Google is adding AI search surfaces and leaving publishers to guess what happens next

Google is expanding the places where AI-generated answers link to publishers. More surfaces, more potential traffic. That sounds like good news until you try to measure it.

Right now, Search Console does not break out clicks that come from AI Overviews separately from standard organic results. So if your traffic shifts, you cannot tell whether AI surfaces are helping or hurting. You are flying on instruments that have not been updated for the airspace you are in.

This is not a minor reporting gap. It is a structural information asymmetry. Google gets to see exactly how users interact with every new AI surface it ships. Publishers get an aggregate number and a shrug.

What Google is actually doing

AI Overviews now appear across a wider range of query types than they did at launch. Google has also been testing AI-organized search result pages, where the classic ten blue links give way to topic clusters arranged by its models. Inside those clusters, publisher links appear as citations rather than ranked results.

Citations can drive clicks. Early publisher reports suggest some do. But the click behavior for a citation in an AI summary is different from a click on a ranked result. Users who click a citation have already read a synthesized answer. They arrive at your page with less intent to read and more intent to verify or go deeper. Engagement metrics look different. Bounce rates look different. Session depth looks different.

If you are using those metrics to brief clients, justify content investment, or make editorial decisions, the old benchmarks no longer apply. The denominator has changed.

The data Google has, and what publishers get

Google knows which impressions came from AI Overviews. It knows which links inside those overviews were clicked, at what position, and by what user segment. It has this data because it built the surface.

Publishers currently see total clicks, total impressions, average position, and average CTR in Search Console. That is the same reporting structure that existed before AI Overviews launched. Nothing has been added to reflect the new surfaces.

The practical result. if your organic traffic drops ten percent this quarter, you cannot determine how much of that drop is attributable to AI Overviews absorbing queries that would previously have produced a click. You can hypothesize. You cannot confirm.

This matters beyond individual publishers. Agencies running SEO programs for clients are being asked to explain traffic changes they do not have the data to explain. The honest answer, currently, is that no one outside Google can give you a precise breakdown.

What an agency actually does with incomplete data

You work with what exists. That means building a measurement layer that does not depend entirely on Search Console.

First, server logs. Raw log analysis shows you every crawl and every request. It does not tell you about AI Overview appearances, but it does give you a ground-truth view of crawl behavior that Google's tools sometimes obscure.

Second, behavioral segmentation in your analytics. Tag your content by query intent category and content type. When AI Overviews appear for informational queries and your informational traffic drops while transactional traffic holds, that pattern is itself a signal. It is circumstantial, but it is directional.

Third, GEO as a parallel discipline. Generative engine optimization is still early, but the core idea is sound: write content that gives AI models something specific and citable, not content that merely ranks for broad terms. If AI surfaces are going to cite publishers, the ones that get cited will be the ones that made the model's job easier. Clear claims. Named authors. Structured context. Dated information.

None of this replaces the data Google is withholding. But it reduces your dependency on a single reporting surface that may never be complete.

The political dimension agencies should track

This is not purely a technical problem. It is also a regulatory one.

The EU's Digital Markets Act creates obligations around data access and self-preferencing for designated gatekeepers. Google is designated. The question of whether withholding AI surface attribution data constitutes a form of self-preferencing is live and unsettled. Several European publisher associations have already raised the topic formally.

In the US, the Department of Justice antitrust case against Google included extensive argument about search data access. A ruling that touches Google's obligations to share data with third parties could, in theory, reach Search Console reporting. That is a long chain of contingencies, but agencies advising clients on search strategy should be watching it.

The practical implication for now. document what you cannot measure. Keep a record of the reporting gaps your team has identified, and when they were identified. If the regulatory environment shifts and attribution data becomes available retroactively or by request, you will want that paper trail.

Where this goes

Google will eventually add some AI surface reporting to Search Console. The pressure from publishers, agencies, and regulators is too sustained to ignore indefinitely. The question is how granular that reporting will be, and whether it will arrive before the current AI surfaces have already reshaped baseline traffic expectations.

For now, the most useful thing an agency can do is be honest about the limits of what is knowable, build measurement systems that do not assume Google's tools are sufficient, and treat GEO as a real discipline rather than a rebranded SEO checklist.

The information asymmetry is real. Work around it methodically. Do not wait for Google to close the gap on your schedule.