How AI Search Works
1
Attention Is All You Need: The Google Paper That Accidentally Ended Google's Search Monopoly.
On June 12, 2017, eight researchers at Google Brain published a 15-page paper. It accidentally made Google's own search business obsolete.
2
How Large Language Models Actually Work: Tokens, Context Windows, and Why Your Content Gets Ignored.
The answer to "how do I get cited by AI search?" is technical, not strategic. It's rooted in how large language models actually process information.
3
What Is RAG: the Only Thing That Matters for AI Search.
Once you see how Retrieval-Augmented Generation works, the rest of AI search visibility clicks into place. It's the mechanism that connects your content to every answer these systems produce.
4
The AI Search Race: ChatGPT, Gemini, Claude, Copilot, and Perplexity. A Technical Timeline.
The models, the milestones, and the architectural decisions that shaped every AI search product in market. A reference for anyone trying to make sense of where this came from.
5
How Google AI Overviews and AI Mode Actually Work: Why Your SEO Rankings Don't Guarantee AI Visibility.
AI Overviews aren't a ranking bonus layered on top of existing positions. In May 2024, Google restructured how most people receive information. The blue links became supplementary.
6
How to Be Surfaced in AI Search: What the Platforms' Own Documentation Actually Says.
Most GEO content produces a list of tactics and implies they work across all platforms. This article goes directly to what each platform has actually published about how its own system works.
7
Why the Bing AI Performance Report Is the Most Underused Tool in GEO.
Microsoft's Bing AI Performance report is the only first-party citation dashboard from any AI search platform. Most people working in GEO aren't using it.
8
The GEO Documentation Gap: Why Three Platforms Haven't Told You How to Rank in Their Systems.
Google published a guide. Microsoft published measurement tooling. Anthropic, OpenAI, and Perplexity have published nothing equivalent. Understanding why is more useful than waiting.
9
Constitutional AI vs RLHF: Why the Alignment Method Affects What Gets Cited.
ChatGPT is aligned with RLHF. Claude is aligned with Constitutional AI. These aren't interchangeable, and the difference affects what each model treats as trustworthy and citable.
10
Training Data Is the New Ranking Factor: The Common Crawl AI Visibility Audit Explained.
Before RAG, before indexing, a site has to be reachable by the crawlers that feed AI training data. The Common Crawl AI Visibility Audit is the framework for checking that upstream layer.
11
Blocking AI Training Can Cost You Search: Cloudflare's September Defaults Explained.
Cloudflare split AI bots into Search, Agent and Training. Multi-purpose crawlers resolve to the most restrictive rule, so a training block can take your search visibility with it.
12
13
The AI Search Statistics Everyone Repeats, and Where They Actually Came From.
Gartner's 25% prediction rests on a survey of 299 people. The 14% AI Mode figure isn't Google's. AI traffic converts 3x, not 23x. The audit, and the method for running it yourself.
14
97% of llms.txt Files Are Never Read. Here's the Data.
Ahrefs studied 137,000 domains and found 97% of valid llms.txt files received zero requests. No major AI provider has adopted it, and Google says you don't need one.
15
Schema Markup Doesn't Drive AI Citations. It Still Matters.
1,885 pages that added schema, against 4,000 matched controls, tested with difference-in-differences. The effect on AI citations was nil. Keep your schema anyway, and here's why.
16
The AI Browser Died. Now Agents Browse as Your Reader.
OpenAI shut down Atlas and moved agentic browsing into a Chrome extension. The winning model runs the agent inside your reader's session, using their logins, which breaks how most people think about blocking bots.
Writing  ·  Series: How AI Search Works  ·  Part 12 of 16

Google Finally Reports AI Visibility. It Still Won't Tell You What It's Worth.

Search Console's Generative AI performance report reached every site on 31 August 2026. It reports impressions and nothing else: no clicks, no CTR, no position, no queries, and no split between AI Overviews and AI Mode. Here's how to use it honestly.

Published5 September 2026
ByThomas Cox
Read time10 minutes
Filed underSeries · GEO Measurement · Part 12 of 16

For two years the honest answer to "can you prove AI search is sending us anything?" was no. Google has now changed that answer to "partly", and the shape of the partly matters more than the fact of it.

The Generative AI performance report landed in Search Console in June 2026 and reached every site by 31 August. It is genuine first-party evidence that your pages appear in AI features, which is more than any of the estimate-based trackers can offer. It is also impressions and nothing else, which means it can prove you were seen and can never prove what that was worth.

What the report actually gives you.

One metric, four dimensions. That's the whole surface area.

The metric is impressions, defined by Google as how many times links to your site were shown to a user in a generative AI feature. The dimensions are pages, grouped by canonical URL; countries, by where the search originated; dates, by day, week or month in Pacific Time; and devices, split across desktop, tablet and mobile.

That's it. It's a visibility counter with four ways to slice it, and within those limits it's reliable, first-party and free, which is three things nothing else in this space can claim at once.

In the report Not in the report
Metrics Impressions. Clicks, CTR, average position.
Dimensions Pages, countries, devices, dates. Queries. There is no way to see what someone asked.
Surfaces AI Overviews and AI Mode, combined into one number. Any split between the two. Discover is a separate report entirely.
Access The Search Console interface, on every verified site. An API. Export is manual.

The missing query dimension is the expensive one.

Of everything absent from that list, the queries dimension is the one that changes what you can do with the data.

Impressions by page tell you which of your content AI features are drawing on. They don't tell you what question it was drawn on for. That's the difference between knowing a page is working and knowing why, and it's the difference between a report you can act on and a report you can only watch.

It's worth noticing who does provide that. Microsoft's Bing AI Performance report exposes grounding queries: the reformulated terms Copilot generated internally when it retrieved your page. As I've written in the piece on that report, those queries are the most useful thing in it, and they remain the only place any platform shows you the retrieval signal rather than just the outcome.

Google will tell you that you were surfaced. Microsoft will tell you what you were surfaced for.

Your AI clicks are already in your data. You just can't find them.

Here's the part that trips people up, and it's the single most useful thing to understand about Search Console right now.

Clicks from AI Overviews and AI Mode are not missing from Search Console. Google's documentation is explicit that sites appearing in these features are included in the overall search traffic, reported in the Performance report under the "Web" search type. The clicks are counted. They are simply blended into your organic totals with no dimension that separates them.

So there is no AI Overviews filter, and there never has been. If someone shows you a Search Console screenshot with an AI Mode search-appearance filter, it isn't real. And if a tool claims to report your AI Mode clicks or CTR from Search Console data, it is inferring them from something else, because that number is not exposed anywhere in the product.

The practical consequence is uncomfortable but clarifying. Your organic performance line already contains AI traffic, mixed in, indistinguishable. A decline you're attributing to a ranking drop might be an AI Overview absorbing the click, and you cannot separate the two from this data.

What you can honestly tell a client.

Three statements are defensible from this report, and it's worth being disciplined about staying inside them.

What you can't say is anything about clicks, CTR, position or queries in AI features, or anything that treats AI Overviews and AI Mode as separately measured. The report combines them, and no amount of confident dashboard design changes that.

How to get something useful out of it anyway.

Four things worth doing, given what the report is rather than what you wish it were.

1

Start the export habit now

There's no API, so exports are manual, and the report is your only historical record. Pull it monthly and keep the files. A year from now the trend line will be the most valuable thing you own in this space, and you can't reconstruct it retrospectively.

Where Search Console, Generative AI performance report
2

Compare AI impressions against organic impressions per page

Pages carrying high AI visibility but low organic visibility are the interesting ones. They're evidence of the decoupling this series keeps returning to: retrieval and ranking reward different things. Those pages tell you what your AI-facing content actually looks like.

3

Pair it with Bing's grounding queries

Google gives you pages without questions. Bing gives you questions. Neither is complete, but read together they get closer to the retrieval picture than either does alone. Verify in Bing Webmaster Tools if you haven't.

Where bing.com/webmasters, AI Performance report
4

Stop reporting AI visibility as a traffic number

The temptation is to convert impressions into an estimated click figure so it looks like the rest of the deck. Don't. The conversion rate you'd need doesn't exist publicly, and inventing one is how this industry ends up repeating numbers nobody can source.

It's easy to be cynical about a report that withholds four of the five things you want. I'd rather read it the other way round. Google has, for the first time, conceded that publishers are entitled to see whether their content is being used, and shipped that to every site for free. That's a floor, not a ceiling.

Use it for what it is. It proves presence, it identifies which pages carry it, and it gives you a baseline you'll be glad you started keeping. Everything beyond that still requires running the prompts yourself.

Frequently asked

Does Search Console show AI Overview data?

Yes, through the Generative AI performance report, which reached all sites on 31 August 2026. It reports impressions only, broken down by page, country, device and date. It combines AI Overviews and AI Mode into a single figure rather than separating them.

Can I see clicks from AI Overviews in Search Console?

No. The Generative AI report reports impressions only. Clicks from AI features are counted, but they're blended into your overall Search totals under the "Web" search type with no dimension that isolates them.

Is there an AI Overviews filter in the Performance report?

No, and there never has been. AI features are included in overall search traffic rather than reported as a separate search appearance. Any screenshot showing such a filter is not genuine, and any tool claiming to report AI Mode clicks from Search Console is inferring them.

Can I separate AI Mode from AI Overviews?

Not in Search Console. The report combines both surfaces into one impressions figure. Google Discover has its own separate generative AI report, but AI Overviews and AI Mode are not split from each other.

Is there an API for the Generative AI performance report?

Not at the time of writing. Export is manual through the Search Console interface, which is why it's worth starting a monthly export habit now. The report is your only historical record and you can't backfill it later.

How should I report AI visibility to stakeholders?

Report impressions and the page-level trend, and say plainly that click and revenue attribution isn't available from any platform. Converting impressions into an estimated traffic figure requires a click-through rate nobody has published. Inventing one is how unsourced statistics get into circulation.

Next in the series

If the platforms won't give you the numbers, the industry fills the gap with its own. Article 13 traces the most-repeated AI search statistics back to where they actually came from. Read Article 13 →

SOURCES ↓
  1. Google Search Console Help, "Generative AI performance report (Search)" - support.google.com/webmasters/answer/16984139
  2. Google Search Central, "AI features and your website" - developers.google.com/search/docs/appearance/ai-features
  3. Google Search Central, "Optimizing for Generative AI Features on Google Search," revised 10 July 2026 - developers.google.com/search/docs/fundamentals/ai-optimization-guide
  4. Microsoft Bing, "Introducing AI Performance in Bing Webmaster Tools," February 2026 - blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools

Let's talk about your visibility.

Invisible in AI answers, slipping in search, or both. Thirty minutes is enough to work out where to start.

Book a call →