For two years, appearing in an AI answer was a black box. In 2026, Google and Microsoft began to open it: official reports finally show where your site is cited by AI. But you need to know what those numbers say — and, above all, what they don't.
Google Search Console: the Search Generative AI reports
On June 3, 2026, Google announced the launch of new Search Generative AI reports in Search Console, with dedicated views for Search and for Discover. The goal: to understand a site's visibility within Google's generative AI features, namely AI Overviews and AI Mode.
Concretely, these reports expose:
- Impressions — how often your URLs appeared in AI features in Search and Discover.
- Pages — which URLs from your site appeared.
- Countries and devices — to place your visibility geographically and by device.
- Dates — tracking over time with hourly, daily, weekly and monthly granularity.
Google notes that this rollout is going to a subset of websites, so it can test and refine before general availability.
Bing Webmaster Tools: the AI Performance report
On its side, Microsoft expanded its AI Performance Report in Bing Webmaster Tools with four capabilities in global preview: Intents, Topics, Citation Share and Compare. Where the original report answered "where is my content cited in AI answers?", these additions answer "why", "on which topics", and "how does my presence evolve relative to other cited sources".
The Intents feature classifies grounding queries (the queries that trigger a citation) into categories: Informational, Commercial, Navigational, Learn and Solve, Research, Creation, Local… For an e-commerce brand, this shows whether it is strong on comparison or shopping intents rather than purely informational ones. Microsoft stresses one point: this is not a single ranking, but a first-party presence signal.
What these tools show — and what they don't
These reports are a real step forward: for the first time, a merchant can track AI presence with first-party data rather than third-party estimates. But they share a limit: they measure a symptom, not a cause.
| What they measure | What they don't tell you |
|---|---|
| Impressions in AI Overviews / AI Mode | Why an AI ignores you on a query |
| Citations and citation share (Bing) | Whether your product claims are proven and consistent |
| Intents and associated topics | Whether an AI crawler is blocked on your pages |
| Evolution over time | Which precise fix would lift a product page |
As the Ahrefs channel summarizes in the video below, "AI visibility" is not a single number: being cited and linked, mentioned without a link, recommended without being named, or fully invisible are different states that call for different actions.
Source: Ahrefs — "What Is AI Visibility? The 3 Types Every Marketer Needs to Know" (April 2026).
The llms.txt myth: what Google actually says
Faced with these metrics, a temptation returns: to find a "magic file" that would guarantee AI visibility. Google is explicit in its official guide to optimizing for generative AI: optimizing for AI search is optimizing for search, period. Its generative features (AI Overviews, AI Mode) rely on its existing ranking and quality systems, through RAG (grounding) and query fan-out.
The consequence: llms.txt is not a signal used by Google Search — it neither helps nor hurts Google rankings. The file may retain value for some third-party LLM crawlers, but selling it as a Google lever is a mistake. What counts for Google remains known and defensible: quality content, consistent structured data, up-to-date Merchant Center and Business Profile, readable and fresh pages.
Official position · Google
"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." Source: Google Search Central.
From measurement to cause: auditing the source
Measuring is necessary; fixing is what moves the numbers. Once Search Console or Bing show you where you lack visibility, you must diagnose why at the source. On Shopify, VerityScore's GEO audit reads your store the way an LLM and a buying agent would, then delivers a fix plan page by page: schema.org Product completeness, proof backing your claims, data consistency, AI crawler accessibility. Where a citations report stops at the observation, a source audit points to the action.
You can test a Shopify store's AI visibility for free (preview, no email) to connect the measured symptom to its fixable cause. It's the missing piece between "I know where I'm absent" and "I know what to fix".
Measurement, cause and UCP
This logic extends that of the Universal Commerce Protocol: visibility measured downstream (impressions, citations) reflects readability built upstream (readable catalog, proof, crawler access, consistency). Google's and Bing's reports are excellent dashboards; they do not replace the work on the source, which remains the only truly actionable lever to be understood, cited and recommended by AI agents.
Frequently asked questions
Do Search Console and Bing give the same numbers?
No. Each tool only sees its own surfaces: Search Console covers AI Overviews, AI Mode and Discover; Bing AI Performance covers Copilot, Bing and partner experiences. They are complementary views, not interchangeable.
Is "Citation Share" a ranking?
No. Microsoft specifies that its signals measure relative presence in citations, not a single rank. It's a share-of-voice measure, to be read over time, not a position to optimize mechanically.
Where should you start?
With the most blocking factor: verify that AI crawlers can access your pages, then that your structured data and proof are complete and consistent. Measurement then confirms the progress.