Search Console's Generative AI Report: The Complete GEO Guide (2026)

Lee Stacey, Bourne Digital · September 2026

For years, the #1 complaint about AI search was simple:

“I have no idea if Google's AI is using my content.”

That changed this summer.

Google Search Console now has a Generative AI performance report. It shows how often links to your site appear in AI Overviews and AI Mode.

It's the first first-party GEO data Google has ever given us.

But it has some big blind spots, and if you read it the wrong way you'll draw the wrong conclusions.

In this guide I'll show you:

• Exactly what the report measures (and how Google counts it)
• What it doesn't tell you
• 6 ways to turn the data into GEO actions
• 6 other things in Search Console you should check if AI visibility matters to you

What Is the Generative AI Performance Report?

The Generative AI performance report is a Search Console report that shows how many times links to your site were shown in Google's generative AI features.

Here's a quick timeline:

• June 2026: Google announced the report and a new AI control, and tested them with a subset of UK sites first.
• 31 August 2026: Google said: “As of August 31, 2026, we've rolled out these insights to all websites worldwide.”

There are actually two reports:

1. Generative AI (Search): AI Overviews and AI Mode. (Search Console Help)
2. Generative AI (Discover): generative AI features in Google Discover. (Search Console Help)

To find them, go to Performance in the left-hand menu and look for Generative AI.

Can't see it? You're not alone. Even after the “worldwide” announcement, Google's help pages said access wasn't yet complete, and sites without enough AI impressions may not get a report at all. If it's missing, skip to Part 2. Most of that still applies.

What the Report Tells You

The report has one metric: impressions.

Google defines them like this:

“Impressions are how many times links to your site were shown to a user in a generative AI feature on Google Search.” (Search Console Help)

You can slice those impressions four ways:

(The Discover version has Pages, Countries and Dates, but no Devices. Search Console Help)

How Google counts impressions (this matters)

There's one detail that trips a lot of people up.

The chart counts by property. The table counts by URL.

If two of your pages appear in the same AI Overview, the chart counts that as one impression for your site. But in the Pages table, each URL gets its own impression. (Search Console Help)

That's why the table total is often higher than the chart total. It's not a bug.

Pro Tip: If the table total is well above the chart total, that's actually good news. It means Google often cites several of your pages in one answer, which is a sign of topical depth.

What the Report Doesn't Tell You

Now for the limitations. And there are quite a few.

❌ No clicks. The report shows impressions only.

❌ No queries. You can't see which prompts triggered your appearance.

❌ No position. You can't tell if you were the first cited source or the eighth.

❌ No mention vs citation split. It counts links to your site. If AI names your brand without linking to you, it doesn't show up.

❌ No Search Labs experiments. Google excludes features still in Labs. (Search Console Help)

❌ No other AI engines. Nothing from ChatGPT, Perplexity, Claude or the Gemini app.

❌ The 1,000-row limit applies, same as the normal Performance report.

Wait, so where are the clicks?

Here's something most people miss.

AI clicks are already in your normal Performance report. They're just blended in.

Google's Performance report documentation says that clicking a link to an external page in an AI Overview or AI Mode counts as a click. And for AI Overviews, “all links in the AI Overview are assigned that same position.” (Search Console Help)

So in the standard report:

• AI Overview and AI Mode clicks are counted
• But you can't separate them from classic blue-link clicks

That's why the Generative AI report is useful. It's the only place in GSC where AI visibility is isolated, even if it's impressions only.

Part 1: 6 Ways to Use the Report for GEO

Let's get practical.

#1: Find your “AI-citable” pages

Open the report, click the Pages tab and sort by impressions.

This is your list of pages Google's AI already trusts enough to link to.

Don't be surprised if it isn't the homepage or the big product pages. Look out for:

• A detailed guide or explainer
• A comparison or “vs” page
• A pricing or specs page
• An old blog post that answers one question really well

Study them. What do they have in common? Clear answers near the top? Specific facts and numbers? Question-led headings?

That's your internal template for citable content.

#2: Calculate your “AI Visibility Ratio” per page

This is my favourite analysis. It takes about 20 minutes.

Step 1: In the Generative AI report, set the date range to the last 3 months, click Pages and export.

Step 2: In the standard Performance → Search results report, use the same date range, click Pages and export.

Step 3: Join the two exports on URL in a spreadsheet and add a column:

AI Visibility Ratio = AI impressions ÷ Web impressions

Now sort. You'll get three groups:

Note: Unavailable values show as ~ or - in the interface but turn into zeros when you export. (Search Console Help) So a “0” in your spreadsheet might mean “below the reporting threshold”, not “never appeared”. Treat low numbers carefully.

#3: Reverse-engineer the prompts behind your AI impressions

The Generative AI report has no queries. But the standard Performance report does.

So here's the workaround:

Step 1: Take a top page from #1.

Step 2: Go to Performance → Search results, filter by that page, then click Queries.

Step 3: Add a regex query filter to surface the conversational queries most likely to trigger AI features.

Search Console supports RE2 regex in query filters. (Search Console Help) Here are three I use all the time:

Question queries:

^(who|what|when|where|why|how|which|is|are|can|does|do|should|will)\s

Long queries (8+ words):

^(\S+\s+){7,}\S+$

Comparison and evaluation queries:

\b(vs|versus|alternative|alternatives|compare|best|top|review|pricing)\b

This won't show you the exact prompts behind your AI impressions. But it gives you a very good picture of the question-shaped demand each cited page is serving.

Pro Tip: Don't dismiss long, oddly specific queries as noise. They're often a great source of real buyer language for FAQs and headings.

#4: Fix your “ranks but isn't cited” pages

Go back to the low ratio, high web impressions group from #2.

These pages already rank. Google clearly thinks they're relevant. But AI features aren't linking to them much.

For each one, check:

• Is there a direct answer in the first 1–2 sentences under each key heading?
• Are the facts specific? (Numbers, named integrations, certifications, dates.)
• Is it up to date? Old stats and old screenshots make a page less useful to cite.
• Does it say something other pages don't? Google's own guidance talks about creating “non-commodity content” with a “unique point of view”. (Google Search Central)

That last one is the big one. If your page says the same thing as the ten pages above it, AI has no reason to cite you specifically.

#5: Compare countries and devices

AI features don't roll out everywhere at the same speed.

Use the Countries tab to see where your AI impressions actually come from. For a B2B company selling into the UK, US and EU, this tells you which markets' AI answers you're part of, and which you're missing.

Then check Devices. If almost all your AI impressions are on mobile but your buyers research on desktop at work (as many B2B buyers do), that's worth knowing.

#6: Annotate everything you change

Search Console has custom annotations. You can add notes to performance charts, up to 125 characters, and they're visible to anyone with access to the property. (Google Search Central Blog)

Every time you make a GEO change, like rewriting a key page, launching a comparison page or getting listed on a major review site, add an annotation.

Three months from now you'll be able to see exactly which changes moved your AI impressions, rather than guessing.

Part 2: 6 More Things to Do in Search Console for GEO

The Generative AI report shows you the outcome.

These six checks make sure nothing is blocking that outcome.

#7: Check the Search generative AI control (seriously)

Alongside the report, Google launched a Search generative AI control. You'll find it under Settings.

It has three options:

1. Include (default)
2. Exclude
3. Inherit from parent (for child properties)

If it's set to Exclude, your content won't appear as links in AI Overviews, AI Mode or Discover's AI features, and won't be used to ground AI answers. Google says exclusion takes effect “within 1-2 days” but can take longer. (Search Console Help)

Google also says the control doesn't affect your regular search rankings. (Search Console Help)

Why check it? Because in a B2B company, lots of people have Search Console access. Agencies, developers, a former marketing manager. It takes one person clicking Exclude “just to be safe” to quietly wipe you out of AI search.

Check it now. Then make it part of your monthly checklist.

#8: Check your robots.txt for Google-Extended

There's a lot of confusion about Google-Extended, so let's be precise.

According to Google's crawler documentation:

• Google-Extended does not affect your inclusion or ranking in Google Search.
• It does control whether your content can be used to train future Gemini models, and for grounding in Gemini Apps and Vertex AI.

(Google Crawling Infrastructure)

So blocking Google-Extended won't remove you from AI Overviews. That's what the control in #7 is for.

But it may reduce your chances of being cited when someone asks the Gemini app a question. That's a GEO decision, not just a legal one.

Use the robots.txt report (under Settings) to confirm what Google is actually reading. Then make a deliberate choice, don't inherit one from a developer's copy-paste.

#9: Make sure your key pages are indexed

This one sounds basic. But it's the foundation of everything.

Google says that to be eligible for generative AI features, “a page must be indexed and eligible to be shown in Google Search.” (Google Search Central)

No index, no citation. It's that simple.

So:

1. Open Indexing → Pages and look through the “Why pages aren't indexed” reasons.
2. Run your top 20 commercial pages (pricing, product, comparison, key guides) through URL Inspection.
3. Use Test live URL and view the rendered HTML. If your key answers only appear after JavaScript runs, check that they appear in the rendered version.

Also check snippet controls. Google says nosnippet, data-nosnippet and max-snippet limit how your content appears in AI features too. (Google Search Central) A max-snippet:50 added years ago “for SEO” could be holding back your AI visibility right now.

#10: Use the branded queries filter as a GEO success metric

Here's a question every B2B marketing director asks:

“If AI mentions us but nobody clicks, how do we know it's working?”

One of the best answers is branded search demand.

When buyers see your name in an AI answer, a lot of them go and Google you. That shows up as branded queries.

Search Console now has a branded queries filter, which uses AI-assisted classification to split branded from non-branded queries. It became available to all eligible sites in March 2026, but only for domain-level properties with enough query volume.

Track branded impressions and clicks month over month, alongside your AI impressions.

If AI impressions go up and branded demand goes up, you've got a much stronger story for the board than impressions alone.

Pro Tip: Google says the classifier isn't perfect, especially for brands with common words in their names. Sense-check the branded list before you report on it.

#11: Use Query groups to map topics

In the Insights report, Search Console now has Query groups, which cluster similar queries using AI. It's only available to properties with enough query volume.

Why does this matter for GEO?

Because AI features answer topics, not keywords. As we covered in our GEO with Google Search guide, AI Mode “fans out” a single question into lots of related sub-searches.

Query groups give you a rough map of the topics Google associates with your site. Compare that map to the topics you want to be cited for. The gaps are your content plan.

#12: Watch your Links report for third-party signals

The Links report shows your top linking sites.

For GEO, don't read this as a link-building report. Read it as a list of third parties that are talking about you.

Ask:

• Are the review platforms, analyst sites and industry publications in my category on this list?
• Are my competitors on sites that I'm not? (Cross-check with a backlink tool.)

AI engines lean heavily on third-party sources. The Links report is a quick first check on whether you have any.

One word of warning. Google's own guidance says that “seeking inauthentic 'mentions' across the web isn't as helpful as it might seem.” (Google Search Central) Earn real coverage. Don't buy it.

What Google Says Not to Bother With

Google published a Guide to Optimizing for Generative AI Features (last updated July 2026). It's refreshingly direct about a few popular GEO “tactics”. (Google Search Central)

llms.txt: “Google Search ignores them.”

Chunking your content: “There's no requirement to break your content into tiny pieces.”

Writing specially for AI: “You don't need to write in a specific way just for generative AI search.”

Structured data: “Structured data isn't required for generative AI search”, though Google still calls it a good idea as part of your overall SEO strategy.

So before you spend budget on an llms.txt project, spend it on the checks above.

(Note: this is Google's position for Google's AI features. Other AI engines may behave differently.)

Your GEO Dashboard in Search Console

Here's the whole thing as a monthly routine:

It's not a perfect picture of AI visibility. No single tool is yet.

But it's free, first-party, and straight from Google. For most B2B companies, that's a far better starting point than guessing.

What's Next for the Report

The Generative AI report is still new. Google has said it's working with site owners to figure out what insights would be most helpful, and plans to add more over time.

So set up the routine above now. When new metrics arrive, you'll already have a baseline to compare them against.

If you'd like a hand turning the data into a GEO plan, take a look at our AI Search Audits.

← Back to all posts