Google's John Mueller agreed with an SEO who called the generative AI performance report an inaccurate snapshot, conceding that position inside AI answers is 'hard to do in a way that makes it useful' and is tracked only as a block.
2026-09-13 · by Moe Ameen
Google's generative AI performance report in Search Console — the report that finished rolling out to sites worldwide on August 31, 2026 — is meant to show how often your pages appear inside AI Overviews and AI Mode. On September 13, 2026, Search Engine Journal reported that Google Search Advocate John Mueller had publicly agreed with SEOs who argue that the report gives an inaccurate picture of how a site actually performs in AI search.
The exchange started on Reddit, where a site owner argued that the report still leans on the old "ten blue links" concepts of impressions and position, which map badly onto an AI answer where a link can be cited inside a synthesized paragraph, hidden behind a "show more" expander, or shown without ever being scrolled into view. Mueller did not push back. He said position "is hard to do in a way that makes it useful, so we're currently tracking it like we do for many search features (as a block), & it's not separated out in the Gen-AI performance report," and noted that "search results pages have a lot of ways for users to interact nowadays, so the old 'position 1–10' is hard to map." Search Engine Journal's Roger Montti reported the acknowledgment.
Stacked up, the gaps are significant. The report shows impressions but no click counts and no click-through rate, so you can confirm a page picked up AI impressions but not whether anyone clicked. Position inside an AI answer is not broken out — it is folded into a block. And the specific prompts that triggered the appearance are not exposed, which Google attributes to user privacy and the way AI responses synthesize multiple sources. Mueller said he was "happy to discuss with the team" what position tracking would actually be useful, but there was no concrete commitment to new metrics; Google has separately said it plans to add metrics over time based on feedback.
Treat this as a candid limitation rather than a bug. The first-party AI-visibility scoreboard exists now, and it is better than the nothing site owners had before — but Google itself is saying the numbers on it do not yet answer the questions creators most want answered.
The trap here is spending real effort tuning to a dashboard Google just admitted is inadequate. When the only visible metric is impressions — no clicks, no CTR, no position, no queries — you cannot run the tight measure-and-optimize loop that classic SEO ran on. So the rational response is to stop trying to reverse-engineer an unmeasurable channel and instead do the two things that still work: publish enough substantive, on-brand material that AI answers have something worth citing, and keep landing on channels where the feedback is real. [Kompozy](/) is the production engine for exactly that posture. Point it at one source and it generates a batch of genuinely distinct pieces from a single [Persona Brief](/glossary/persona-brief) — a long-form [Blog Article](/glossary/output-buckets) and text an AI-search engine can actually index and quote, plus [Persona Shorts](/glossary/persona-shorts), [Clipped Shorts](/glossary/clipped-short), and brand-exact [Carousel](/glossary/hyperframes) posts — instead of you hand-making a corpus big enough to be worth quoting.
Then it publishes where you can see results. [Autopilot](/glossary/autopilot) schedules the batch across the eight social platforms plus a blog and an email newsletter to Mailchimp, behind a per-post review gate. Those surfaces report clicks, opens, saves, and watch time — the after-the-impression signals the AI report withholds — so your read on what's working comes from channels with honest analytics rather than a block-position estimate. Use the Search Console AI report for the one thing it's good at: a coarse "are we being surfaced at all" pulse. Use Kompozy to make sure that being surfaced is backed by a body of work across measurable channels, so a quarter where the AI numbers are murky doesn't leave you flying blind.
Effectively, yes. Search Engine Journal reported on September 13, 2026 that Google Search Advocate John Mueller had agreed on Reddit with SEOs who said the Search Console generative AI performance report gives an inaccurate picture of AI-search performance. He conceded that position inside AI answers is "hard to do in a way that makes it useful," so it is tracked only as a block and not separated out in the report. He offered to discuss more useful position tracking with the team but did not commit to new metrics.
The report shows impressions — how often your pages appear in AI Overviews and AI Mode — but no click counts, no click-through rate, and no per-answer position (position is folded into a block). It also does not expose the specific prompts that triggered your appearance, which Google attributes to user privacy and the way AI answers synthesize multiple sources. So you can see that a page was surfaced, but not whether it earned any clicks or where it ranked within the answer.
A modern results page and an AI answer offer many ways to interact — a link can be cited inside a synthesized paragraph, tucked behind a "show more" expander, or shown without ever being scrolled into view — so the old "position 1–10" model maps badly onto them. Mueller said Google currently tracks these appearances as a block, the same way it handles many search features, because a single ranked position is not meaningful in that layout.
Don't build a content strategy around climbing a metric Google itself calls inadequate. Instead, publish enough substantive, on-brand material that AI answers have something worth citing, and lean on channels where analytics are real — blog clicks, email opens, and native social engagement. A tool like Kompozy generates a full batch of on-brand content from one source and publishes it across the eight social platforms plus blog and email, so your read on performance comes from measurable channels while AI-search reporting matures.