Bing Topics Portfolio: how to find the AI search queries where competitors are cited and you are not (and what a test of 972 of them found)

What is Bing Topics Portfolio?

On 13 October 2026 Microsoft’s Krishna Madhavan, Saral Nigam and Vivek Jain announced Topics Portfolio, a beta view inside the AI Performance report in Bing Webmaster Tools. It takes the grounding queries where your site already appears in AI answers and groups them into broader themes. Microsoft’s own examples are football, taxes and AI chatbots.

Grounding queries need a sentence of explanation, because they are not what people type. In the February launch post, Microsoft defined them as the key phrases the AI used when retrieving content that was then referenced in an AI-generated answer. In other words, they are the searches Copilot runs on the user’s behalf before it writes its answer and picks its sources.

The new part is the second half of the view. Until now, AI Performance only showed where you were cited. Topics Portfolio adds Opportunity Queries: in Microsoft’s words, "the strongest queries across your whole portfolio where other sites appear and yours does not". As PPC Land noted, it is the first time the report has shown queries where other sites are cited and yours is not. PPC Land and Search Engine Roundtable both covered it on launch day.

How to read the Topics Portfolio report, column by column

Open Bing Webmaster Tools, go to AI Performance and look for the new tab next to Citations & Pages (in the interface it is labelled Topic Portfolio, without the "s"). If it is missing, Microsoft says there is not yet enough data for your site. You can choose a one-, two- or three-month window.

Microsoft is unusually frank about the limits. The announcement says topic classification "can be imperfect", that topics with too little activity are hidden, so smaller sites may see few or none, and that branded queries can appear as opportunities. Its summary line is worth pinning above your desk: "Treat the numbers as directional signals for planning, not as a precise ledger."

Paralax’s analysis adds a useful caution on Sites Cited. A high count does not mean a topic is closed. What matters is how concentrated the citations are among the leading sources, and the column cannot show that. Do not skip a topic just because many sites appear in it.

Are Bing’s opportunity queries any good? What a test of 972 found

Four days after launch, Polish SEO and GEO consultant Ewelina Podrez-Siama published the first independent audit of the feature. She pulled every site in her Bing Webmaster Tools account, 51 in all, her own and clients’. 33 had citations between July and September 2026; 21 of those showed topics. From those 21 she exported 972 opportunity queries and assigned each one a language by hand.

The findings are specific. The opportunity lists are genuinely personalised: of 75 pairs of sites sharing the same topic, 60 pairs had not one query in common. But they are not filtered by language or market. A Madeira travel guide was offered Vietnam and Laos; a keto recipe blog was offered pancakes and tiramisu in English; a French tile-tools shop, used as a control, got "grout" and "epoxy grout" in English rather than French. Her conclusion is that the problem affects any language other than English, not just Polish.

Two more findings matter for anyone planning content from this report. The lists move: the median overlap between the September-only list and the full-quarter list was 48%, so a one-month window can change under you. And topics cover only part of your AI visibility: a median of 28% of each site’s citations fell inside its topics. The Citations & Pages tab remains the complete picture.

The most useful finding was about a site’s own grounding queries. For the fashion retailer, which has eight language versions, Italian queries produced 42% of the citations inside its topics, although the home page is Polish. That tells you which language version is actually earning AI visibility, and it also showed that Copilot was answering returns-policy questions by citing the shop’s own page. If that page is out of date, the AI repeats it to customers.

Bing AI Performance vs Google Search Console’s AI reports

Google now has its own view. On 3 June 2026 Hillel Maoz and Moshe Samet introduced generative AI performance reports in Search Console, with a note added later that they reached all websites worldwide on 31 August. The two tools measure different things, and neither covers ChatGPT or Perplexity.

The practical reading: Google tells you where your pages show up in its AI features; Bing tells you why (the queries) and, now, where you are missing. Use Bing’s query data to inform content decisions that also apply to Google, and Google’s impressions data to judge whether those changes moved anything in the larger ecosystem. For ChatGPT, Gemini and Perplexity, you still need a fixed set of prompts that you run on a schedule.

How to use opportunity queries without wasting a content budget

1. Check you can see the report at all. Verify your site in Bing Webmaster Tools, make sure Bingbot is not blocked by your CDN or firewall, and submit new URLs through IndexNow, which Microsoft recommends for keeping content current. A technical SEO audit finds the blocks that security tools add silently.
2. Export the three-month window, every month. One month is too short; the test found the list can change by half. Save both the topics and the opportunities so you can compare month to month.
3. Filter before you plan. Remove opportunities in languages you do not publish in, for countries you do not serve, for products you do not sell, and for competitors’ brand names. Expect only a handful to survive. That is normal.
4. Read your own grounding queries first. They show what Copilot already associates you with, which language version earns citations, and which pages it uses to answer questions about your prices, policies and services. Fix anything out of date on those pages before chasing new topics.
5. Compare your topics with what you sell. A missing topic you should own matters more than a long list of opportunities. If AI sees you as a "recipe blog" when you are a keto specialist, your positioning, not your keyword list, needs work.
6. Build pages that are easy to cite. For the survivors, Microsoft’s own guidance applies: depth and expertise, evidence for claims, fresh content, and "clear headings, tables, and FAQ sections". This is the core of answer engine optimisation and of good content strategy.
7. Measure across engines, not in one dashboard. Track Bing citations, Google AI impressions and a fixed prompt set for ChatGPT and Perplexity side by side. A reporting setup that combines them stops one beta report from driving your whole plan.

Why bother with Bing data if your traffic comes from Google?

Because, of the two search engines’ reports, Bing’s is the one that shows the queries behind your AI citations and names answers you are missing from. Microsoft’s March post describes its grounding as powering "nearly every major AI assistant in the market", a claim it does not break down, so do not read Copilot data as a proxy for ChatGPT. What does transfer is the diagnosis: if an AI system does not associate your site with a topic you sell, or cites an outdated page about your policies, the fix on your own site helps in every engine.

If you have already read our pieces on the AI Overviews links update and AI referral traffic versus Google organic, this is the measurement half of the same story: knowing which AI answers you appear in, and which you do not, is where the work now starts.