Why Google’s AI search reporting is still not enough

Google has now acknowledged what many SEO and GEO teams were already feeling: Search Console’s AI search reporting is useful, but incomplete. It can show that your pages appeared in generative Google surfaces, yet it still cannot tell you the one thing teams actually want to know: how visible your individual link really was inside the answer.

That matters because AI search changes the unit of measurement. In classic SEO, rank was a rough proxy for attention. In AI search, inclusion, prominence, framing, and click potential can all diverge. A page can be counted, assigned a strong position, and still have weak commercial value.

• Google’s new report is mainly an exposure report, not a full visibility report.
• In AI Overviews, every cited link can inherit the same position as the whole AI block.
• Impressions do not always mean a user meaningfully saw your citation.
• For GEO teams, native Google data is helpful, but not enough to guide strategy alone.
• The right response is to build a layered measurement stack, not chase one dashboard.

What did Google actually admit?

Google admitted that the current position reporting for AI search is hard to make genuinely useful. John Mueller said the metric is currently tracked as a block, which means the reporting still reflects the AI feature’s placement on the page more than the visibility of any one cited URL.

That is consistent with Google’s own documentation. In its June 2026 Search Central announcement, Google introduced dedicated Search Console reporting for generative AI features in Search, including AI Overviews and AI Mode. The rollout started for a subset of sites and became globally available on August 31, 2026. The report gives site owners a filtered view of impressions by page, country, device, and date.

The important word there is filtered. The report is not a separate traffic universe. Google says the data is already included in the overall performance report, which means teams should not treat it like a clean new source of truth. It is a narrowed lens on the same reporting framework.

Mueller’s acknowledgement matters because it moves the conversation from suspicion to confirmation. This is no longer just an SEO complaint about imperfect tooling. Google itself is effectively saying that the old reporting logic does not map neatly onto modern AI search behavior.

A simple example makes the problem obvious. Imagine your pricing page appears as a secondary citation inside an AI answer at the top of the page. Search Console may assign that page the position of the AI feature itself. On paper, that can look like elite visibility. In reality, the link may have been peripheral, partly hidden, or less likely to attract attention than the label suggests.

Why is the current report still insufficient?

It is insufficient because AI search creates layers of visibility that the current Search Console model compresses into legacy metrics. The report can tell you that a URL was present. It cannot fully express how prominent that URL was, how central it was to the answer, or whether the user was likely to notice it.

Google’s help documentation on impressions, clicks, and position explains the issue in plain terms. For AI Overviews, standard impression rules apply, and the AI Overview occupies a single position in search results. All links inside that overview are assigned that same position. That is neat for reporting consistency, but weak for decision-making.

Here are the three biggest gaps.

• Impression inflation risk: a page can earn an impression because the result was served, even if the user did not meaningfully engage with that citation.
• Hidden visibility risk: links that sit behind expandable elements can have some practical exposure value while still being measured differently.
• Position flattening: a prominent citation and a marginal citation can inherit the same position number because the block, not the link, is being ranked.

This is exactly why old SEO language starts to wobble in AI interfaces. Position 1 used to mean something reasonably intuitive. In an AI Overview, that same label can mean at least three different things: your page was central to the answer, your page was merely included, or your page was technically present somewhere inside a top-placed block.

For AI Mode, the story is only slightly better. Google says AI Mode follows standard Search position methodology, but the experience itself is interactive, multi-step, and based on subtopic exploration. Follow-up questions are counted as new queries. That makes sense from Google’s product perspective, but it means the marketer looking at one dashboard still does not get a clean picture of session-level influence or answer-level prominence.

If you want a deeper breakdown of how these numbers get misread, our article on why Search Console AI data is misleading marketers is worth reading next.

Why does this matter so much for GEO teams?

Because GEO is not just about being present. It is about being chosen, understood, and cited in a way that helps the brand. That is a different standard from classic reporting, where a high ranking URL often did enough to justify the KPI.

In generative search, a page can appear without owning the answer. A brand can be mentioned without being cited. A citation can drive influence without a click. And a click can arrive without clean attribution. Once you accept that, you stop asking whether Google showed your page and start asking what role your page played in the generated response.

That is why AI search visibility needs a broader definition than native platform reporting can offer. Visibility is not only exposure. It is exposure plus prominence, brand mention quality, source citation, answer framing, and downstream business effect.

Take a practical example. A comparison page for a B2B software brand appears in an AI Overview for a high-intent query. Search Console logs the impression and assigns a strong position because the Overview sat near the top of the page. But the answer primarily recommends two competitors, your page is a secondary source, and the user never clicks. The dashboard says presence increased. The commercial reality is that you may still be losing the buying conversation.

This is also why the old debate of “does AI search traffic convert?” misses the point. Sometimes the user clicks. Sometimes they do not. Sometimes the answer changes brand preference before the visit ever happens. For teams trying to connect visibility to outcomes, our analysis of AI Overview traffic that GA4 may be hiding adds another layer: even when traffic exists, native analytics can still understate it.

The measurement problem is no longer theoretical. It is operational. If you cannot separate presence from prominence, and prominence from commercial value, your reporting will look precise while pushing the team toward the wrong conclusions.

BotRank's Take

Google’s new AI reporting is still useful. It just is not enough to serve as the operating system for GEO. The mistake would be to treat Search Console as if it had already solved AI measurement. It has not. It solved one part of the problem: confirming that your pages can appear in Google’s generative surfaces. That is helpful, but it is still only the outer layer.

The BotRank feature that matters most here is AI Visibility tracking. It lets teams run stable prompts across multiple models, compare how often the brand appears, track which competitors get named, and observe how representation changes over time. That matters because Google’s own report cannot tell you whether your page was central to the answer, whether your brand was positively framed, or whether another model tells a completely different story on the same topic.

The honest view is simple: Search Console should stay in the stack, but it should not be the stack.

What should teams measure instead?

They should measure layers, not one headline number. Search Console can own the exposure layer for Google. After that, teams need a second layer for answer quality, a third for source diagnostics, and a fourth for business impact.

A practical stack looks like this.

• Exposure: how often your pages appear in AI Overviews or AI Mode.
• Answer role: whether your brand is central, peripheral, or absent from the generated answer.
• Source quality: which of your pages are being cited and whether they actually support the claim.
• Business effect: clicks, assisted conversions, branded search lift, or pipeline influence.

This is where Source Analysis becomes valuable. In AI search, the page you want cited and the page Google or another model actually trusts are often not the same. Teams need to see which assets are being used, whether those pages mention the brand clearly, and whether the cited evidence is commercially useful or just technically present.

Consistency matters too. If you keep changing your test prompts, your benchmark breaks. That is why Prompts Studio is such a practical piece of the workflow. It gives teams a reusable panel of prompts so they can compare like with like over time instead of reinventing the measurement every reporting cycle.

Once the data is visible, the hard part is turning it into work. A recurring issue in GEO programs is that teams can see the gap but do not know how to prioritize fixes. That is where Recommendations help: not by replacing editorial judgment, but by converting noisy visibility findings into concrete actions.

If you are rethinking the KPI layer itself, our article on the AI visibility metrics brands still miss is a strong next read. The short version is that mentions, citations, rankings, and conversions are not interchangeable signals. Treating them as one score only hides what actually changed.

How should brands react right now?

Start with restraint. Do not throw out Search Console’s AI report, but do stop over-reading it. It is useful for page discovery, pattern spotting, and directional exposure trends. It is weak as a proxy for true citation prominence or commercial success.

Then rebuild your reporting around questions that match how AI search actually works.

• Which pages appear most often in Google’s AI surfaces?
• Which of those pages are visibly associated with our brand?
• Which answers mention us, cite us, or recommend a competitor instead?
• Which appearances correlate with traffic, leads, or assisted conversions?
• Which missing pages or weak pages should we improve first?

A concrete example: suppose your support documentation suddenly starts appearing in AI Mode for product comparison prompts, while your commercial landing pages disappear. Search Console may show rising exposure, and a shallow review might call that progress. A better diagnosis would ask whether Google is finding your brand useful only for factual support content, while failing to trust your higher-intent pages for recommendation-style answers. That is a content architecture problem, not a reporting win.

The same logic applies to executives. If leadership sees rising AI impressions, the tempting story is “we are winning AI search.” Sometimes that will be true. Sometimes it will mean only that your URLs are present inside more generated answers. Those are not the same outcome.

The brands that handle this well will not be the ones with the prettiest dashboard. They will be the ones that separate signal from interpretation fastest, then act on the right layer. In AI search, better measurement is not an analytics luxury. It is what keeps the team from optimizing for the wrong thing.

FAQ: what should marketers do with Google’s AI search reporting?

Is the Search Console AI report useless?

No. It is useful for understanding where your pages appear in Google’s generative search features, especially by page, device, country, and date. The problem starts when teams treat it as a full visibility or business-impact report.

Can a position 1 label be misleading in AI Overviews?

Yes. In AI Overviews, Google assigns all links inside the Overview the position of the whole AI block. That means a highly visible citation and a barely noticed citation can share the same reported position.

Should we stop tracking AI impressions?

No. Keep tracking them as an exposure signal. Just pair them with answer review, source analysis, and outcome metrics so the team does not confuse presence with impact.

Does this reporting include everything happening in Google’s AI experiences?

No. Google’s documentation says Search Console does not include data from Search Labs experiments because those experiences are still in active development. That is another reason the native report should be treated as partial, not exhaustive.

What is the practical next step for GEO teams?

Use Search Console as the first layer, not the final layer. Then add prompt-based testing, citation review, and page-level source diagnostics so you can see not only whether you appeared, but whether the answer actually worked in your favor.

The takeaway is blunt: Google finally gave the market AI search reporting, then admitted it still cannot express the whole reality of AI visibility. If your team wants a truer picture of how your brand appears across answer engines, build that measurement stack now, before a flattering position number turns into a false sense of progress.