Bing Just Gave You a List of AI Queries You Don't Win. That Part Was Easy.

Bing Just Gave You a List of AI Queries You Don't Win. That Part Was Easy.

On October 13, Bing did something Google won't do. It handed site owners a list of the AI-answer queries where other sites get cited and theirs doesn't.

The new Topics Portfolio tab inside Bing Webmaster Tools' AI Performance report groups the queries your site is cited for into thematic clusters, then lists the strongest queries where other sites appear and you don't. Microsoft's Bing team announced it as a beta, with authors Krishna Madhavan, Saral Nigam and Vivek Jain explaining that it shows "where your site has room to grow."

Search Engine Roundtable covered the launch. As far as I can tell, no major search engine has handed you the queries you're missing before; Google's Generative AI report shows impressions only, no queries, topics, or missed opportunities.

But a gap list is not a fix. The report is the easy half; turning it into citations is the hard half, and it does not require writing 40 pages you didn't need.

What the report shows you

First, the vocabulary the protocol below relies on.

Grounding queries. These are the short search phrases Bing's AI uses to fetch and cite sources for an answer in Copilot or Bing's AI summaries, plus select partner integrations. They are retrieval wording, not necessarily what users typed.

The three topic metrics. Each topic gets Topic Size (approximate total citations across all sites, which reads as AI search demand), Sites Cited (approximate number of sites appearing, which reads as competition), and Queries (approximate number of distinct grounding queries). All three are approximate; Microsoft's line is to treat them as directional signals for planning, not a precise ledger.

Opportunity queries. Below the topic table, Top Opportunity Queries lists the strongest queries where other sites appear and yours doesn't, with citation volumes as ranges. Click a topic and the queries split into Your Grounding Queries (where you already appear) and Opportunity Queries (where you don't).

The tab only appears if Bing has enough data on your site; smaller sites may see a shorter list or nothing at all.

What Bing's report covers that Google's doesn't

Bing's own June 2026 Topics report only grouped queries where your site was already cited. Topics Portfolio adds demand data from across all sites plus the missed queries, which makes it the first version of the report that works as a planning tool. Neither report covers ChatGPT, Perplexity or the Gemini app, and neither has click data yet: as Barry Schwartz put it on the announcement, "No click data yet, sorry folks."

At First Rank, this changes how I start a GEO engagement. I used to answer the which-queries-are-we-losing question by hand; now Bing hands me the opening list and the intake hours go to the answer-versus-citation split. A citation is not a click, and the report won't tell you which citations move revenue, but it tells you where to look.

The 5-step protocol I use on a gap list

The discipline of the first three steps comes from Mustafa Aydemir's Medium walkthrough of the report: record the window, don't treat the metrics as a ledger, and separate a missing answer from a missing citation. This workflow drove a +98% LLM-traffic lift across a 30-day GEO sprint for a client campaign. I run it alongside my fixed prompt-panel recurrence scoring (10 money questions, four engines, logged in and out, three samples per question, Monday cadence) for the engines Bing doesn't cover.

Step 1: Record the window before you read anything. The report lets you show one, two, or three months of data. Write down which window you picked; the list shifts with it.

Step 2: Read the metrics the right way. Topic Size divided by Sites Cited is not your chance of winning, so don't do that math. Topic Size is a demand signal; Sites Cited is a crowd signal. Start with a topic connected to something your customers actually need. A huge unrelated category is a distraction no matter how big the number looks.

Step 3: Sort every opportunity into one of two buckets. Separate a missing answer from a missing citation: two different problems that need different fixes. For each opportunity query, ask whether the answer exists anywhere on your site.

Bucket A is a missing answer. That's a content decision, and even then the right answer is usually one page that earns five related citations, not five pages.

Bucket B is a missing citation. You do cover it, and Bing still didn't cite you. It's almost never fixed by writing. When a page answers the query but never gets cited, I check, in this order: is the page indexed and crawlable, are Bing's meta directives telling it to stay out, and is the answer retrievable. Per Bing's meta-tag guidelines, as GeoToolbox's walkthrough summarizes: NOARCHIVE prevents content from being used in Copilot responses and grounding results, and NOCACHE limits Copilot to the URL, title and snippet. No content edit will move a directive-blocked page until the directive is lifted.

Step 4: For bucket B, fix the citation, not the topic. Most bucket B problems are structural. The answer is buried under three screens of marketing copy. The page answers five questions and the grounding query needs one. The fix is usually to make the answer extractable: a direct, short answer near the top, a clear heading that matches the retrieval wording, and structured data that agrees with the visible text. Then re-crawl and wait for the next window.

For bucket A, write the page only if the intent fits your business. Opportunity queries can include branded queries, sometimes a competitor's brand name, and AI-generated topic labels can be wrong: a GeoToolbox walkthrough shows "que es claude" filed under Baby Names and "claude precios" under E-commerce. Exercise judgment before you chase someone else's name.

Step 5: Run the cross-check the report can't give you. Topics Portfolio covers Copilot, Bing AI summaries, and unnamed partner integrations, not Google AI Overviews, ChatGPT, Perplexity, or Gemini. Run your bucket A and bucket B conclusions through my recurrence method: the same fixed set of money questions, across the engines the report can't see, logged in and out, three samples per question, noting which model answered. If Bing's gap matches the other engines, the page is the problem. If only Bing misses it, the retrieval wiring is the problem.

Keep the limits in your head while you work

The report is a beta built on AI-generated topic labels. Every number is approximate, niche topics may not appear, and branded queries can show up as opportunities, including a competitor's brand: exercise judgment before you chase someone else's name. Competing domains are not listed, and there is no click data. Citations, traffic, and clicks all need to be recorded to get the full picture.

None of this makes the report useless; it is a starting signal, which is what Microsoft says it is.

The honest close

Sort your opportunity list into missing answers and missing citations. Fix the citations first. Then run the rest through the prompt panel on Monday and see which engines agree with Bing.

Published on the dalindegraff.com blog.