Query Fan-Out: A Data-Driven Approach to AI Search Visibility

Google’s AI Mode search doesn’t just process your query—it explodes it into multiple sub-queries, fanning out across Google’s knowledge graphs and its web index to synthesize comprehensive answers. This “query fan-out” process represents the most significant shift in search behavior since mobile-first indexing, yet it seems as if we’re all flying blind, unable to measure how well our content performs against this new reality.

After reviewing foundational patents like US 2024/0289407 A1 (Search with stateful chat) and newer insights from WO2024064249A1 (Systems and methods for prompt-based query generation for diverse retrieval)—the latter highlighted by Michael King’s analysis—and studying AI search patterns for several months, I’ve updated our practical framework and the accompanying Colab tool. The breakthrough isn’t just understanding query fan-out—it’s being able to more accurately simulate its initial stages and score your content against it.“

The Query Fan-Out Reality: Beyond Deterministic Rankings

When someone asks Google’s AI “What’s the best sustainable marketing strategy for small e-commerce businesses?”, the AI doesn’t search for that exact phrase. Instead, it decomposes the query into multiple sub-queries that vary based on user context:

• What makes a marketing strategy sustainable?
• Which marketing channels work best for e-commerce?
• What are the budget constraints for small businesses?
• How do sustainable practices impact customer acquisition?
• What are successful case studies in sustainable e-commerce marketing?

But here’s the critical insight from my work since the SGE (Search Generative Experience) days: these follow-up questions aren’t the same for everyone. They’re deeply contextual, stochastic, and impossible to predict deterministically.

The problem for marketers isn’t just optimizing for sub-queries—it’s accepting that there’s no way to “rank” or even track visibility in traditional terms. The game has fundamentally changed.

A New Framework: Predicting Questions

Traditional SEO tools struggle with the dynamic nature of query fan-out. What we need is a way to probe how our content might fare when an AI deconstructs user intent. Our updated Colab simulator now takes a more AI-native approach to this challenge:

(Future Vision) Validation Function & Reward Policies: While the current tool simulates and scores, the long-term vision remains developing robust validation functions—reward policies within DSPy to ground and improve predictions for actual contextual follow-ups based on user interactions and knowledge graphs

AI-Powered Entity & Context Understanding: Instead of traditional scraping, the tool leverages Google’s Gemini model and the url_context tool that simulate how AI Mode might interpret URLs. Gemini identifies the main ontological entity and extracts key grounded content chunks it finds relevant from the page.

This more closely reflects how AI search systems digest content: not by parsing HTML from top to bottom, but by grounding themselves in semantically rich, citation-worthy chunks.

Chain-of-Thought Synthetic Query Generation: Informed by the entity identified by Gemini, and guided by principles from patents like WO2024064249A1, the simulator uses a DSPy Chain-of-Thought (CoT) module with Gemini. This CoT module first reasons about the different types of information facets and query types (Related, Implicit, Comparative, etc.) relevant to your entity, then generates a diverse set of synthetic fan-out queries. This is a step beyond simple keyword expansion, aiming for a more reasoned decomposition.

Semantic Coverage Assessment: Using the Gemini-extracted content chunks (or a Gemini-generated summary if direct chunks aren’t available), we assess how well your page’s key information covers these synthetically generated fan-out queries using semantic similarity (embeddings).

The goal isn’t to predict every possible query—it’s to build semantic infrastructure that increases our accuracy in predicting the next likely question based on user context. Gianluca Fiorelli has been working on this for years recommending us to retrieve all the query refinements (Topic filter, PAA, People Also Search For, Images Search tag queries etc) that Google presents for a seed query.

🧱 How Google Chunking Works—and Why It Matters

Google doesn’t analyze content as a whole. Instead, it segments documents into smaller, meaningful chunks. Based on our research, Google appears to use a hybrid strategy that includes:

• Fixed-size chunking, used to fit content into model limits like the 2048-token cap for gemini-embedding-001
• Recursive chunking, to split unstructured content based on paragraphs → sentences → words
• Semantic chunking, where related sentences are grouped together
• Layout-aware chunking, which segments content based on HTML structure (headings, lists, tables) and is the default in Vertex AI Search via LayoutBasedChunkingConfig

Among these, layout-aware chunking is most crucial for ingestion. It aligns with how documents are visually structured, how humans process them, and how Google’s AI determines passage-level relevance.

Now, before, sharing how we’re planning to scout for follow-questions, let’s review the core principles behind the framework.

The Strategic Framework for AI-Search Optimization: 10 Core Principles

Based on testing and analysis of successful AI-visible content, these are the strategic principles that actually move the needle:

1. Focus on Ontological Core, Not Content Volume

The shift isn’t about producing more content—it’s about building a robust semantic foundation. Your ontological core (as introduced by Tony Seal) is the structured representation of essential domain knowledge. It’s what enables AI systems to generate dynamic, context-aware responses.

Since follow-up questions vary by user, you can’t prewrite content for every possibility. Instead, invest in semantic infrastructure that can adapt and scale—anchored in entities, relationships, and intent—so your content stays relevant, no matter the path the conversation takes.

2. Build Dynamic, Conversational Content Architecture

Since follow-up questions are both contextual and stochastic, your content must be conversational and adaptive—not merely static or exhaustive. Rather than chasing keyword coverage, shift your focus to semantic relationships and knowledge graph structures.

Our data shows that content built on a strong ontological foundation can respond to 3x more contextual variations than traditional long-form copy. Every entity relationship opens a new conversational path; every attribute becomes a new potential answer.

3. Prioritize E-E-A-T and Structured Data Rigorously

Google has made it clear: E-E-A-T is critical for AI-generated responses. But beyond traditional trust signals, you also need to provide explicit semantic context through schema markup.

The combination of human credibility and machine readability is now essential for maximum visibility. In an AI Agent–driven world, trust and authenticity aren’t optional—they’re foundational.

4. Adapt to Conversational and Intent-Driven Queries

AI excels at understanding natural language—your content strategy should reflect this by focusing on underlying intent, not just keyword phrases. It’s not about ranking for exact terms anymore; it’s intent and semantic similarity that win the game.

Our initial tests on SGE show that content optimized for conversational queries achieves 40% higher coverage in fan-out simulations.

5. Develop Prediction Models for Contextual Follow-ups

Universal coverage isn’t the goal. Rather than optimizing for individual queries, focus on building data—and eventually models—that can predict likely follow-up questions based on user context.

This is where the validation function becomes essential: it should reward predictions that not only anticipate the next contextual turn but also align with your ontological core—your brand values and marketing position.

The objective is to increase prediction accuracy for user-specific follow-ups, not to cover everything.

6. Excel with “Zero-Click” and “Citation-Based” Visibility

Success metrics are shifting. Being cited in AI-generated answers can be more impactful than traditional clicks—especially for brand authority and consideration. But these are still proxy metrics.

What truly matters is the bottom line: optimize for conversions and leads. That’s the North Star—everything else is secondary.

Track mention frequency and sentiment in AI responses, not just CTR. Then connect the dots: Are these mentions driving trial signups if you’re SaaS? Purchases if you’re e-commerce?

SEO is no longer just about rankings—it’s a function of business development.

7. For E-commerce, Focus on Rich Product Data and Comparative Content

AI-driven search doesn’t just answer direct queries—it supports users through complex decision-making journeys. To win in this environment, your product content must address the comparative sub-questions that AI systems generate: How does this product differ? What are the key features and use cases? Why choose this over another?

Pages enriched with structured data and comparison matrices—highlighting specifications, benefits, and alternatives—perform significantly better. Learn more on how you can convert product data (enriched with GS1 Digital Link) with additional visibility and sales.

8. Monitor Paid Search Performance and Adapt Strategies

AI Overviews are reshaping the search landscape—including paid visibility. Closely monitor CTR fluctuations for query types where AI delivers comprehensive answers, and reallocate budgets where necessary. Always remember that an effective landing page reduces the cost per click.

Prioritize high-intent queries—those where AI responses lead users to take action, not just feel informed.

9. Stay True to Your Values and Listen Strategically

As AI reshapes how we create, optimize, and discover content, staying grounded in your core values is more important than ever. Prioritize transparency, fairness, and privacy—not just for compliance, but to build enduring trust in an increasingly automated world.
At the same time, adopt a clear listening strategy: structure your data infrastructure around well-defined personas. These personas serve as the foundation for inferring the most relevant follow-up questions, guiding content generation and optimization in a way that truly resonates with your audience.

10. Experiment and Iterate Continuously

The AI search landscape evolves rapidly. Use tools like our query fan-out simulator to regularly assess your content’s AI visibility and adapt strategies based on actual coverage data.

The Simulator: Understanding Your Content’s AI-Mode Readiness

Understanding query fan-out is one thing—simulating how your content might be perceived and deconstructed by an AI is the practical first step. Our updated Colab notebook aims to provide this initial insight.

This new release enhances the simulation by:

• Leveraging Gemini for Initial Understanding: Uses Google’s Gemini (via its url_context-like grounding capability or direct analysis) to identify the main entity of your URL and extract key content chunks from the page. This replaces traditional HTML scraping and parsing, getting us closer to how an AI might “see” your page.
• Sophisticated Query Fan-Out Simulation with DSPy CoT: Employs a DSPy Chain-of-Thought (CoT) module powered by Gemini to generate synthetic fan-out queries. This module first reasons about different relevant query types (informed by patent insights like WO2024064249A1) before producing a diverse set of sub-queries.
• Semantic Coverage Scoring: Assesses how well the Gemini-identified content chunks from your URL cover these diverse synthetic queries using embedding-based similarity.

In AI Mode, Google may also retrieve up to five chunks before and after a relevant one to provide context. This means your content should not only make sense at the chunk level, but also flow coherently across sections. A well-structured layout improves both chunk visibility and contextual stitching.

The tool, built with DSPy (my preferred framework for robust AI system programming), allows you to:

1. See what primary entity Gemini extracts from your URL.
2. Inspect the reasoning process (thanks to CoT) behind the generation of fan-out queries.
3. Analyze the types of fan-out queries generated for your entity.
4. Get a coverage score indicating how well your page’s key content (as understood by Gemini) addresses these potential AI sub-queries.

📍 Technical Pipeline:
URL → Entity Extraction → Query Fan-Out → Embedding Coverage → AI Visibility Score

While predicting every exact contextual follow-up is impossible, this tool helps you understand your content’s readiness for an AI-driven search environment that relies on such decomposition. It provides a data point for how “AI-friendly” and “synthesis-ready” your content might be.

🧪 This tool is experimental. The way LLMs, including Gemini, perform grounding and interpret URLs can evolve. We welcome your feedback as we continue to refine this approach with the community.

Access the Query Fan-Out Simulator – A free tool to measure your content’s AI-Mode Readiness.

.

From Coverage to Context: The Real Challenge

The notebook reveals something more profound than coverage gaps—it shows the impossibility of deterministic optimization in an AI-driven search world. Most content addresses only samples of possible follow-ups, and that’s not a bug, it’s the feature.

What matters isn’t covering every possible query variation—it’s building semantic infrastructure that can predict and respond to contextual follow-ups accurately. When you understand that there’s no way to “rank” in the traditional sense, you start focusing on what actually matters: the ontological core.

This is why content becomes dynamic and conversational. In the end, the ontological core—your fundamental knowledge structure about your domain—is all you really need to focus on. Everything else is prediction and adaptation.

The brands that understand this shift from deterministic rankings to contextual prediction will dominate AI search. The question isn’t whether you can cover every query variation—it’s whether you can build semantic infrastructure that predicts and adapts to user context.

Ready to start building your ontological foundation? The simulator is your first step to understanding how contextual follow-ups work—and why traditional SEO thinking no longer applies.

👉 Access the Query Fan-Out Simulator (3rd release) — a free tool to measure your content’s AI search visibility [runs on Google Colab].

👉 Book a demo to see how we can help you transform your content strategy with knowledge graphs and AI-powered optimization.

References:

For further reading on the evolution of Google’s AI Mode and its implications for search and SEO, see:

• Gianluca Fiorelli’s “Why AI Mode Will Replace Traditional Search as Google’s Default Interface” on I Love SEO
• Aleyda Solís’s breakdown of the Query Fan-Out approach
• Dennis Goedegebuure’s perspective on SEOs as natural-born marketers in the age of AI
• Michael King’s in-depth analysis of the AI Mode patent
• The official patent document for “Search with stateful chat” (US20240289407A1)
• The WO2024064249A1 patent “Systems and methods for prompt-based query generation for diverse retrieval”