Bing Webmaster Tools AI Performance report: how to read it

You open AI Performance in Bing Webmaster Tools from the menu of a verified site, or go straight to the Bing Webmaster Tools AI Performance address, bing.com/webmasters/aiperformance (Search Engine Roundtable published it the day after launch).

From scratch, it's four steps:

1. Sign in at bing.com/webmasters with a Microsoft, Google or Facebook account.
2. Add the site. Importing it from Google Search Console verifies it automatically; otherwise add it by hand and verify ownership with an XML file, a meta tag or a CNAME record.
3. Allow about 48 hours (Microsoft's estimate) for data to appear.
4. Pick the site, open AI Performance, and switch the list between grounding queries and pages.

AI Performance metrics are citations, cited pages and grounding queries, plus a page table and a timeline built from the same data.

Microsoft's AI Performance help page defines each one.

Total Citations and Average Cited Pages

Total Citations and Average Cited Pages are the volume metrics. Total Citations counts how many times your content was shown as a source in AI answers during the date range (with nothing on where in the answer), and Average Cited Pages is the average number of unique pages cited per day, so one tells you how loud you are and the other how wide. The Pages view lists citations URL by URL, which makes it your top pages report.

Grounding queries

Grounding queries are the searches behind your citations, and Microsoft defines them as "the key phrases the AI used when retrieving content" that ended up cited. They're short, grouped versions of those searches and NOT the user's prompt, the count beside each one is how often you were cited for that phrase, and one phrase can map to several pages. Since March 2026 you can click a grounding query to see the pages it cited, or a page to see its grounding queries, though not both at once.

Intents, Topics, Citation Share and Compare are the four preview features Microsoft added to AI Performance on June 16, 2026, and three of them slice the grounding query data.

Intents labels each grounding query with the need behind it, from Informational and Commercial to Comparison, Planning and Local (fourteen example labels, which Microsoft admits can misfire). Topics groups related grounding queries into themes (Microsoft's example is Solar Energy), and seeing the pages behind a topic means exporting and filtering the data yourself. Citation Share is your site's percentage of all citations shown for one grounding query, share of voice for a single search, without naming the other sites. Compare overlays an earlier period on the chart.

Citation Share is the one I'd watch. Microsoft reads a falling share as citations spreading across more sources; I read it as other pages winning that search, fixed the way you fix any search you're losing.

The Bing AI Performance report doesn't show clicks, traffic or rankings: Microsoft's help page says a citation represents no clicks, traffic or engagement, so you can be cited all month without a single visit.

It doesn't show prompts, full answers or why a page was picked, Citation Share hides the other domains, and it says nothing about ChatGPT, Gemini, Claude, Perplexity or Google's AI features, none of which Microsoft names as a source.

To use Bing grounding queries for SEO, treat them as a list of searches you already win: each is a search an AI ran against Bing's index before citing your page, so protect those, fix the ones landing on the wrong page, and chase the neighboring searches you don't win yet.

This is query fan-out with the lid off. Google says AI Overviews and AI Mode may run several related searches to build one response, and grounding queries are Microsoft's version of those searches, reported back to you. The model writes the answer, but a search index decides what it reads. That's SEO.

My monthly routine would be:

1. Export grounding queries with their pages and sort by citations.
2. Search the top phrases in Bing and see what ranks around you.
3. Find phrases cited to the wrong page and make the right page the stronger match with its title, headings and internal links.
4. Use Topics and Intents to spot themes where you're cited for informational searches but missing from commercial ones.

Bing ONLY shows Microsoft's side, though. To see what other assistants search, run the same prompt through our query fan-out tool: it shows the searches ChatGPT, Claude and Gemini actually run (one run, or three to spot the ones that repeat) and your site's position in Google's top 10 for each. Both lists belong in an AI visibility audit.

The Bing AI Performance report vs Google Search Console question comes down to one difference: Bing shows you the searches behind your AI citations, and Google doesn't.

Google's page on AI features and your website says AI Overviews and AI Mode appearances count toward your overall Search Console traffic, inside the Performance report's Web search type, and that report has no Search appearance filter for them. On June 3, 2026 Google added a separate generative AI performance report (every site since August 31) that isolates those impressions by page, country, device and date, with no query dimension and no clicks, a gap Barry Schwartz flagged on Search Engine Roundtable the next day.

Neither report gives you clicks you can pin on an AI answer, and for content work Bing's wins, because you can act on a query and you can't act on an impression count. To see the two side by side, page by page, drop both exports into our AI performance report.

Bing AI Performance data is a sample, not a log, which explains most of the confusion about it.

Microsoft's help page says low activity may not surface, totals can differ between views, and pages and grounding queries may be sampled over slightly different windows, so the same phrase and page can show different counts depending on which you filter first. Microsoft calls that expected. Compare within one view and read the trend, not the digit.

The bigger miss is what the report confirms. It only reflects content eligible for indexing, and Microsoft's own reasons a page never gets cited (less relevant to what people ask, thin, outdated, or outperformed by other sources) read like a ranking checklist. A page that isn't indexed can't be cited and a page that loses the search won't be, so I'd start an empty report with indexing and Authority (indexing trouble is usually an Authority problem) before touching the copy.

This Bing AI Performance report FAQ covers cost, ChatGPT, history, exports and empty reports.

Is the Bing AI Performance report free?

The Bing AI Performance report is free: it's part of Bing Webmaster Tools, which Microsoft describes as free, and all it asks is a verified site.

Does the AI Performance report include ChatGPT?

There's no sign the AI Performance report includes ChatGPT: Microsoft names Copilot, AI-generated summaries in Bing and unnamed partner integrations, and ChatGPT appears nowhere in its documentation. Microsoft does say its grounding powers nearly every major AI assistant, but the report never says which partners count, so treat ChatGPT separately (here's what search engine ChatGPT uses).

How far back does AI Performance data go?

AI Performance data goes back as far as the dashboard's date ranges, which Microsoft lists as 7 days, 30 days and 3 months, plus custom ranges inside that history. Export it monthly if you want a longer record.

Can you export Bing AI Performance data?

You can export Bing AI Performance data as CSV or Excel (grounding queries, page-level citations and the time series), filtered the way your view is. There's no documented API yet: Microsoft's Fabrice Canel said so at launch, and the API reference still had no AI Performance methods when we checked in October 2026.

Why does AI Performance show no grounding queries for my site?

AI Performance shows no grounding queries when there's too little citation activity in the date range to surface, and Microsoft says that's expected, not a penalty. Widen the range first, then check that your pages are indexed in Bing and rank for the searches an AI would run.