Table of contents
Introduction
English version of an article first published in Polish on 18 October 2026. Read the Polish original.
On 13 October 2026 Bing added a Topics Portfolio tab to the AI Performance report. On paper it is one more view of the same data. In practice it is something Search Console has never offered: a list of queries where AI cites other sites but not yours. Free of charge and drawn from real traffic, not from prompts invented by a tool.
I read Bing's description, but what mattered more was what the tab shows on Polish websites. I checked it on 17 October on a dozen or so domains. The screenshots and descriptions below cover, among others, a site about prescription medicines, a site about sick leave and benefits from ZUS (the Polish social insurance institution), a parenting site and a site about vaccinations.
The result is more interesting than the description and the data would suggest. A large share of the opportunities Bing offers a Polish site concern other markets. Topics are assigned by a model, which now and then files cosmetics under a topic about eye diseases. Even so, I think this is the most interesting free data on AI visibility available today. You just need to know how to sift it.
You will learn:
- What exactly Topics Portfolio shows and how to read Topic Size, Sites Cited and Queries.
- How the new tab differs from the topics Bing added in June.
- Why this is the first time a free webmaster tool shows queries where your site is absent.
- What I saw on the accounts of Polish websites and why some of the "opportunities" were in English, Chinese, Spanish or Portuguese.
- How many of 250 opportunities suit a Polish site as they stand, and how to check the rest against Polish data.
- Which topic classification errors the Bing model makes today and what the company does not say about its numbers.
- How to use this data in content planning without chasing queries from markets you do not write for.
What does Topics Portfolio show?
The tab sits in Bing Webmaster Tools, in the AI Performance report, next to the Citations & Pages tab. In the interface it is called "Topic Portfolio" with a BETA badge, in Bing's announcement "Topics Portfolio", and in one place in that same announcement simply "Topics". They are all the same thing.
The starting point is grounding queries: the queries the model used when looking for content for an answer. The AI Performance report has collected them since February. The new tab groups them into broader topics and shows three numbers for each that describe the whole market, not just your site.
Topic
A broad topic to which the Bing model has assigned a group of your grounding queries. The names are always in English, on Polish sites too.
Topic Size
The approximate total number of citations across all sites in this topic over the selected period. Bing calls it a measure of demand in AI search. These are citations, not searches and not people.
Sites Cited
The approximate number of sites cited for queries in this topic. It tells you how crowded the topic is. You will not see the names of those sites.
Queries
The approximate number of distinct grounding queries in the topic, across the whole market, not just yours.
Below the topic table is the Top Opportunity Queries block. According to Bing, these are the strongest queries across the whole portfolio where other sites are cited and yours is not. Each one comes with its topic and a citation range, for example "10K-50K". On the accounts I checked, the list had exactly 50 entries.
Clicking a topic opens its card. At the top are three tiles: Topic Size, your citations in the topic and the number of queries. Below them are two tables. Your Grounding Query shows the queries in this topic for which you are already cited. Opportunity Query lists queries from the same topic for which others are cited. Bing puts it neatly: together they show the two halves of a topic, what you already have and what is still open.
Your own citations are given precisely, for example 5.5K or 259. Everything that concerns other sites comes in ranges. The time windows on offer are one, two and three months.
How is it different from the June Topics?
If you read my piece on measuring AI visibility in Bing Webmaster Tools, the name Topics may sound familiar. On 16 June 2026 Bing added four things to AI Performance: Intents, Topics, Citation Share and Compare. Those Topics, however, are a label on each query for which you are already cited. A column in a table, next to the intent and the citation count.
The June Citation Share calculates your share of citations for a given query, so the denominator includes other sites too. It still covers only the queries where you are present. Bing stated plainly at the time that it is not a ranking or a competitor leaderboard, and that it does not show the domains that hold the rest of the share.
The October tab adds two things that view did not have. The first is the size of a topic across the whole market: how many citations were handed out to all sites and how many sites fit into it. The second is the queries where you are absent. Bing itself writes that these are three things the report could not show before: a view of the whole portfolio, ecosystem context and specific opportunities.
A year of AI reports from Google and Bing
-
27 October 2025
Google: Query groups
queries grouped by AI in Search Console Insights, clicks only and your own queries only
-
10 February 2026
Bing: AI Performance
citations, cited pages and grounding queries in Bing Webmaster Tools
-
3 June 2026
Google: generative AI reports
impressions in AI Overviews and AI Mode, without queries and without clicks
-
16 June 2026
Bing: Intents, Topics, Citation Share
intent and topic labels on your queries, plus your share of citations
-
13 October 2026
Bing: Topics Portfolio (beta)
ecosystem-wide measures and the queries where you are absent
Queries where you are absent
Search Console has always shown the queries for which your site appears. It does not show the queries for which it does not appear at all. To find them you have to reach for a paid tool and compare yourself with competitors on its data, in other words run a content gap analysis. Bing Webmaster Tools worked the same way.
Topics Portfolio changes that rule in its own corner of the market. The queries in Opportunity Queries come, as Bing writes, from real grounding activity in Bing and not from a generic list. These are not phrases that a tool typed into an assistant itself to see whom it would cite. They are queries for which the model really looked for sources and really cited them. It just cited someone else.
Among Microsoft's services this is not the first data of its kind. Clarity, in its Citations report (generally available since 13 May 2026), counts the queries for which a domain could have appeared, whether or not it was cited. I described it in my article on AI visibility in Microsoft Clarity. In Bing Webmaster Tools itself, though, it is new, and a list of specific queries broken down by topic is far more practical than an aggregate metric.
A correction to my June article
In June I wrote that Google offers no equivalents of metrics such as Topics. That was imprecise. Since 27 October 2025 Search Console Insights has had Query groups: queries grouped by AI. The mechanism resembles Bing's: the groups are computed by a model, they can change over time, each one can be rated with a thumb, and only sites with a large number of queries see them. The difference lies in what is counted. Query groups show clicks from your own queries. They do not show the size of a topic across the market or the queries where you are absent.
The generative AI reports in Search Console, rolled out from June and available to all sites since 31 August, show only impressions in AI Overviews and AI Mode, broken down by page, country and device. No queries and no clicks. I wrote more about them in my article on AI reports in Search Console. In AI measurement, then, Bing is currently ahead of Google, and clearly so.
A test on Polish websites
The coverage of the announcement that I read mostly described Bing's post itself. Of those reports, only the Italian site DeepSEO showed a real account, using a gardening website as its example. I did not find a Polish test, so I ran my own.
I analysed the domains on my Bing Webmaster Tools accounts. I leave out those that had no citations at all in three months. For now, only some of the rest see the Topic Portfolio tab. The others are sent straight back to Citations & Pages when they open it.
Parenting site
38.2K citations
yes: 20 topics in 1M
Prescription medicines site
26.3K citations
yes: 17 topics in 1M, 21 in 3M
Sick leave and ZUS site
6.1K citations
yes: 6 topics
Kitchenware shop
1.5K citations
yes: 1 topic
Cosmetics shop
194 citations
no
Vaccination site
134 citations
yes: 1 topic
Company website
89 citations
no
Sites with a handful of citations
1–24 citations
no
Citations from the Citations & Pages view, 3M window, as of 17 October 2026. The last column says whether the site has the Topic Portfolio tab.
Bing does not say who gets the tab or in what order. It only writes that smaller sites may see a shorter list or may not see the feature at all. It is easy to conclude from this that everything depends on the number of citations, but my data does not support that. The vaccination site has the tab with 134 citations, and the cosmetics shop with 194 citations does not.
It is more likely that Bing is switching the feature on gradually, as with most new things in beta. For now some sites have it, and over time it will appear on more and more domains. If your site does not have it today, that does not have to mean it is too small. It is simply worth looking at the report from time to time.
Something else is interesting. Once a site has the tab, a topic can appear with very little presence. On the medicines site, one of the topics in the September window rests on a single query with three citations. On the vaccination site, one query, "szczepionka na ospe dla dzieci" (chickenpox vaccine for children), was enough to build the "Vaccines & Immunizations" topic and a full list of 50 opportunities.
For the medicines site, Citations & Pages shows 26.3K citations in three months. Topic Portfolio has 17 topics in the September window and 21 in the July–September window. The largest topic is "Medications & Prescriptions": a Topic Size of 500 million to 1 billion citations, 500,000 to 1 million queries and 50,000 to 100,000 sites cited. The site has 5.5K citations in it.
Those numbers alone say one thing: this is not the Polish market. It is hard to believe that Poland has 50,000 to 100,000 sites cited for queries about medicines. Topic Size and Sites Cited describe the topic at the scale at which Bing sees it, and Bing sees it globally. Bing does not spell this out in the announcement, but the order of magnitude leaves no doubt.
Opportunities from all over the world
This is the most important observation from my test, and I did not find it in any of the coverage. The queries for which a Polish site is cited are in Polish. The opportunities Bing offers it are not limited to the Polish market. On some sites this is easy to miss; on others it is obvious at first glance.
On the medicines site, the site's own queries include "leki na zawroty głowy" (medicines for dizziness), "czopki na hemoroidy na receptę" (prescription haemorrhoid suppositories), "ozempic zamiennik" (Ozempic substitute) and "jęczmień na oku" (stye on the eye). The opportunities are diclofenac, xarelto, ivermectin, "knee brace", "cough syrup", "earplugs", "eye drops for dry eyes" and the Spanish "lágrimas artificiales". Most of that list consists of international drug names, which people in Poland type the same way as everywhere else, so they may be real opportunities. The citation range next to them, however, describes demand across the whole Bing ecosystem, not in Poland. The rest are plainly English phrases that a Polish site can do little with.
On the site about sick leave (known in Poland as L4) it is even clearer. The site's own queries sound as if they came from a Polish forum about ZUS: "zwolnienie lekarskie" (sick leave) with 90 citations in September, "świadczenie pielęgnacyjne 2026" (carer's benefit 2026), "urlop ojcowski ile płatny" (how much paternity leave pays), "zus okres wyczekiwania" (ZUS waiting period), "przerwa między zwolnieniami 60 dni" (60-day gap between sick leaves). And the opportunities?
Selected opportunities for the sick leave site, 1M and 3M windows, 17 October 2026.
The biggest opportunity in the three-month window is the minimum wage in Portuguese, with a range of 100,000 to 500,000 citations. Next come Spanish income tax, the minimum wage in the Philippines, French maternity leave and the minimum wage in Chinese. Among the ten strongest opportunities in the leave topic there is not a single Polish query, but there is Taiwanese mental health leave.
The parenting site is similar. The whole list of 50 opportunities contains not one Polish query; instead there are English, Spanish, French and German ones, and one in Chinese.
The kitchenware shop shows that the problem affects e-commerce too. "Kitchen utensils", "serrated knife" and "serving tray" are good queries, just not for a Polish shop. The Polish equivalents of these queries are not among the 50 opportunities. The most interesting case is the vaccination site. It has one Polish grounding query about chickenpox, and Bing mostly offers it Dutch questions about reimbursement of the shingles vaccine and about vaccinations before travelling to Thailand or Sri Lanka.
Bing writes that topics and opportunities are personalised, and under each table stands the sentence "Topics and opportunity queries are personalized based on your context". On the accounts I checked, the personalisation covers the topic but not the market or the language. Bing knows what you write about, but does not take into account whom you write for. The announcement is silent on languages and countries, and all of Bing's examples are in English, so the problem is easy to miss if you test an English-language site.
There is one more detail that shows how raw this data is. "Salário mínimo 2026", "salario mínimo 2026" and "salario minimo 2026" are three separate entries among the opportunities. Bing does not merge variants that differ only in accents. The same happens in Polish: on the vaccination site, "szczepionka rsv", "rsv szczepionka" and "szczepionka na rsv" (all three mean RSV vaccine) are three separate opportunities.
For a Polish site this means something simple. The list of opportunities is raw material, not a finished content plan. How much of it can be taken as it stands depends on the industry, and the rest has to be translated and checked against Polish data. I counted this precisely on the full lists.
I sifted 250 opportunities. How many of them suit a Polish site?
Individual examples can mislead, so I took the full lists. All the sites that have the tab, 50 opportunities each in the September window, 250 queries in total. I assigned each one to a language and judged whether a Polish site can do anything with it. Then I checked the Polish equivalents of the strongest opportunities in Senuto (a Polish SEO tool) to see whether the topic has demand behind it on Google in Poland.
Which languages Bing offers
Languages of the opportunities in the 1M window (September), 50 per site, 17 October 2026. "Universal" means drug names, brands and terms that a Polish user types the same way, such as hcg, asperger or pampers premium care.
I counted 14 languages in the 250 opportunities. English leads with 116, almost half. Next come Spanish (26), Dutch (19), French (9), Chinese (7) and Portuguese (6), and there are also single queries in Swedish, Vietnamese, German, Italian, Hungarian, Persian and Ukrainian. There are 7 Polish queries, just under 3%, and all of them come from one site.
The mix of languages changes with the topic, not with the language of the site. For taxes and the minimum wage, English and Spanish dominate; for vaccinations, Dutch; for medicines, international names. All these sites publish in Polish, and Bing simply shows them the topic at world scale.
The sieve: how many opportunities remain
I put each opportunity into one of four groups. Polish queries in the site's topic are the obvious opportunities. Universal ones are names and brands that people in Poland type the same way, so they count too. Queries in a foreign language, even in the right topic, give a Polish site nothing as they stand. The fourth group is queries that are off-topic or wrongly assigned, such as the cosmetics with retinal.
Medicines site
Polish 0 universal 39 foreign 8 off-topic 3
39 of 50
Parenting site
Polish 0 universal 9 foreign 41 off-topic 0
9 of 50
Vaccination site
Polish 7 universal 1 foreign 42 off-topic 0
8 of 50
Sick leave and ZUS site
Polish 0 universal 0 foreign 50 off-topic 0
0 of 50
Kitchenware shop
Polish 0 universal 0 foreign 50 off-topic 0
0 of 50
Total usable as they stand
56 of 250, 22%
56 of the 250 opportunities (22%) are usable as they stand. The spread between sites is huge, though. The medicines site keeps 39 of 50, because it lives off names that sound the same in every country. The sick leave site and the kitchenware shop keep none, because their topics are described differently in every language.
For medicines, Senuto changed my classification. Eleven opportunities are substance names in English spelling, such as atorvastatin or ciprofloxacin. I assumed that Poles search for the Polonised forms. Some do: "semaglutyd" has more than nine times as many searches as "semaglutide". But "atorvastatin" and "rosuvastatin" are typed more often than "atorwastatyna" and "rosuwastatyna", and "ciprofloxacin" just as often as "cyprofloksacyna". That is why I count them as universal in the sieve.
Average monthly searches on Google in Poland: English spelling (left column) and Polish spelling (right). Senuto data, 18 October 2026.
A foreign query is a pointer to a topic
The most important conclusion came at the third step. A foreign-language opportunity does not have to be written off straight away. It is enough to translate it into an intent and check whether anyone in Poland searches for the same thing.
On the left the opportunity from Bing Webmaster Tools, in the middle the Polish equivalent, on the right the average monthly searches on Google in Poland according to Senuto, 18 October 2026. This is demand on Google, not in Copilot.
The kitchenware shop has not a single Polish opportunity, yet behind its English list stand topics with tens of thousands of Polish searches a month: kitchen scales, knife sharpeners, induction pots. The sick leave site gets the English "compassionate leave", and its Polish equivalent, "urlop okolicznościowy", is searched for 18,000 times a month in Poland. Not every opportunity passes this test. The Dutch questions about the shingles vaccine translate into a mere 210 searches in Polish, and vaccinations before a trip to Sri Lanka into barely ten.
An opportunity from Bing tells you in which topic AI cites others. It becomes a keyword only after it has been translated and checked against Polish data. Demand on Google is not a measure of the questions put to Copilot, but it is a good filter. It is hard to expect a topic that almost nobody in Poland searches for to be a frequent subject of conversations with an assistant.
What only the full lists reveal
- Polish opportunities rank high. On the vaccination site, three of the seven strongest opportunities are in Polish, in the 1K-5K and 500-1K citation ranges. If there is anything in Polish on the list, that is where to start.
- 50 entries are fewer than 50 topics. On the sick leave site, the minimum wage in various countries and languages takes up 13 places, and tax brackets 9. Bereavement leave appears twice, each time with a different typo: "berevement leave" and "bereavment leave".
- Some opportunities are tied to one country. "Philippines minimum wage september 2026", Dutch reimbursement of the shingles vaccine or "高雄親子景點", attractions for children in Kaohsiung, Taiwan, are topics a Polish site cannot serve even after translation.
When AI assigns the wrong topic
Bing is honest in warning that the classification is done by AI and can be imperfect, so a query may land in a topic you would have named differently yourself. On Polish data such mistakes show up straight away.
- Retinal is not the retina. The opportunities "medik 8 crystal retinal" and "celimax retinal shot" are cosmetics with retinal, a vitamin A derivative. The model assigned them to the topic about eyes, because of the retina of the eye.
- Obesity filed under eyes. The site's own query "otylosc iii stopnia" (class III obesity) ended up in "Eye & Vision Health".
- ICD-10 codes as medical billing. The site has a "Medical Billing & Coding" topic. In the US that is the industry of billing insurers. Here it grew out of queries such as "otylosc icd10" (obesity ICD-10), which are about the disease codes Poles see on a sick note.
- Insects on a site about medicines. The "Insects & Pollinators" topic arose from one query with three citations. Its opportunities include "apis mellifica", a homeopathic remedy, and "lepidium meyenii", which is maca, a plant.
There is also a problem less visible than individual mistakes. Topics overlap. "Ozempic" belongs both to the topic about weight-loss drugs and to the topic about prescription medicines. "Mounjaro" sits in three topics, and "helicobacter pylori" in two. On top of that there is a general "Health" topic that covers most of the others. Adding up Topic Size or citations across topics therefore makes no sense, because the same queries are counted several times.
The queries from which topics are built are interesting too. On the sick leave site one of them reads "pue zus logowaniekiedy liczy sie nowy okres zwolnienia lekarskiego 182 dni" (a login query for the ZUS online portal, followed by "when does a new 182-day sick leave period start"). These are two queries glued together without a space. A person is unlikely to type that; it looks like the trace of a model generating several queries for one answer. It is a good illustration of what Bing's help page says: grounding queries are not full questions or user prompts.
What does Bing not tell you?
Bing itself asks you to treat these numbers as a directional signal for planning, not as a precise ledger. After the test I know where that caution comes from. Here are the things worth knowing before you show this data to a client or a boss.
The windows are calendar months
On 17 October the buttons read "1M (Sep)", "2M (Aug-Sep)" and "3M (Jul-Sep)". October will appear once it has ended, so the data lags by up to six weeks. The announcement does not mention this. DeepSEO pointed it out as well.
The ranges are wide
"10M-50M" is a fivefold spread, and so is "100M-500M". A move of one range can mean growth of a few per cent or of several times over. The announcement speaks of "approximate" numbers, and in practice all three ecosystem measures are ranges.
Topic Size is citations, not people
A large topic can mean many queries or many sources in each answer. Bing does not separate the two. PPC Land pointed this out.
Nobody knows who takes the citations
The opportunities tell you that a query goes to others, but not to whom. Bing does not show competitor names, just as it did not show them with Citation Share.
"Strongest" without a definition
Nobody knows how Bing picks the strongest opportunities or why the list ends at 50. On the accounts I checked it had exactly 50 entries. For the topic with three citations there were 11.
Competitors' branded queries
Bing admits that they end up among the opportunities. Clarity has had a split into branded and non-branded queries since 3 August; Bing Webmaster Tools does not have one.
An export without the period
The CSV file from the topic table has the columns Topic, Volume Range, Queries Range and Sites Range. In the interface, Volume Range is called Topic Size. The file does not say which period it covers, and its name contains only the download date.
No documentation and no date for leaving beta
In mid-October the AI Performance help page did not yet describe the new tab. There is no date for leaving beta, and AI Performance itself still carries a BETA badge.
I would add one question to this list that I did not see in the coverage. Some grounding queries in AI Performance have looked suspicious from the start. Analysts at Hive Digital described queries stuffed with keywords and more than a thousand citations against three impressions in Bing. In their view these are traces of AI visibility monitoring tools, which query assistants in bulk. If they are right, part of the "demand" that Bing shows as Topic Size may be generated by the SEO industry checking on itself. I am not saying that this is the case. I am saying that nobody has ruled it out.
Where does the data come from?
Above every AI Performance view stands the note "Citation sources: Microsoft Copilots and Partners". Bing's help page breaks this down into three sources: Microsoft Copilot, AI summaries in Bing results and select partner integrations. Bing has never named the partners.
With these reports one question comes up immediately: is ChatGPT in here? The honest answer is that nobody knows. OpenAI's help article says that ChatGPT search sometimes uses other search providers, and Microsoft's privacy policy is listed among those providers' policies. That is still too little to claim that citations from ChatGPT reach Bing Webmaster Tools. Neither OpenAI nor Microsoft says so. Jordi Ribas of Microsoft wrote in February that Microsoft's grounding powers almost every major AI assistant on the market, but he did not name any.
That is why I read Topics Portfolio as a picture of the Microsoft ecosystem and unnamed partners, not of AI search as a whole. It is a large enough slice to draw conclusions about topics. It is too small to carry results over one to one to ChatGPT or AI Overviews.
In Poland that slice is not marginal, although it depends on the device. According to StatCounter, in September 2026 Bing had 7.38% of all searches in Poland, 13.96% on desktop and 1.27% on mobile. Google had 89.56% overall. From January to September 2026, Bing's share in Poland stayed between 7.12% and 8.02%. For B2B companies whose customers work on computers with Windows, Edge and Microsoft 365, Copilot is at hand every day. For a shop that lives off mobile traffic, it is a much smaller channel.
Free Bing versus paid tools
Gap analysis for AI visibility is a paid product today. Semrush has a report in its AI Visibility Toolkit that looks for prompts and topics where others are cited and you are not. Ahrefs Brand Radar shows citations against competitors. The difference from Bing is fundamental: the paid tools measure answers to prompts that they send to assistants themselves. Bing shows queries that passed through its grounding in normal traffic.
Bing Topics Portfolio
free
Data: real grounding queries from Copilot, Bing and partners
Queries where you are absent: yes, without competitor names, in ranges
Google Search Console
free
Data: your own queries with clicks; impressions only in the AI reports
Queries where you are absent: no
Microsoft Clarity
free
Data: citations from grounding plus prompts you define yourself
Queries where you are absent: partly, with the domains cited alongside you
Semrush AI Visibility Toolkit
from 99 USD/mo, 25 prompts
Data: prompts the tool sends to assistants
Queries where you are absent: yes, with competitor names
Ahrefs Brand Radar
from 199 USD/mo
Data: a prompt index modelled on searches
Queries where you are absent: yes, with competitor names
Prices from the vendors' price lists as of 17 October 2026, the cheapest plans that include the feature.
The paid tools have two advantages. They show competitor names and they cover assistants that Bing does not measure: ChatGPT, Gemini, Perplexity or AI Overviews. They also have a drawback that is rarely mentioned: they see only what they ask about themselves. If you do not enter a prompt, you will not find out that someone is asking about something you had not thought of. Bing needs no such assumptions, because it shows queries that really occurred.
I do not treat this as a choice. Bing gives you a free list of topics and queries you would not have come up with yourself. A paid tool lets you check who wins on them and what it looks like in other assistants. The two sources complement each other.
How should you use it?
If your site has the Topic Portfolio tab, here is the order I suggest starting with. If it does not, the same list will be useful once the tab appears. Until then you have Citations & Pages and Clarity.
01. Start with the topics you are already in
Bing advises starting with the topics where you have the most citations, and it is good advice. That is where the opportunities are closest to what you already write, and where the model has the most data about you.
02. Sift the opportunities by language and market
Throw out queries about specific other countries, such as the minimum wage in the Philippines. Do not write off foreign-language opportunities in your topic straight away: translate the intent and check Polish demand in a keyword tool. Without translation, only 22% of the opportunities in my test were usable.
03. Throw out competitors' branded queries
Bing admits that they end up among the opportunities. The name of a medicine or a product can be a fair opportunity; the name of a competitor's shop is not.
04. Check whether you have an answer
An opportunity makes sense if the site can answer the query better than the authors cited today. With medical, legal and financial topics there is also the question of the author's expertise, not just the text.
05. Save the export with the period in its name
The export file contains no information about the time window, and its name carries only the download date. Add the period to the file name straight away, because in a month you will not be able to tell the 1M view from the 3M one.
06. Rate topics with the thumbs
Every topic and every opportunity has a thumbs up and a thumbs down. Bing writes that these signals go straight to the team working on classification. Nobody knows whether they change what your site sees, but it is the only way to make errors like "retinal = eye" disappear.
Once you have sifted the opportunities, it is best to compare them with what Search Console and a keyword research tool say. A query that has demand on Google in Poland, and for which Bing cites competitors, is a good candidate for an article or for expanding an existing page. A query with no Polish demand but a high citation range in Bing usually turns out to be a phrase from another market.
My view
I had got used to free search engine tools showing you your own backyard. Everything that happens beyond it was the domain of paid databases. Bing has just moved that boundary, at least within its own AI ecosystem. It is a big step forward, but I would not move this data into a content plan without sifting it.
Two things matter to me. First: the missing market filter is not a cosmetic issue but a flaw that changes the meaning of the whole feature for Polish sites. An opportunity for a site about Polish sick leave should not read "minimum wage in the Philippines". Second: Bing admits its limitations openly, in the announcement itself, and provides thumbs for reporting errors. That is rare and deserves credit.
For the first time a search engine shows, free of charge, the queries for which AI picks someone other than you. That is a big deal. But as long as Bing cannot tell a Polish site from a Portuguese one, this list is working material for a specialist, not a finished plan for a site owner.
FAQ: common questions
What is Topics Portfolio in Bing Webmaster Tools?
It is a tab in the AI Performance report, announced as a beta on 13 October 2026. It groups the grounding queries for which a site is cited in AI answers into broader topics. For each topic it shows the total number of citations across all sites (Topic Size), the number of sites cited (Sites Cited) and the number of queries (Queries). Below the topic table are the Top Opportunity Queries: queries where other sites are cited and yours is not.
Why can I not see the Topics Portfolio tab?
Most likely because it is a beta that Bing is rolling out gradually. On the domains I checked, only some sites saw the tab and it did not depend on the number of citations: a site with 134 citations in three months had it, and a site with 194 did not. Bing also writes that topics with insufficient site presence or ecosystem activity are not shown. If you do not see the tab today, it is worth looking again from time to time, because it may appear on more domains every week.
Are Opportunity Queries worth writing content for?
Not all of them. On the Polish websites I checked, Bing did not limit opportunities to the market of the site. For the site about sick leave there was no Polish query in the top ten, and for the parenting site and the kitchenware shop there was none among all 50 opportunities. Some are also competitors' branded queries, which Bing itself admits. Treat the opportunities as a list to sift. In my test 56 of 250 (22%) were usable as they stood. Foreign-language queries are not worthless, though: once translated into the Polish intent, some of them lead to topics with high demand on Google in Poland.
How is Topic Size different from search volume?
Topic Size is the approximate total number of citations across all sites in a topic, not the number of searches or users. One AI answer can contain several citations, so a large topic can mean many queries or many sources in each answer. Bing gives the number as a range, for example 100M-500M.
What period does the data in Topics Portfolio cover?
You can choose one-, two- and three-month windows, but they are counted in calendar months. On 17 October 2026 the 1M view showed September, 2M showed August and September, and 3M showed July, August and September. October will appear only once it has ended.
Does Topics Portfolio show data from ChatGPT?
Nobody knows. Bing Webmaster Tools names Microsoft Copilot, AI summaries in Bing and select partners, which it does not identify, as the citation sources. OpenAI writes that ChatGPT search sometimes uses other search providers, but neither OpenAI nor Microsoft confirms that citations from ChatGPT reach this report.
Does Google have an equivalent of Topics Portfolio?
Partly. Since 27 October 2025 Search Console Insights has had Query groups: queries grouped by AI, with thumbs for rating the groups. Query groups count clicks, however, and cover only the queries for which the site appears. They show neither the size of a topic across the market nor the queries where the site is absent.
Worth remembering
- Topics Portfolio (beta) was announced on 13 October 2026. It groups grounding queries into topics and shows ecosystem-wide measures for each: total citations across all sites, the number of sites cited and the number of queries.
- The most important new element is Opportunity Queries: queries where AI cites other sites but not yours. Search Console has never shown queries for which a site does not appear.
- On the Polish websites I checked, the opportunities were not limited to the Polish market. For the parenting site and the kitchenware shop, not one of the 50 opportunities was a Polish query; the vaccination site had seven. They need sifting before any content planning.
- Of 250 opportunities, 56 (22%) suit a Polish site as they stand: Polish queries plus names and brands that a Polish user types the same way. The rest are pointers to topics. Once translated, some of them have high demand on Google in Poland, for example "waga kuchenna" (kitchen scale) with 27,000 searches a month and "urlop okolicznościowy" (compassionate leave) with 18,000.
- Topics are assigned by a model, and it makes mistakes. Cosmetics with retinal ended up in a topic about eyes, and a single query can belong to three topics at once, so adding up results across topics makes no sense.
- The time windows are calendar months. On 17 October the "1M" view showed September, so the data lags by up to six weeks.
- The tab is a beta that Bing is rolling out gradually. For now only some sites see it, and citation counts do not explain who does: a site with 134 citations in three months has it, and a site with 194 does not. It is reasonable to expect it to reach more and more domains over time.
- The data comes from Copilot, AI summaries in Bing and unnamed partners. Neither Microsoft nor OpenAI confirms that it includes citations from ChatGPT.
Summary
Topics Portfolio is the first free search engine tool that shows the queries for which AI cites others and not you. The data comes from real grounding in the Microsoft ecosystem, not from prompts invented by a tool, and that is its greatest value.
On Polish sites that value takes work to extract. The opportunities come from all over the world, topics are sometimes assigned oddly, the numbers are ranges, and the data ends with the last full month. If your site has the tab, going through it and sifting the opportunities will take an hour. Only then does the real work begin: content that answers the chosen queries better than the pages cited today. That takes months, and no report can shorten it.
Sources and materials
Announcements and documentation from Microsoft and Google, industry coverage and market data on which I base the facts, quotations and dates given here. The screenshots and site data come from my clients' Bing Webmaster Tools accounts, without names:
- Bing Webmaster Blog – Introducing Topics Portfolio (Beta) in AI Performance (13 October 2026)
- Bing Webmaster Blog – Introducing AI Performance in Bing Webmaster Tools Public Preview (10 February 2026)
- Bing Blogs – New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare (16 June 2026)
- Bing Webmaster Tools – help: AI Performance
- Bing Blogs – Elevating the Role of Grounding on the AI Web (12 February 2026)
- Search Engine Journal – Bing Webmaster Tools Sorts AI Search Queries Into Topics (13 October 2026)
- Search Engine Roundtable – Topics Portfolio In Bing Webmaster Tools AI Performance Reports (13 October 2026)
- PPC Land – Bing shows sites the AI search queries where rivals get cited and they don’t (13 October 2026)
- Exploding Topics – Bing Topics Portfolio: Find Topic Gaps for AI Search (14 October 2026)
- DeepSEO – Topics Portfolio di Bing Webmaster Tools (14 October 2026)
- Hive Digital – How Grounding Queries and Query Fan-outs Drive LLM Visibility
- Google Search Central Blog – Query groups in Search Console Insights (27 October 2025)
- Google Search Central Blog – Generative AI performance reports in Search Console (3 June 2026)
- Microsoft Clarity – Citations now generally available (13 May 2026)
- Microsoft Clarity – Topic Insights (9 July 2026)
- Microsoft Clarity – Branded and non-branded queries (3 August 2026)
- OpenAI – Searching the web with ChatGPT (help article)
- StatCounter – search engine market share in Poland (September 2026)
- Senuto – Google search volumes in Poland (as of 18 October 2026)
- Ahrefs – Brand Radar pricing
- Semrush – AI Visibility Toolkit (knowledge base)
Want to know where AI cites competitors instead of you?
I will go through your Bing Webmaster Tools and Clarity data, sift the opportunities for your market and point out the topics where your site has a realistic chance of being cited.
See AI SEO servicesHow this text was written
I will not pretend it was written entirely by hand, by candlelight. For the research, the proofreading and the code of this page I used a language model. It is a good tool for digging through documentation and catching the typos you no longer see on the third read of your own text. The English version was also prepared with a model, from my Polish original.
What the model did not do: it did not choose the topic, form the arguments or decide what this data means for your site. That stays with me, as does the last word on what goes into the text and what comes out of it.
Where the model helped
- digging through sources and organising data
- proofreading and the English version of the text
- code, page layout and infographics
What stays with me
- the choice of topic and what is worth saying about it
- the arguments, the reading of the data and the conclusions
- fact-checking and the decision on what goes into the text