Google Now Shows Where AI Surfaces Your Site. Here Is What the Report Will Not Tell You

For two years, every site owner has been asking the same question: people clearly see AI answers instead of blue links now, so are we in those answers or not? Until this month, the honest response was a shrug. You could run experiments, buy third party trackers or trust your gut, but Google itself said nothing.

That changed quietly. Google announced Generative AI performance reports in June, and around 11 August the report appeared in every Search Console property in the world. No banner, no fanfare. If you manage a website, you now have official data on your AI visibility, and you may not have noticed.

What the report actually shows

Open Search Console and look under Performance. There is a new entry called Generative AI. It isolates, for the first time, the impressions your pages earn inside AI experiences: AI Overviews, the summaries that sit on top of many results pages, and AI Mode, the conversational search Google keeps expanding.

You can break the numbers down by page, country, device and date. That alone answers questions that were unanswerable in July. Which of your pages does AI actually pull into its answers? Did that technical article you published quietly become your most cited asset? Is your AI visibility growing while your classic clicks stay flat?

Before you panic about a drop

One warning before you read your own graph. Google confirmed a logging error that under counted Generative AI impressions for data starting 13 August. The dip many sites see around that date is a measurement artefact, not a visibility crash. It is a useful first lesson in reading this report: before you diagnose a trend, check whether the instrument moved.

The three silences

Now the part Google is quieter about. The report deliberately withholds three things, and each silence changes how much you can conclude.

No queries. You can see that a page was surfaced by AI, but not for which questions. You cannot map prompts to pages the way you map keywords to rankings. Optimisation stays partly blind: you know AI likes a page, you do not know why.

No clicks. The report counts impressions inside AI experiences, but does not tell you how many people clicked through from them. This is the uncomfortable one. AI answers are built to be read, not clicked. Google showing you visibility while hiding follow through keeps the hardest question open: what is an AI citation actually worth to your business?

No API. The data is not exposed through the Search Console API. If you want history, you export by hand, month after month. For agencies and reporting tools, that is a deliberate speed bump.

What to do with it this month

Imperfect data is still data, and this is the first official measurement of AI visibility anyone has had. Five practical moves:

• Record a baseline now. Screenshot or export the report monthly. There is no API, so the history you keep is the history you will have.
• Compare shares, not totals. Put AI impressions next to classic search impressions. If the AI share grows while clicks stagnate, your visibility is migrating into answers. That is strategy shaping information, not trivia.
• Study the page list. The pages AI cites are rarely the pages you would have guessed. Clear, well structured, question answering content tends to win. Learn from your own winners and shape new content the same way.
• Do not judge the 13 August week. The logging bug sits exactly there. Start clean readings from late August.
• Watch what the core update does to it. Google began rolling out its August core update on 26 August. Core updates now move AI citations as much as rankings. The report is where you will see it.

Why this matters more than it looks

We wrote recently about what the numbers say on answer engine visibility: AI assistants send a small share of traffic today, but they increasingly decide who gets found at all. The missing piece was official measurement. This report is that piece, incomplete on purpose, but official.

Visibility without clicks is becoming a currency of its own. Brands will argue about whether an AI citation sells anything. The ones who win that argument will be the ones who measured theirs from the start.