The Search Console AI report: what it tells you about Google AI Mode, and what it doesn't

Since 31 August, every website on the planet has a new report in Google Search Console. The Search Console AI report shows how often your pages turned up inside AI Overviews and Google AI Mode.

It is the first time anyone outside Google has been able to measure this on their own site. That is genuinely new, and it is worth turning on.

It is also, quietly, much less than it looks like. What follows is what the report contains, the four things it leaves out, what the opt-out costs, and the one thing I would actually trust it for.

Read it as a visibility signal, not as a traffic number — and be careful what you conclude from a line going up.

What the Search Console AI report actually shows

Google calls it the generative AI performance report. It announced the new Google Search Console feature on 3 June 2026 and finished rolling it out on 31 August. The footnote on the announcement is one sentence: "As of August 31, 2026, we've rolled out these features to all websites worldwide."

You will find it in Google Search Console under Performance, as a separate view next to the regular report. There are two of them — one for Search, one for Discover.

Here is everything in it:

Google's own description of the metric: "Impressions are how many times links to your site were shown to a user in a generative AI feature on Google Search."

That is it. Four dimensions, one number. Not in it: clicks, CTR, position, queries, and any split between AI Overviews and AI Mode.

The four things the Google Search Console AI report doesn't tell you

It doesn't tell you whether anyone clicked

There is no click column. No CTR. No position. Google says it is "continuing to work with website owners to understand what insights and data would be most helpful," and mentions "adding additional metrics over time." So clicks may come. They are not here now.

This matters more than it sounds. An impression in an AI answer is not like an impression in the blue links. In the blue links, an impression means your title and description were on screen and a click was the obvious next move. In an AI answer, an impression can mean your page was one of several links tucked beside a paragraph that already answered the question.

The number the report gives you is the one number that cannot distinguish between those two.

It doesn't tell you what anyone searched for

The regular Performance report has a Queries tab. This one does not. The dimension list is Pages, Countries, Dates, Devices — and that list is complete.

So you can see that your devlog appeared in an AI answer 43 times last week. You cannot see whether it appeared because someone asked about your game, or because someone asked a general question about devlogs and Google pulled a sentence from you.

Those are two completely different pieces of news, and the report gives you the same row for both.

It doesn't tell you which feature you appeared in

This is the one that surprised me.

The help page lists what the report covers: an intro line, then two bullets — AI Overviews, and AI Mode. Together. There is no filter, no tab, no comparison view. One number covering both.

AI Overviews is a box that sits above the ten blue links, on a normal results page, for someone who may well scroll past it. AI Mode is a separate conversational surface that someone deliberately opened instead of searching normally. Different people, different intent, different odds that anyone ever reaches your site.

The report adds them together.

Google leaves itself room here, too, right under the list: "We expect to update this list over time as we develop Google Search." Which is fair — but it means a chart you compare month over month may not be counting the same thing at both ends.

It doesn't count the way you'd assume

Two of your pages showing up in the same AI answer counts as one impression in the chart. Google spells this out: "if two results from the same site appeared in a generative AI search results feature, they count as a single impression in the chart total."

If your strategy is to be cited repeatedly, the metric is designed not to reward you for it.

Why you might not see the generative AI performance report at all

There are three reasons the report can be missing, and Google lists all three:

1. "Not all properties have access to the report, as we're rolling out over time." (Yes — this still sits on the help page underneath the note saying the rollout finished on 31 August.)
2. "Your site hasn't received enough impressions in generative AI features on Google Search."
3. You have excluded your site from generative AI features yourself.

Reason two is the one that will bite an indie dev. The site with four hundred visitors a month, the one that most needs to know whether AI Search is doing anything for it, is exactly the site that may open Search Console and find nothing there.

An empty report is not evidence that you are invisible in AI Search. It is evidence that Google won't show you the number. Those are not the same finding, and only one of them is actionable.

Google is measuring one thing and promising another

The blog post announcing all of this makes a case for AI Search being good for websites. It has numbers: "AI Overviews now has over 2.5 billion monthly active users, while AI Mode has surpassed one billion monthly users." Google's own figures, not an independent count — but they are the scale we are talking about.

Then, a paragraph later, on the same page:

We've increased the number of inline links directly within responses and added helpful website previews to encourage people to click through.

Click through. On the same page that announces a report with no click data in it.

I don't think that is a conspiracy. I think it is what a first version looks like. But it is worth naming, because it tells you what the report is for: it is a visibility number, published by the company whose visibility it measures, at the moment that company would like you to believe visibility is converting.

If someone shows you a chart that says AI Search sent them traffic, ask which column that came from. There isn't one.

The AI Search opt-out, and what it actually costs

Next to the report there is an opt-out toggle that site owners control: Settings → Search generative AI. Three states — include (the default), exclude, or inherit from a parent property. Google announced it on the same day as the report, 3 June 2026, as something it was "beginning to test"; it is now live alongside the report.

Turning it off does what it says: "You won't receive any traffic or impressions from these features." Both. For a discovery blog, that is almost always the wrong trade — you are switching off a surface in exchange for nothing you can measure.

Three things about it that are easy to miss:

It doesn't stop Google using your content. Only displaying it. Google: "Your site's content may still be used to power Google Search as a whole, like helping our systems better understand the language of users' queries and the language on pages on the web."

It is not an AI training opt-out. Explicitly: "This control doesn't affect AI training; to limit training of the models used to generate responses in Search generative AI features, use Google-Extended."

It is inherited downwards. A property inherits the setting from its nearest parent that has one — otherwise from the top-level domain property. If you run a subdomain and flip the switch on the main domain, you have flipped it on the subdomain too, without visiting it. Worth knowing if you are splitting a site across subdomains for any reason.

And then there is the sentence everyone quotes: "This control isn't used as a ranking or inclusion signal affecting other parts of Search."

That is Google saying so. There is no way for you, me, or anyone outside Google to verify it. It is not a lie and it is not a fact — it is a statement of intent from the party with the incentive. Treat it the way you would treat any other unverifiable assurance: as the best information available, and as information from an interested party. If you are weighing the opt-out, weigh it on the traffic and impressions you would definitely lose, not on a ranking risk you cannot measure either way.

The thirty-six hours of AI answers nobody measured

One small thing that explains why I would not build a strategy on any AI visibility number yet, including this one.

On 2 September, Google AI Mode started running on Gemini 3.8 Flash and largely stopped including links in its answers. Gagan Ghotra spotted it, Glenn Gabe confirmed it, and Google's Robby Stein replied in public: "Hey — thanks for flagging. This isn't working as intended, and we'll roll out a fix soon." It was fixed by the morning of 4 September.

Roughly thirty-six hours in which citations in one of the two surfaces this report covers were close to nothing.

No tracker put a number on it. The evidence is screenshots from two people. It is in your Search Console data as a dip you have no way to explain, mixed in with the other surface, with no query breakdown to check against.

That is not a reason to ignore the report. It is a reason to read a month of it before you read a week of it.

So what is this AI visibility number actually good for?

Three things, and they are real — none of which replaces your ordinary SEO reporting:

A baseline. You now have a date from which you can say "this is how visible we were." That is worth more the earlier you start it, and it costs nothing.

Page-level surprises. The Pages dimension is the most useful part. The post you thought was your weakest often turns out to be the one AI answers keep reaching for — usually because it answers a narrow question plainly, in text that is actually in the delivered HTML. That is a content signal you cannot get anywhere else.

A check on your own claims. If you have done anything specifically for AI visibility — an llms.txt file, a round of SEO work aimed at getting cited in generative answers — this is the first tool that can tell you whether the AI impressions moved. Not whether it made you money. Whether the impressions moved.

What it is not good for: proving AI Search sends you traffic, comparing yourself to anyone else, or making a decision that depends on knowing which feature you appeared in.

Worth remembering what Google itself said about all this, on Bluesky in late August: "I'm not quite sure what you're asking; from our POV there's nothing really special you need to do for generative AI responses in search." John Mueller — including the first half of the sentence, which usually gets cut. He was not certain what he was being asked. That is roughly where the whole field is right now: a new dashboard, a lot of confidence, and one metric.

Turn it on. Look at it monthly. Don't build a business plan on it yet.

FAQ

Sources

• Google Search Central Blog — Hillel Maoz and Moshe Samet, "Introducing Search Generative AI performance reports in Search Console", 3 June 2026, with the 31 August 2026 worldwide-rollout note
• Google Search Console Help — "Generative AI performance report (Search)"
• Google Search Console Help — "Search generative AI control"
• Google Blog — Mrinalini Loew, "New opportunities, control and insights for website owners", 3 June 2026, updated 31 August 2026
• John Mueller on Bluesky
• Search Engine Roundtable — "Google To Fix Citation Bug With Gemini 3.8 Flash In AI Mode", 3 September 2026 (secondary source)