# Topics Portfolio: AI Positioning Evolution Through Thematic Clusters 14 October 2026 — Alessandro Caprai --- After years of thinking in terms of keywords, SEO faces a paradigm shift that many have intuited but few have yet seen in structured data form. Topics Portfolio, the beta feature within AI Performance on Bing Webmaster Tools, represents the first concrete implementation of a vision that shifts SEO and GEO analysis from individual queries to thematic clusters, finally offering an overview of where our site is cited by AI and where competitors prevail instead. This isn't just an additional report, but an operational answer to a question I'm often asked: how do you measure visibility in a context where users no longer click on a classic SERP, but receive a synthesized answer directly from artificial intelligence? Topics Portfolio attempts to provide a first answer, and it's important to understand its logic, metrics, and limitations. ## How Topics Portfolio Works: From Prompts to Grounding Queries The report sits alongside Citations & Pages, but radically changes perspective. Instead of listing individual cited pages or exact queries, Topics Portfolio groups information into macro-themes. This approach reflects how generative models actually work: they don't execute a literal query entered by the user, but translate the intent into one or more synthetic searches, called grounding queries, which are submitted to Bing's index to retrieve authoritative sources. In other words, what we see in Topics Portfolio aren't the exact prompts of end users, but the intermediate searches that the AI executes internally to construct the response. This is a fundamental distinction: we're observing the behavior of the retrieval system, not that of the user. This allows us to understand which semantic entities the model deems relevant for a given topic, and therefore where we need to position ourselves. ### Key Metrics: Beyond Impressions, Citations Matter The heart of the report consists of four main metrics, all expressed in estimates and ranges given the beta nature of the tool: **Topic** Represents the themes covered by the domain. These aren't editorial categories we choose, but semantic clusters automatically detected by the system based on actual citations received in AI responses. **Topic Size** Indicates the total volume of citations that a given topic generates in the AI Search ecosystem. It's a proxy for demand: the higher the Topic Size, the more that topic is requested by users through Copilot and other Microsoft tools. **Sites Cited** Shows how many competitors are cited on the same topic. It's a direct measure of competitive density: a high value indicates a crowded topic, where gaining space requires established authority or distinctive content. **Queries** Represents the total of distinct queries that define the topic. It's an indicator of semantic breadth: the more different queries are part of a topic, the more articulated the theme and the more layered editorial coverage it requires. But the most strategic data, the one on which to build editorial plans, are the **Opportunity Queries**: searches where AI already cites competitors but not our site. This is the true gold of the report, because it reveals immediate content gaps to fill. These aren't hypotheses or indirect competitive analysis: they're gaps certified by the system itself. ## Conditional Visibility and Access Limitations One aspect that needs immediate clarification: Topics Portfolio isn't available to everyone. The tab activates only for sites with sufficient history and volume of AI citations. Microsoft hasn't made the exact minimum requirements public, but from tests I've conducted, it emerges that you need an established presence in the index, a solid backlink profile, and an already detectable citation frequency. Data is filterable over 1, 2, or 3 months, with daily updates and a 48-72 hour delay. This is acceptable latency for a beta tool, but must be considered when evaluating rapid interventions or trying to measure the impact of recent editorial changes. ## Three Structural Limitations to Know While Topics Portfolio represents a significant step forward, there are three structural limitations that shouldn't be ignored. ### 1. Closed Ecosystem: Microsoft Only The report tracks exclusively the Microsoft ecosystem: Copilot, Bing's AI summaries, and integrated partners. Google AI Overviews, Claude, Perplexity, and all other AI engines remain outside. This doesn't make the data useless, quite the opposite: RAG (Retrieval-Augmented Generation) logics are similar across various systems, so Topics Portfolio can be used as an indirect indicator for other platforms too. But it should be considered a partial snapshot, not a complete picture of AI visibility. ### 2. Zero-Click Only: No Real Traffic Data Topics Portfolio provides only Zero-Click data, meaning citation frequency. There are no metrics on impressions, clicks, CTR, or actual traffic. This reflects a precise reality: in AI responses, users often don't click on the source. But it also means the report can't be used to measure conversion or direct commercial effectiveness. It serves to understand where we're semantically present, not to track ROI. ### 3. Changes Editorial Plans: Less Volume, More Precision As confirmed by tests conducted on DeepSEO, using Topics Portfolio radically changes the editorial approach. There's no longer a need to produce large volumes of generic articles hoping to cover the maximum number of possible queries. Effectiveness shifts to surgical interventions: identifying Opportunity Queries, mapping missing entities, and updating existing content to cover those specific gaps. It's a shift from quantitative to qualitative logic, where semantic depth matters more than publication frequency. And this has implications for teams, workflows, and success metrics as well. ## How to Integrate Topics Portfolio into SEO Strategy If you have access to Topics Portfolio, the first step is to map already covered topics and compare them with the current editorial strategy. Surprises often emerge: topics we believed central turn out marginal in AI citations, while secondary topics obtain unexpected visibility. The second step is to prioritize Opportunity Queries. Not all have the same value: Topic Size (to understand potential demand) must be crossed with Sites Cited (to assess competitiveness). Queries on high-demand, low-competition topics are the ones to attack first. The third, more complex step is working on entities. Opportunity Queries reveal which concepts, people, products, or events our site doesn't cover in a sufficiently structured way. Writing an additional paragraph isn't enough: you need to enrich the semantic graph of the content, insert structured data, update relationships between entities. Finally, a continuous monitoring cycle must be set up. Topics Portfolio is in beta, data fluctuates, citations change rapidly. A dedicated dashboard is needed, with alerts on significant variations and periodic reviews of emerging topics. ## Conclusions: A First Tool, Not the Last Topics Portfolio isn't the definitive solution to the problem of AI visibility measurability, but it's the first public tool offering structured data on how generative models cite content. It has clear limitations, restricted ecosystem, and partial metrics, but introduces a necessary perspective shift: from query optimization to thematic cluster optimization. For those working with technical content, complex editorial, or highly competitive sectors, it's a tool that deserves immediate attention. Not to replace traditional approaches, but to complement them with new logic that responds to how AIs retrieve and cite sources. And above all, it's a directional signal: Microsoft is investing in AI-native metrics, and this anticipates what we'll see on other platforms in coming months. Better to start familiarizing yourself now, while it's still possible to experiment without excessive competitive pressure.