On 13 October 2026, Microsoft added to Bing Webmaster Tools, its free console for site owners, a beta tab called Topics Portfolio in the AI Performance report. It groups by topic the queries where a site is cited in the answers of Copilot, Microsoft's AI assistant, and in Bing's AI summaries, measures the size and the competition of each topic across all sites, and lists opportunity queries where other sites are cited while the site under review is not. It is the first official tool from a search engine that tells a site which questions asked of an AI it is missing from. Bing warns that its figures are approximate and that small sites may see nothing. For a marketing team, the point is to turn that list into a content plan, after sorting it.
The facts
- In its post of 13 October, Bing describes a Topics Portfolio tab next to Citations & Pages in the AI Performance report, visible only when the site has enough data. Each topic carries three measures: Topic Size (total citations for the topic across all sites), Sites Cited (number of sites cited) and Queries (number of queries). Below the table, Top Opportunity Queries lists the strongest queries across the portfolio where other sites appear and yours does not, with volumes shown as ranges.
- The AI Performance report has existed since February 2026: total citations, cited pages, a sample of grounding queries, a timeline, and no click data at all. In June, Bing added Intents, Topics, Citation Share and Compare; those first Topics only grouped queries where the site was already cited, and Citation Share, presented as an observational metric, does not expose competitor domains.
- Bing states four limits: AI-generated topic classification that can be imperfect, topics hidden when data is insufficient, competitor brand queries that may appear as opportunities, approximate figures to be read as directional signals for planning. Search Engine Journal noted on 14 October that the AI Performance help page did not yet document the tab and that no date is given for leaving beta. According to StatCounter, Bing accounted for 5.29% of searches worldwide in September 2026, Google for 89.94%.
Four columns and an opportunity list: what the report displays
Everything rests on grounding queries: the searches Bing's AI runs on the web before writing its answer and choosing the sources it cites. The AI Performance report has shown them for the site's pages since February; Topics Portfolio groups them into topics, football, taxes or AI chatbots in Bing's examples, and compares each topic with the whole set of sites cited for it.
A topic's page shows three indicators (topic size, the site's citations within the topic, number of queries) and two tables: Your Grounding Queries, the topic's queries where the site is already cited, and Opportunity Queries, those where other sites are cited and it is not. The analysis window runs from one to three months. Every row carries a thumbs up and a thumbs down that go to the team in charge of classification.
| Measure | What it indicates, according to Bing | What we do with it |
|---|---|---|
| Topic | A cluster of related grounding queries where the site appears | Check that the topic matches the offer; flag errors with the thumb |
| Topic Size | Total citations for the topic over the period, across all sites: AI search demand | Rank topics against each other, without forecasting traffic |
| Sites Cited | Approximate number of sites cited for the topic: its level of competition | Spot large topics with few contenders |
| Queries | Approximate number of distinct queries in the topic | Gauge the breadth of the subject to cover |
| Opportunity Queries | Queries where other sites are cited and yours is not, volumes as ranges | Sort, then run each query in Copilot to see who is cited |
From February to October, the report moves from observation to planning
The February version answered one question: where is my content cited? It counted citations, listed cited pages and a sample of grounding queries, and drew the timeline. Bing already specified that these figures indicated neither ranking nor placement within the answer.
The June update added the why: the intent behind each query (information, purchase, learning, local), topics limited to queries already cited, the site's share of citations for a query, calculated over all citations shown for that query, and the comparison between two periods. Bing called Citation Share an observational metric: it does not expose competitor domains, does not represent traffic share and assigns no quality score.
Topics Portfolio adds what the report did not show: the overall shape of the site's presence in AI search, the weight and competition of each topic across all sites, and specific queries where the site is absent. The report used to describe a state; it now points to gaps, which makes it a planning tool for GEO, the optimisation of visibility in AI-generated answers.
The limits Bing states itself, and those it leaves open
Bing ships the feature with its beta label and asks users to treat the figures as directional signals for planning. Topics are classified by an AI and a query may land in a box the publisher would have named differently. Topics with insufficient site presence or insufficient ecosystem activity are hidden: a small site sees a short list, or no tab. Competitor brand queries may appear among the opportunities, Bing gives the example of a running shoes site that would see a rival brand's query. Opportunity volumes are deliberately shown as ranges.
The post does not say whether the opportunity tables name the cited sites; the June Citation Share metric does not. It gives no threshold for sufficient data, no date for leaving beta and no indication about classification quality outside English. No click or traffic data comes with the citations: a citation in an AI answer is not a visit, and measuring the effect remains a job for your own analytics tools.
Six steps to turn the opportunity list into a content plan
Bing gives a reading order in three moves: pick a one- to three-month window, start with the topics where citations are already concentrated, then work down the opportunity queries. The steps below extend that method with our own practice; only the starting point and the feedback thumbs come from Bing.
- Access. Verify the site in Bing Webmaster Tools, then open AI Performance and look for the Topics Portfolio tab. If it does not appear, Bing lacks data on the site.
- Starting point. Read first the topics where the site is already cited, over the three-month window, the longest offered, to smooth out variations.
- Reading. Cross Topic Size with Sites Cited: a large topic with few contenders and a large saturated topic do not call for the same effort. Both figures serve to rank topics against each other.
- Sorting. Remove from the opportunity list the competitor brand queries, which Bing admits can surface, and those outside the offer. What remains is a set of questions to which Bing's AI found an answer somewhere other than your site.
- Testing. Run each retained query in Copilot and in Bing, note which pages are cited and in which format (guide, comparison, FAQ, product sheet), then create or enrich the matching page.
- Measurement and feedback. Track the effect in your analytics tools, since the report counts no clicks, and flag classification errors with the thumbs, which Bing says steer its fixes.
Bing's weight remains modest: 5.29% of searches worldwide in September 2026 according to StatCounter, against 89.94% for Google. The signal is worth more than that scope, and this is our reading: few official, free tools show what an AI engine searches for on a topic and how many sites it cites, and the list of missing queries informs demand even for those who target Google first. On the Luxembourg market, multilingual and populated with modest-sized sites, many publishers will see only a handful of topics, and Bing says nothing about classification in French or German. Since verifying a site is free, it costs only the time of signing up.
AIxH's view
Our AI visibility audits rely on answers recorded in ChatGPT, Perplexity, Gemini and Copilot, question by question, and on what Search Console shows of a site's presence in Google's AI since 31 August, without the queries. Topics Portfolio brings first-hand data from the Microsoft side, with the queries and the gaps, to be read with the same caution: the tool offers leads, the sorting and the decisions remain human. To find out on which questions AI engines cite your competitors, our GEO agency in Luxembourg crosses these reports with recorded answers and draws up a prioritised content plan.
