
If the number is small or close to zero, treat that as informative in a different way. It likely means your content is not currently being retrieved for the kinds of questions that trigger these AI features in your category, which is a different and in some ways more urgent problem than a visibility measurement gap.
What not to do with this data yet
Do not draw conclusions from a single week. This is newly global infrastructure, and Google’s own historical pattern with new reporting features is that data takes time to stabilise as more properties gain access and reporting pipelines mature.
The useful read is a trend over four to eight weeks, not a snapshot. Is your share of AI impressions climbing, flat, or falling relative to your total visibility over that window. That trajectory tells you far more than any single day’s number, particularly while the feature is still rolling out unevenly across different sites.
Frequently asked questions
1. Does this show me if ChatGPT or Perplexity cite my site?
Answer: No. This report is scoped entirely to Google’s own AI features, AI Overviews, AI Mode, and Discover’s generative surfaces. It has no visibility into any AI platform outside Google’s own products. Measuring citation inside ChatGPT, Claude, Gemini, or Perplexity still requires testing those platforms directly, since none of them currently offer a first party reporting tool of their own.
2. Does a high AI impression count mean I am getting more traffic?
Answer: Not necessarily, and this is the central limitation to keep in mind. The report measures impressions only. BrightEdge’s industry data shows click through rate falling even as impressions rise broadly, so a strong impression number does not guarantee a proportional lift in actual visits.
3. Can I see which search queries triggered my AI impressions?
Answer: No, and every source covering this launch agrees on that point. Query level detail is one of the more consistently requested features Google has not shipped, alongside click data.
4. Should I use the opt out toggle?
Answer: For most businesses, no. Google has confirmed opting out does not improve or protect your standing in standard Search results, so opting out only removes a visibility channel without gaining you anything in return. It exists mainly for edge cases such as publishers protecting content behind a specific business model, not for typical service or B2B businesses trying to be found.
5. Why did it take Google so long to build this, given the September 2025 rumour showed clear demand?
Answer: Google’s public position throughout that period was that AI Overview data belonged inside standard performance reporting rather than a separate view. The shift to building a dedicated report suggests the internal case changed, likely under the same pressure reflected in the Semrush and BrightEdge data, impressions climbing, clicks falling, and no way for site owners to see why.
If you want visibility that goes further than what Google’s own tools can show, including whether ChatGPT, Claude, Gemini, and Perplexity are actually citing you, that is exactly what we check in a free AI Visibility Audit.