For two years, the hardest question in AI visibility has been a simple one: which questions is the AI answering without me? You could guess, you could sample prompts by hand, or you could pay a tool to guess for you.
On October 13, Microsoft started answering it directly. A new beta tab in Bing Webmaster Tools, called Topics Portfolio, lists the queries where Microsoft's AI surfaces cite other sites and not yours.
What Microsoft actually shipped
Topics Portfolio sits inside the AI Performance report, next to the existing Citations and Pages views. According to reporting from PPC Land and Search Engine Roundtable, it takes a site's grounding queries, the searches Copilot and Bing's AI summaries run behind the scenes to build an answer, and groups them into AI-classified topics.
Each topic gets three approximate numbers: Topic Size (total citations across all sites for that topic), Sites Cited (how many sites share those citations), and Queries (how many distinct grounding queries sit under it). You can look back one, two or three months.
Click into a topic and you get two tables. One is "Your Grounding Queries", the ones where you were cited. The other is "Opportunity Queries".
The gap list is the real news
Opportunity Queries are the ones where, in Microsoft's words, "other sites appear and yours does not." There is also a portfolio-wide view of the top ones.
This matters because it is first-party data. Until now, every "you are missing from these AI answers" list came from a third party running its own prompts and inferring the rest. This one comes from the engine that did the citing.
It is also built on grounding queries rather than user prompts. That is a useful distinction. A buyer types one long question; the engine breaks it into several shorter searches and cites what those searches return. The gap list shows you the searches, which is the level where a page either gets retrieved or doesn't.
What it leaves out
The tab is honest about being a beta, and the gaps are worth knowing before you plan around it.
- No names. You see how many sites are cited for a topic, not which ones. You learn that you are losing, not who is winning.
- No share figure. There is no share-of-citations or share-of-voice number. Opportunity volumes appear as ranges, not counts.
- Topics are built from your own presence. Two sites in the same market can see different topic lists, so you cannot compare screenshots with a competitor.
- Small sites may see little or nothing. Topics without enough activity are hidden, and only sites with "sufficient data" get the tab at all.
- Competitor brand names can show up as opportunities. Microsoft's own example is a rival's branded product query appearing in the list. There is no filter for it.
The classification is done by a model Microsoft does not name, and the company says it "can occasionally be imperfect." Each row has a thumbs up or down, though it is not stated whether voting changes what you see.
How this compares with Google
Google's generative AI report in Search Console launched on June 3 and reached every site by the end of August. It covers AI Overviews and AI Mode, and reports impressions, pages, countries, devices and dates.
It does not report queries. There is nothing in it equivalent to a grounding query, and nothing that tells you where a competitor was cited in your place.
So the engine with the smaller audience is now the one giving site owners the more useful report. That is a familiar pattern: Bing Webmaster Tools has often shown data Google held back. The practical upshot is that Bing's gap list is currently the best first-party proxy available for how answer engines in general carve up your topic, even if most of your buyers are asking somewhere else.
What to do with it this week
If you have a verified site in Bing Webmaster Tools, open AI Performance and check whether the tab is there. If it is:
- Export the top Opportunity Queries and strike out the competitor-branded ones. Those are not yours to win.
- For each remaining query, check whether you have a page that answers it plainly. Most gaps turn out to be a missing page, not a ranking problem.
- Note the date and the ranges before changing anything. A before-reading you did not write down is one you cannot compare against later.
Then remember what the list cannot tell you: it covers Copilot and Bing, not ChatGPT or Claude, and it never names the sites that took the citation.
The bigger picture
Answer engines are slowly starting to show their work. A year ago, citations were a black box; now one engine publishes a gap list and another publishes impressions. Neither tells you the whole thing, and neither tells you which competitor the buyer was pointed to instead of you. That part still has to be measured by asking the questions and reading the answers.
ClawWorld runs AI visibility work for B2B startups. Start with a free baseline report: 10 buyer questions, who ChatGPT and Claude name instead of you, and the pages they cite.