For most of the past two years, measuring visibility in AI search has involved a fair amount of guesswork.
Teams build prompt lists, run the same questions through different assistants and record which brands or publishers appear. The approach can be useful, but the prompts are chosen by the person doing the measurement.
Bing has now added something different.
Its new Topics Portfolio beta in Bing Webmaster Tools uses Bing’s own AI Performance data to group queries by topic and surface Opportunity Queries where other sites appear in AI answers while yours does not.
That is a much more interesting starting point than another visibility score.
It tells a publisher where Bing’s own systems are already finding someone else useful enough to cite.
The harder part is deciding whether the missing citation is actually a problem worth fixing.
Bing is exposing the gap instead of making you guess
Topics Portfolio sits inside Bing Webmaster Tools' AI Performance reporting.
Microsoft says it groups grounding queries into topics and shows information such as Topic Size, Sites Cited, Queries and opportunity queries.
The values are presented as ranges rather than exact counts, and reporting can cover one to three months.
That limits how precisely the data can be treated. A range is not a search-volume number, and it should not be converted into a precise market-share calculation.
Even with that limitation, the feature changes the starting point for AI visibility work.
Instead of asking, “Which prompts should we monitor?”, a publisher can begin with, “Where is Bing already citing somebody else?”
Those are not the same question.
One begins with a synthetic test set. The other begins with observed activity inside the search engine.
The opportunity list is not a content brief
The obvious mistake would be to export the missing queries and start commissioning articles.
Microsoft itself gives reasons not to do that.
Classification errors can occur. Some sites may not have enough AI Performance data to see Topics Portfolio at all. And competitor-branded queries can appear in the opportunity list.
Suppose a software company sees a missing query for a rival’s pricing.
Technically, it may be an opportunity. Other sites appear and yours does not.
Editorially, it may be worthless.
The user is looking for information about a specific competitor. Creating a thin page just to chase that query could add another piece of content without making the site more useful.
The same problem applies to queries outside your market, products you do not sell, support questions you cannot answer and topics that have no commercial or editorial value.
So the first job is not optimisation.
It is rejection.
If half the opportunity list disappears after a sensible review, that is not a failure. It is the filtering doing its job.
Existing pages may be more important than new ones
The most useful opportunities may be queries where the right page already exists.
Imagine Bing repeatedly cites other publishers for a question your site already answers.
Before creating another URL, compare the existing page with what is being cited.
Is the answer buried halfway down the article? Is the page three years old? Does the competitor give the comparison in two paragraphs while yours takes 800 words to reach it? Are important facts trapped in vague copy instead of stated directly?
Those are much more practical questions than whether a page has been “optimised for AI.”
Citation gaps can expose ordinary editorial weaknesses.
A page may rank reasonably well in conventional search and still be less useful as a grounding source because the answer is difficult to extract.
That is one reason the new Bing report is worth paying attention to. It can send editors back to pages that already have authority, links and history rather than encouraging an endless cycle of new content.
The same principle applies to the wider changes we are tracking in AI search. Our recent Gemini 4 Argon analysis makes a similar distinction: observe what changed on the surface before inventing a theory about what changed underneath it.
Competitor citations are useful even when you do nothing
There is another value in the report that has little to do with fixing pages.
It shows which sites Bing repeatedly finds useful around a topic.
That can reveal a publisher you were not watching, a specialist source that keeps appearing, or a competitor page that answers a question unusually well.
You do not have to copy any of it.
But looking at the cited page can tell you what kind of material the query seems to reward.
Maybe it is a first-party dataset rather than another explainer. Maybe it is a comparison page with current pricing. Maybe the winning source is short but unusually direct. Maybe it includes primary documentation while everyone else cites each other.
That kind of competitive reading is more valuable than a generic checklist about article length or schema.
The point is not to reverse-engineer a universal citation formula.
It is to understand why a particular page is useful for a particular question.
Do not turn Bing into a universal AI visibility score
Topics Portfolio is Bing data.
It does not tell you which pages Google AI Mode will cite. It does not reveal ChatGPT’s source selection, Gemini’s citation behaviour or Perplexity’s results.
A publisher could close several Bing citation gaps and see no equivalent movement elsewhere.
That would not make the work pointless. It would simply mean the measurement belongs to the surface that produced it.
This distinction is becoming more important as search teams add more AI visibility dashboards.
Combining every platform into a single score can make reporting look cleaner while hiding what actually changed.
Bing has done something useful here because it is exposing more of its own data.
Publishers should take advantage of that without pretending it represents the rest of the AI search market.
For now, the best use of Topics Portfolio is straightforward: find the missing queries, throw away the bad ones, inspect the pages that are already winning, and fix existing content where the gap is real.
The opportunity list is valuable because it narrows the search.
It should not be allowed to write the editorial calendar.