What AI Performance is, and what it shows

Microsoft built the first free first-party dashboard for AI citation, plugged it into the existing Bing Webmaster Tools account most WooCommerce stores already have, and made it visible to anyone who logs in.

Microsoft will tell you what Copilot cites. Microsoft will not tell you what those citations earn you. That’s the binary every WooCommerce store needs to understand before opening the dashboard.

AI Performance shipped to public preview on February 9, 2026, sitting inside the same Bing Webmaster Tools account WooCommerce stores have used for organic Bing reporting for years. Sign in at bing.com/webmasters/aiperformance and the panel populates with five metrics: total citations across Copilot and Bing AI, average cited pages per day, citation counts by page URL, the grounding queries Bing used to retrieve your content, and citation trends over time (Bing Webmaster blog, February 2026).

This is the first first-party AI citation dashboard from any major search platform. Google Search Console has no equivalent.

The context for the launch is the same context every other AI announcement in 2026 is operating in: AI-driven traffic to retail rose 187% from January to December 2025, with agentic browser traffic up roughly 8,000% year-over-year (HUMAN Security Quadrillion Report, 2026). Citation traffic is the fastest-growing input to retail discovery, and Microsoft has decided to give every site owner a free look at the demand signal feeding their content into that surface.

Grounding queries: the most underrated panel

The grounding queries panel sounds like an SEO curiosity. It is actually the most useful intelligence in the dashboard, because it shows exactly what Copilot thinks your store sells.

When a shopper types “I need running shoes for flat feet under $100” into Copilot, Bing doesn’t run that string against the index. It reformulates the prompt into one or more retrieval queries — the grounding queries — and runs those. The grounding query for that shopper might be “best stability running shoes under 100 dollars 2026” — and AI Performance shows you every grounding query that surfaced your domain.

For a WooCommerce store this is a category audit. If your /product/runner-x page is being surfaced for “best stability running shoes under 100 dollars” but never for “running shoes for flat feet,” then Copilot has classified your page on the supply side (price + product type) but not on the demand side (the foot-condition language the shopper actually uses). That gap is content work, and it is now measurable for the first time.

The other thing the grounding query panel exposes is what Copilot will not retrieve. If your store sells trail-running shoes but no grounding queries with “trail” surface your domain, you have a content surface visible in the dashboard as a zero — and the zeros are usually where the gap is.

You may be interested in: Microsoft Activate Is May 19 — Your UET Tag Just Became a Second AI Engine

The four features previewed on April 27

The preview at SEO Week NYC was Microsoft showing the next two quarters of the roadmap. None have shipped, none have firm dates — but their existence reshapes what AEO measurement is about to mean.

At SEO Week NYC on April 27, 2026, Microsoft Principal Product Manager Krishna Madhavan previewed four queued features (Search Engine Land, April 2026):

1. Citation Share. For each grounding query that cited your domain, what percentage of total citations came from you versus competitors? This converts AEO performance into a public scoreboard. When it ships, any logged-in user can see how your store stacks up on the queries you both compete for.

2. Grounding Query Intent. Fifteen predefined labels, including Learning, Informational, Navigational, Research, Comparison, Planning, Conversational, and Content Filtered. The Content Filtered label is the one to watch — it tells you when Copilot wanted to retrieve from your domain but the content didn’t pass a safety or quality filter.

3. Grounding Query Topic Labels. Topical categorisation of grounding queries — useful for content planning, especially identifying topic clusters where your domain has zero coverage versus where it dominates.

4. GEO Recommendations. Algorithmically generated suggestions on content and structure gaps. Microsoft’s framing was that these will read your existing content against the grounding queries you didn’t surface for and recommend the structural or semantic changes most likely to close the gap.

The click-data gap Microsoft has chosen not to fill

Every Microsoft AI dashboard in 2026 — AI Performance, Clarity AI Visibility, Microsoft Advertising — shows citations and stops at citations. The click loop is a structural omission, not a temporary one.

Here’s the thing: CXL and Momentic independently flagged the same gap in their 2026 testing of AI Performance. Citation visibility without click data is a half-loop. Microsoft can tell you a Copilot answer cited your /product page; Microsoft cannot tell you whether the shopper who saw that citation visited, asked a follow-up, bought from a Brand Agent inside Copilot, or walked away.

Microsoft dashboard	Shows	Hides
Bing Webmaster Tools AI Performance	Grounding queries, citation counts, trends	Click-through, conversion
Microsoft Clarity AI Visibility	Page-level citation, competitor share, no-click influence	Conversion downstream
Microsoft Advertising reporting	Paid placements, UET conversions	Organic AI citations
Google Search Console	Traditional search performance	AI citations entirely

Third-party tools that claim to estimate click-through from AI citations are guessing — they extrapolate from impression counts and historical CTR curves that don’t apply to AI surfaces. The only honest answer to “did this citation earn anything” comes from joining citation data to your own conversion stream.

Citation visibility without click data is a half-loop, and no first-party Microsoft tool is going to close it. The closing has to be done on your side of the wire.

The WooCommerce instrumentation pattern

The fix is the same fix every server-side tracking conversation has been pointing at since the start of the AI Max era: capture the inbound signal before the browser drops it, write it to a warehouse, join it to the order.

When a shopper does click through from a Copilot answer, the inbound URL carries a referrer header from a Bing or Copilot subdomain, and increasingly a URL parameter Microsoft has been rolling out in pilots. Client-side tools see those signals for the duration of one page render before consent banners, ad blockers, and Safari’s cookie ceiling start eating them.

A server-side capture layer reads referrer and URL parameters at the WordPress request level, before browser data loss kicks in. The visit gets stamped with an AI-surface dimension — Copilot, Bing AI, Brand Agent, organic — and that dimension carries through to the WooCommerce order record. In BigQuery, the citation counts from AI Performance and the order counts from the server-side stream join on URL plus week.

Transmute Engine™ is the piece that does this for WooCommerce: a first-party Node.js server that runs on your subdomain (e.g., data.yourstore.com) and captures every session including AI-referrer fingerprints. The inPIPE WordPress plugin reads WooCommerce hooks and batches events to the Transmute Engine server, which routes simultaneously to GA4, Facebook CAPI, Google Ads, and BigQuery. AI Performance reports the grounding-query side; Transmute writes the conversion stream that the grounding-query data joins against.

You may be interested in: Microsoft Clarity Now Shows Which WooCommerce Pages Copilot Cites

Why this is also your ChatGPT and Google AI Mode roadmap

Grounding queries are a Microsoft artefact, but the demand they reveal is platform-agnostic. The same shopper language drives retrieval inside ChatGPT and Google AI Mode — your AI Performance dashboard becomes a free proxy roadmap for both.

CXL’s testing through 2026 showed that grounding-query patterns inside Bing AI correlate strongly with retrieval behaviour inside ChatGPT and Google AI Mode — same shopper language, same retrieval logic, slightly different model outputs. The grounding queries that surface your domain in AI Performance are a close proxy for the queries that surface your domain inside competing AI surfaces, even though ChatGPT and Google won’t show you their version of the data.

That makes AI Performance a free AEO content roadmap. Pick the grounding queries where you have low citation count or where Citation Share (when it ships) shows a deficit, build content that explicitly addresses that intent, and you’ve improved your retrievability across every major AI surface — not just Bing.

Key Takeaways

• AI Performance is live and free: in public preview at bing.com/webmasters/aiperformance since February 9, 2026 — first first-party AI citation dashboard from any major platform.
• Grounding queries are the panel that matters: they show how Bing reformulates shopper language, which is the gap between what you sell and what AI thinks you sell.
• Four new features previewed April 27: Citation Share, Grounding Query Intent labels, Topic Labels, and GEO Recommendations — all queued, none shipped.
• No click data, by design: Microsoft has declined to expose Copilot click-through; the citation-to-revenue join has to be built on your side.
• Server-side capture closes the loop: AI-referrer dimension stamped at the WooCommerce hook, written to BigQuery, joined to grounding-query data in SQL.

Frequently asked questions

References

• Bing Webmaster blog. “Introducing AI Performance in Bing Webmaster Tools Public Preview.” blogs.bing.com, February 2026.
• Search Engine Land. “Bing Previews AI Citation Share For Webmaster Tools.” searchengineland.com, April 2026.
• HUMAN Security. “Quadrillion Report 2026.” humansecurity.com, March 2026.
• Google Search Central. “Documentation.” developers.google.com, 2026.

Grounding-query insight is free; the revenue join lives in your warehouse. See how Seresa writes WooCommerce orders plus AI referrers into BigQuery side by side.