Google has activated AI performance reports in Search Console for all websites worldwide. As of August 31, 2026, any site owner can see how their content appears in AI Overviews and AI Mode.
Sounds great. And it genuinely is an important step. But once you open the reports and look at what data you actually get, the picture shifts.
We spent the last few days analyzing the new reports for the properties we manage. Here is what we found, what is conspicuously absent, and what you should do right now.
What the new AI reports in Search Console actually show
The AI reports in GSC give you something you never had before: impression data from Google's AI features. These are separate from classic search impressions and cover two sources: AI Overviews (the generated answers that appear directly in search results) and AI Mode (the conversational search experience launched this year).
The data breaks down across three dimensions: page, country, and date. You can see which pages on your site appear most often in AI-generated responses, where those impressions come from geographically, and how they trend over time. For the first time, you have official visibility into how Google's AI treats your content.
Until now, AI visibility was a completely opaque zone. No official data existed. You could manually monitor a few queries, use third-party tools with variable accuracy, or simply guess. Now you have at least a foundation of real data to build something concrete on.
A recent article on Search Engine Journal confirms the global rollout and highlights an important detail: while Google announced the complete deployment, some properties with insufficient AI impressions may not see the reports yet. So do not panic if they are not immediately visible.
The missing data changes the entire calculation
Here is the real problem: the reports contain no click data.
You know Google's AI displayed your page. You do not know whether anyone clicked through from that AI response to your site. It is like having a digital billboard at a busy intersection. You count how many cars pass by, but you have no idea how many drivers stopped, read the message, and walked into your store.
Impressions alone do not tell you whether your AI presence generates actual traffic or just checks a box in a report.
For us as an agency, this creates a difficult conversation with clients. The inevitable question is: "OK, we are visible in AI. But does it drive conversions?" And the honest answer, even with these new reports, remains: we cannot know precisely. We can see presence, but not impact.
This is a temporary limitation. The UK's Competition and Markets Authority (CMA) explicitly required click data and imposed a deadline: March 2027. So clicks will eventually appear in Search Console. But until then, the AI reports are a half-open window: you see something, but not everything that matters.
The opt-out control: why you probably should not use it
Alongside the reports, Google activated a global opt-out control. Sites can choose to not appear in Search AI features, with no ranking penalties in classic results.
Many people's first reaction: "Great, I will stop AI from stealing my traffic." But that logic does not hold long-term.
AI Overviews are not going away. If you opt out, your competitors will fill that space. And as users grow accustomed to generated answers, being present in AI responses becomes an authority signal, not just a traffic channel.
We recently discussed how marketing teams need to restructure for AI search. Opting out is the opposite of restructuring. It is the equivalent of blocking search engines in 2005 because "they copy our content."
There are clear exceptions. If you produce premium original content (proprietary research, studies, unique datasets) that AI reproduces in full without attribution, opt-out can be a temporary tactical decision. But for 95% of sites, giving up AI visibility is not a strategy. It is a retreat.
A relevant technical detail: the current control works at the site level. By March 2027, Google will need to implement page-level controls. When that becomes available, you will be able to protect pages with unique value while keeping the rest of your site visible in AI. That is where this feature becomes genuinely useful, because real-world situations are rarely all-or-nothing.
What you can do with these reports right now
We do not have click data, but we have impressions. And impressions, used correctly, yield useful insights. Here is what we recommend to our clients and what we have already started implementing:
Compare AI impressions with classic impressions. If a page has 10,000 classic impressions and 200 AI impressions, Google does not consider that content relevant for generated answers. If the ratio is reversed, you have a clear signal: the engine prefers that content for AI. Both scenarios tell you something actionable about how to optimize your pages.
Set a baseline now. Today's AI impressions have no historical equivalent, so there is nothing to compare them against yet. But next month, you will be able to see whether your AI visibility is growing or shrinking. Export the data, save it in a spreadsheet, create the reference point you will measure from.
Identify discrepancies. Pages that appear in AI but do not rank well in classic search (and vice versa) reveal what type of content the generative engine prefers versus the traditional algorithm. We wrote about how Google is building tools directly into search, and these discrepancies confirm: AI evaluates content through a different lens than the classic algorithm.
Do not make strategic decisions on impressions alone. AI impressions are a visibility indicator, not a performance indicator. It is the difference between being seen and being chosen. Until we have click data, resist the temptation to restructure your SEO strategy based solely on AI impressions.
The AI reports in Search Console are a new window. It is not a large window and it does not show the full landscape, but two months ago you did not have it at all. Monitor, compare, build your baseline. And when Google adds the data that truly matters (likely in the first half of 2027), you will be ready to act on a complete picture.
