Bing's Topics Portfolio: what bing webmaster tools now shows about AI citations

What Bing Webmaster Tools added on 13 October#

Microsoft added a beta feature called Topics Portfolio to the AI Performance report in Bing Webmaster Tools on 13 October 2026. It groups the grounding queries where a site appears into broader thematic clusters, so an owner can see the shape of their presence across themes rather than one query at a time. Microsoft's own examples of themes are football, taxes and AI chatbots, and the view is personalised: the topics and the opportunity queries come from your site's own activity, not a generic list (the Bing Webmaster blog, the Webmaster blog archive).

The mechanics are simple once you log in. Inside any topic, three cards sit at the top: the Topic Size, your own citations within the topic, and the number of queries in it. Beneath them, the queries split into two tables. Your Grounding Queries lists what you already appear for; Opportunity Queries lists the strongest queries where other sites appear and yours does not, with volumes shown as ranges rather than exact numbers (Bing Webmaster blog, October 2026, Bing Webmaster Tools).

Two caveats are built into the product. The feature only shows if there is sufficient data for your site, and Microsoft asks owners to treat the numbers as directional signals for planning rather than a precise ledger (PPC Land's report). Citation volumes as ranges are a tell: this is a beta shaping up in public, and the sane posture is honest curiosity rather than a KPI built on evidence.

Search Engine Roundtable carried the launch the same day, crediting Krishna Madhavan of Microsoft, one of the three authors on the blog post (Search Engine Roundtable, the original post on X). The whole feature sits inside Bing Webmaster Tools behind a sign-in, which keeps the audience honest. For anyone who reads release notes for a living, that is the signal that matters: this is the same team that shipped the underlying report, still iterating in the open.

Grounding queries and the two halves of a topic#

Grounding queries are the searches that end in a citation. When Copilot or Bing's AI summaries answer a question and your page is part of the evidence, that query is a grounding query for your site. Until this week, the AI Performance report in Bing Webmaster Tools showed them one at a time, which is a list. Grouped into topics with competitor gaps beside them, it becomes a map, and the second half of the map is the part nobody has sold before.

The launch post frames the split cleanly: your grounding queries are what the site already owns, and the opportunity queries are what is still open, drawn from the same topic and marked by the absence of your domain (Bing's launch post). That is content planning with a ruler instead of a hunch, and it sits next to the Citations & Pages view rather than replacing it. Both live behind the same Bing Webmaster Tools sign-in.

Context is what makes the map worth reading. Bing's audience is smaller than Google's, and the panel counts disagree about exactly how much smaller: Statcounter put Bing at 9.89 percent of US search in September 2026 against Google's 85.19, while Datos, a Semrush company, counted Bing's US share peaking at 5.33 percent in December 2025 and settling back to 3.2 to 3.8 percent in early 2026 (Statcounter's US share data, PPC Land on the Datos report).

The measurement gap this beta starts to fill#

Ask marketers what is hardest to measure, and AI visibility is near the top of the list. In a Semrush survey of 481 marketers, 45 percent named visibility in AI-generated answers among their hardest metrics, behind only the impact of AI on pipeline and revenue at 49 percent. Forty percent still check their AI presence by typing prompts by hand, and only 32 percent call AI search visibility clearly measurable (the Semrush study). A free baseline inside Bing Webmaster Tools is a useful corrective to that fog.

Academic work is catching up from the other direction. An observational study of 900 US adults found that about 18 percent of Google searches produced an AI Overview, and that browsing sessions ended on 26 percent of pages with an Overview against 16 percent without one (the AI Overview click study). A second paper on ChatGPT citations argues that generative engine optimisation should split retrieval visibility from citation visibility, and that run-to-run variance is large enough to plan for (the ChatGPT citations paper). Both are early evidence, and both point the same way: the questions are real, and the scoreboards are young.

That is the honest frame for this beta in Bing Webmaster Tools. Generative AI visibility is not a solved metric with a standard currency, and anyone selling a single number for it is selling certainty the field does not have. What Microsoft shipped is narrower and more useful: a list of the themes where a site is cited, a list of the queries where it is not, and a set of ranges that make no promises it cannot keep.

How to use the Topics Portfolio this week#

The beta rewards a specific kind of work: slow, comparative, unglamorous, and exactly the sort of thing an answer engine notices. Here is the order we would run it for a client site, before any budget decisions follow.

Start by checking presence. Sign in to Bing Webmaster Tools, open AI Performance, and see whether the Topics tab appears at all; Microsoft gates it on sufficient data, so a missing tab is itself a finding about the site's AI footprint (Bing's getting-started notes). Then pick a window of one, two or three months and start where your citations already concentrate, working down the themes rather than across them.

Next, treat the two tables as a single pipeline. The Grounding Queries table tells you which existing pages are earning citations and for what language, which is the raw material for internal linking and consolidation. The Opportunity Queries table is a brief list: each row is a question where the answer exists, a competitor is holding it, and no page of yours covers it. Ranges, not ranks, so judge by theme rather than by a single row.

Finally, write down a baseline before you touch anything. A dated export of the topics, your citations and the pass-through to your own analytics is all it takes; the value comes when the next window opens and the movement is legible (PPC Land's walkthrough). For teams already running a generative ai visibility programme, this is a free rung on the ladder that paid tools occupy higher up (SEO and GEO services). None of it needs noise; the scent does the work.

One habit keeps the work honest: pair the crawlable with the human. Answer engines cite pages that read like sources, so the opportunity list should feed the same editorial system that feeds everything else, not a separate pile of AI-shaped pages. The searches behind the numbers are people asking questions, and that never changes. A fox reads the hedgerow, not one branch of it.

The honest read on the portfolio#

Solid facts: Microsoft shipped Topics Portfolio as a beta inside AI Performance on 13 October 2026; it groups grounding queries into themes, adds opportunity queries where rivals are cited instead, and shows volumes as ranges; it only appears when a site has sufficient data. The beta is free inside Bing Webmaster Tools, which is the point of it. The measurement context is equally solid: 45 percent of surveyed marketers struggle to measure AI answer visibility, Bing's share numbers differ by panel, and academic work shows AI Overviews change click behaviour. Uncertain: how stable the topic groupings are across windows, whether the ranges tighten after beta, and what Microsoft adds next. We will not pretend to know.

The playbook in one line: export the baseline this week, work the opportunity queries through your existing editorial process, and re-read the portfolio next month against your own analytics rather than against a screenshot. The scoreboards will keep changing; the trail worth keeping is the dated one you can compare. That is how the den keeps warm in autumn: dated trails, no guesses.

For the discovery side, our SEO and GEO services practice covers the full stack from technical foundations to answer-engine work, and the content marketing team builds the editorial engine the opportunity list needs. When you want the audit run against live data, talk to folkfox is the fastest way in.

Frequently asked questions#