Measurement

Bing now lists the AI queries where rivals are cited and you aren't. It still won't name them.

By Arnav Mukherjee, founder of TofuBofu · October 22, 2026

In September I wrote that Bing's Citation Share tells you your size and withholds the shape of the field. You learn you hold 8 percent of the citations for a grounding query. You don't learn whether the other 92 percent belongs to one rival who owns the topic or to forty sites with a sliver each, and those two situations need opposite responses.

Five weeks later Microsoft shipped half an answer. On 13 October three people from Microsoft Bing announced Topics Portfolio, a beta tab in the AI Performance report. It tells you how crowded each topic is and lists the queries where other sites get cited and you don't. It still won't tell you who those sites are. Here's what it shows, how a B2B team should work it, and where it stops.

What did Microsoft ship on 13 October?

Topics already existed in the June release. The new tab sits next to Citations & Pages and turns it into a portfolio. In Microsoft's words, it "groups the grounding queries where your site appears into broader thematic clusters", personalised to your site rather than drawn from a generic list. You pick a one, two or three month window.

Each topic row carries four columns. Here they are with Microsoft's own definitions:

• Topic. A broad cluster of related grounding queries where your site appears.
• Topic Size. The approximate total citations across all sites for the topic in the window, Microsoft's measure of how much AI Search demand it represents.
• Sites Cited. The approximate number of sites appearing for queries in the topic, "which indicates how crowded the topic is".
• Queries. The approximate number of distinct grounding queries in the topic.

Below the table sits Top Opportunity Queries: "the strongest queries across your whole portfolio where other sites appear and yours does not." Click into a topic and you get three cards (Topic Size, your own citations in it, query count) and two tables. Your Grounding Queries is what you already hold. Opportunity Queries is what's still open inside that topic.

Every row has a thumbs up and thumbs down. Microsoft says those votes go straight to the team tuning the classification, so use them.

Which part closes the gap Citation Share left?

Two columns do real work, and they answer different halves of the question I left open in September.

Sites Cited gives you the shape of the field, at topic level. A topic with a large Topic Size and few sites cited is held by a handful of sources. You won't displace them with one blog post; you need to become one of the few. A topic with many sites cited is fragmented, and a single well-structured page can earn a place. That's the "one rival or forty" distinction, finally visible, even if only per topic and only as an approximation.

Opportunity Queries give you a gap list from the engine's own data. Until now every gap list in this category came from the outside: a tool asks an engine a question and records who got named. This one comes from what Microsoft's AI surfaces actually grounded on. That's a different kind of evidence and a stronger one for the surface it covers.

What hasn't moved is the line Microsoft drew in June: Citation Share "does not expose competitor domains". Topics Portfolio keeps that line. Sites Cited is a number. Opportunity Queries say "other sites", plural and unnamed. You can now see that a topic is crowded and that you're missing from a query, and you still can't see whose page you're up against.

How should a B2B team work it in one sitting?

Microsoft's own advice is to start with the topics where your citations are already concentrated, then work down the opportunity queries. That's right, and for a B2B firm I'd make it sharper. This order is my read, not Microsoft's.

1. Set the window to three months. A B2B site has fewer citations than a publisher, and the short windows will be thin.
2. Find the topics that match what you sell. The classifier will also cluster your blog's side topics. Ignore those first; a citation on a buying topic matters more than any count on a tangent.
3. Sort those by Sites Cited. Low count with real Topic Size means a few sources own it. Plan to become one of them, with one definitive page, before you write anything else.
4. Map every opportunity query to a page. If the page exists, make its answer self-contained so it can be lifted as a passage. If it doesn't, you've just written a content brief.
5. Drop rival brand queries. Microsoft warns that branded queries can appear as opportunities. You won't get cited for another company's name, so don't spend a page on it.
6. Vote on the misfires. A topic that doesn't describe your business gets a thumbs down. The classification is AI-generated and Microsoft says it can be wrong.

Keep one discipline from the June release: these rows are grounding queries, the searches the system wrote for itself, and not what your buyer typed. Microsoft defines the idea through grounding, "the source material and web evidence the system uses to support and cite its response." Treat an opportunity query as evidence about how the machine looks for sources. Don't paste it into a keyword plan as if a buyer said it. We made the same point about which queries are worth tracking.

What can it still not tell you?

Four limits, three of them stated by Microsoft itself, which is why I trust the feature.

• Who's winning. No competitor domain in Citation Share or Topics Portfolio. For a B2B firm that's the question that decides the next page you write, because the answer to "who's cited instead of us" tells you what the engine thinks a good answer looks like.
• Precise numbers. Topic Size, Sites Cited and Queries are approximate, and opportunity volumes are "deliberately shown as ranges". Microsoft's own phrase: "directional signals for planning, not as a precise ledger."
• Small sites. Topics with insufficient site presence or ecosystem activity aren't shown, so "smaller sites may see a shorter list or may not even see this feature." Plenty of B2B services firms will open the tab and find little or nothing. That's not proof you're doing fine; it's the absence of a reading.
• Every other engine. The AI Performance report covers "Microsoft Copilot, Bing, and select partner AI experiences". In our own paired corpus of 34 questions answered by both Google AI Mode and Bing Copilot, across 4 reports and 3 brands, the two engines didn't share a single cited domain on 12 of the 34. Those brands scanned because they already suspected a problem, and one of the three is a consumer brand, so read it as a count from a small sample. It's still enough to say a gap list from Copilot won't match Google's.

So run both kinds of measurement. The engine's own logs are the only place your true citation picture exists, and Bing hands you a bigger piece of it than Google does: Search Console's generative AI report groups data by pages, countries, dates and devices, with no query dimension at all. An outside scan is the only way to see who's named instead of you, on the engines your buyers use most. Microsoft keeps moving toward the first. It still hasn't done the second. Our September breakdown of Citation Share covers the metric this tab is built on.

Frequently asked questions

Sources and further reading

• Microsoft Bing: Introducing Topics Portfolio (Beta) in AI Performance, by Krishna Madhavan, Saral Nigam and Vivek Jain, published 13 October 2026, read at source 22 October 2026: the feature, the four column definitions, Opportunity Queries and the beta limitations quoted above
• Microsoft: New AI Visibility Insights in Bing Webmaster Tools, published 16 June 2026, read at source 7 September 2026: Citation Share, the scope line, the grounding definition and the "does not expose competitor domains" limitation
• Google Search Console Help: Generative AI performance report, read at source 22 October 2026: the four dimensions (Pages, Countries, Dates, Devices), impressions as the metric, and no query dimension
• Our own paired citation corpus: 34 questions answered by both Google AI Mode and Bing Copilot, across 4 completed reports and 3 brands, from scans run between 24 June and 24 August 2026. Every one of those brands ran a scan because somebody already suspected they were missing from AI answers, so the sample is selected on the problem being measured. One of the three brands is a consumer brand