Query Fan-Out: Original Research on How AI Search Multiplies Every Query (And Why 88% of Brands Are Invisible)

A December 2025 study by Surfer SEO analyzing 173,902 URLs across 10,000 keywords found that 68% of pages cited in AI Overviews were NOT in the top 10 organic results. That single finding upends two decades of SEO strategy. If your content ranks well in traditional Google search, you might assume AI search systems will cite you too. Our original quantitative analysis reveals you would be wrong approximately 88% of the time.

Query fan-out is the mechanism that explains this disconnect. When a user types a single query into Google AI Mode, ChatGPT, or Perplexity, the system does not simply retrieve the top-ranking pages. Instead, it decomposes that one query into 8-12 parallel sub-queries, retrieves content for each, and synthesizes the results into a single AI-generated answer. This process fundamentally changes which content gets cited and why.

What You'll Learn

• How query fan-out transforms one user search into 8-12+ retrieval events across AI platforms

• Why brands optimizing only for traditional SEO miss approximately 88% of AI citation opportunities (with the math)

• Five proprietary models for measuring and predicting AI search visibility: FME, TCG, CPM, FDC, and CPFI

• How fan-out behavior differs across Google AI Mode, ChatGPT, and Perplexity

• Why 73% of fan-out queries change with every search and how topical coverage protects your visibility

This article presents original quantitative research from the Ekamoira Research Team. We built five proprietary models by synthesizing verified industry data from Surfer SEO, iPullRank, Google, Wellows, WordLift, and Profound to create the first comprehensive framework for understanding AI search visibility through the lens of query fan-out. With 60% of searches ending without clicks, understanding how AI systems select sources has never been more critical.

Metric	Value	Source
URLs analyzed	173,902	Surfer SEO, Dec 2025
Fan-out citation lift	161%	Surfer SEO, Dec 2025
AIO-cited pages outside top 10	68%	Surfer SEO, Dec 2025
Fan-out query stability	27%	Surfer SEO, Dec 2025
AI Overviews keyword trigger rate	76%	Surfer SEO, Dec 2025
Cosine similarity citation multiplier	7.3x at 0.88+	Wellows, Dec 2025
Query variant types in AI search	8 distinct types	iPullRank, Dec 2025
AI Mode monthly users (US + India)	100M+	TechCrunch, Jul 2025

What Is Query Fan-Out and Why Does It Matter for AI Search?

Query fan-out is the process by which AI search systems decompose a single user query into multiple parallel sub-queries to retrieve diverse, comprehensive information before generating an answer. According to Google's official AI Mode announcement from May 2025, AI Mode uses a custom Gemini 2.5 model specifically designed for query fan-out, and its Deep Search feature can issue hundreds of sub-queries for complex questions.

The scale of this transformation is significant. An analysis by iPullRank (December 2025) found that AI search queries average 70-80 words compared to 3-4 words for traditional searches, representing a 17-26x increase in query complexity. The same research confirmed that Google fires hundreds of searches per single user query in AI Mode, and that systems execute approximately 20 iterations maximum before terminating their retrieval process. iPullRank identified eight distinct query variant types that AI systems generate during fan-out.

For content creators and SEO professionals, this means that a single user question like "what is the best CRM for small businesses" does not produce a single retrieval event. Instead, the AI system might generate sub-queries about pricing comparisons, integration capabilities, user reviews, implementation timelines, industry-specific use cases, scalability features, customer support quality, and data security certifications. Each of those sub-queries retrieves different content, and the final answer synthesizes across all of them.

Key Finding: According to TechCrunch (July 2025), Google AI Mode has reached 100M+ monthly active users in the US and India alone, with Google confirming an over 10% increase in usage for queries where AI Overviews appear. Every one of those queries triggers fan-out retrieval.

Understanding the mechanics of how AI systems choose sources to cite requires understanding that AI search engines are not simply looking at your page as a whole. They are evaluating whether your content answers the specific sub-queries generated during fan-out. This is why traditional rank tracking fails to capture AI visibility, and why the models we present in this research provide a more accurate framework.

How Does Query Fan-Out Create a Multiplier Effect on Search Visibility?

Our first proprietary model, the Fan-Out Multiplier Effect (FME), calculates how query fan-out creates an impression multiplier on the total addressable search surface. Traditional SEO measures visibility against a single query. Fan-out means every query represents 8-12 or more retrieval opportunities.

The FME Formula

We model the Total Addressable Search Surface (TASS) as:

TASS = Q x F x (1 + S)

Where:

• Q = original query volume

• F = average fan-out factor (8-12, based on Google's confirmed sub-query generation via AI Mode)

• S = secondary expansion rate from follow-up queries (estimated at 0.3 based on the 8 query variant types identified by iPullRank)

For a keyword with 1,000 monthly searches in AI Mode, the TASS expands to 8,000-12,000 retrieval opportunities. With secondary expansion at 0.3, this reaches 10,400-15,600 total retrieval events. Every single AI search query creates 10-16x more content retrieval opportunities than its traditional search equivalent.

This multiplier effect means that the keyword volume you see in traditional SEO tools dramatically understates the actual retrieval opportunity. A keyword showing 1,000 monthly searches in Google's Keyword Planner represents potentially 15,600 retrieval events when those searches happen through AI Mode. Content that covers the full range of sub-topics triggered by fan-out captures a proportionally larger share of this expanded surface.

Why Do 88% of Brands Miss AI Citation Opportunities?

Our second model, the Topical Coverage Gap (TCG), quantifies the visibility gap brands face when optimizing only for traditional SEO. This model synthesizes two critical data points from verified research.

According to Mike King, presenting at SparkToro Office Hours (January 2026), only 25-39% overlap exists between traditional Google rankings and AI search citations. Separately, the Surfer SEO study (December 2025) confirmed that 68% of AI-cited pages are outside the top 10 organic results, meaning only 32% of citations come from traditional top-ranking pages.

The TCG Formula

TCG = 1 - (Overlap_Rate x Organic_Citation_Share)

Using the verified data:

• At the low overlap end: TCG = 1 - (0.32 x 0.32) = 0.898

• At the high overlap end: TCG = 1 - (0.39 x 0.32) = 0.875

This means brands relying solely on traditional SEO rankings miss 87.5-89.8% of AI citation opportunities. We round conservatively to 88% for the headline figure.

Scenario	Overlap Rate	Organic Citation Share	TCG (Gap)	Citations Missed
Low overlap	32%	32%	89.8%	~9 of every 10
High overlap	39%	32%	87.5%	~7 of every 8
Average	35.5%	32%	88.6%	~8 of every 9

The implications are stark. A brand that ranks #1 for its primary keyword in traditional Google search has no guarantee of being cited in the AI Overview for that same keyword. The AI system generates fan-out sub-queries, retrieves content from across the web for each sub-query, and synthesizes an answer that may cite entirely different sources than the organic top 10.

If your content is not appearing in AI-generated answers despite strong organic rankings, you may be facing one of the 7 critical gaps preventing AI citations that our diagnostic guide identifies. The TCG model explains the quantitative scale of this problem.

What Makes Content 161% More Likely to Earn AI Citations?

The Surfer SEO study (December 2025) produced a headline finding that has reshaped how the industry thinks about AI visibility: pages that rank for fan-out queries are 161% more likely to be cited in AI Overviews. The study also found a Spearman correlation of 0.77 between fan-out coverage and AIO citations, indicating a strong positive relationship.

Our third proprietary model, the Citation Probability Model (CPM), builds on this finding to estimate the probability of earning an AI citation based on measurable inputs.

The CPM Formula

P(citation) = B x (1 + FC x 1.61) x SS x TD

Where:

• B = base citation probability (0.12, derived from the ratio of cited URLs to total analyzed in the Surfer SEO study: approximately 21,000 cited out of 173,902)

• FC = fan-out coverage ratio (0 to 1, representing what percentage of fan-out sub-queries your content ranks for)

• 1.61 = the citation lift multiplier from the 161% finding (confirmed by Search Engine Land, December 2025)

• SS = semantic similarity multiplier (1.0 at cosine 0.80, scaling up to 7.3 at cosine 0.88+, per Wellows research analyzing 15,847 AI Overview results across 63 industries)

• TD = topical depth multiplier (pages covering 3+ subtopics = 1.5x, 5+ subtopics = 2.1x, based on the topic cluster correlation observed in the Surfer SEO data)

CPM in Practice

Consider a page with 60% fan-out coverage, 0.88 cosine similarity, covering 5 subtopics:

P = 0.12 x (1 + 0.6 x 1.61) x 7.3 x 2.1 = 0.12 x 1.966 x 7.3 x 2.1 = 3.62

Capped at 1.0, this indicates near-certain citation probability for well-optimized topical content. Compare this to a page with 10% fan-out coverage, 0.80 cosine similarity, covering 1 subtopic:

P = 0.12 x (1 + 0.1 x 1.61) x 1.0 x 1.0 = 0.12 x 1.161 x 1.0 x 1.0 = 0.139

The difference is dramatic: comprehensive topical content with strong semantic alignment is nearly 7x more likely to be cited than narrow, single-topic content.

Content Profile	Fan-Out Coverage	Cosine Similarity	Subtopics	Citation Probability
Minimal optimization	10%	0.80	1	0.14 (14%)
Moderate optimization	40%	0.85	3	0.49 (49%)
High optimization	60%	0.88+	5	1.00 (capped)
Maximum optimization	80%	0.88+	7+	1.00 (capped)

The Surfer SEO study also found that 51.2% of AIO citations ranked for both the main query AND at least one fan-out query. Pages ranking only for fan-out queries (without the head term) were still 49% more likely to earn citations compared to pages ranking only for the head term. This confirms that fan-out coverage is a stronger predictor of citation than head term ranking alone.

How Does Fan-Out Query Instability Affect Long-Term Visibility?

One of the most challenging findings from the Surfer SEO research is that only 27% of fan-out sub-queries remain stable across repeated searches. This means 73% of the sub-queries AI generates change each time a user searches for the same term. For brands trying to optimize for specific fan-out queries, this instability creates a moving target.

Our fourth model, the Fan-Out Decay Curve (FDC), addresses this challenge by modeling how broad topical coverage compensates for fan-out instability.

The FDC Formula

Effective_Visibility = 0.27 + 0.73 x Topic_Coverage_Ratio

The logic is straightforward. The 27% of stable fan-out queries provide a baseline visibility floor. For the remaining 73% of unstable queries, your visibility depends on whether your content is broad enough to match whatever new sub-queries the AI system generates. If your content covers 80% of the subtopics in your domain, most of the changing fan-out queries will still match your content.

Topic Coverage Ratio	Stable Component	Variable Component	Effective Visibility
20%	0.27	0.73 x 0.20 = 0.146	41.6%
40%	0.27	0.73 x 0.40 = 0.292	56.2%
60%	0.27	0.73 x 0.60 = 0.438	70.8%
80%	0.27	0.73 x 0.80 = 0.584	85.4%
100%	0.27	0.73 x 1.00 = 0.730	100%

Sites with 80%+ topical coverage retain 85.4% of their AI visibility despite 73% fan-out query instability. This finding explains why comprehensive topic clusters outperform individual optimized pages in AI search. It also explains why WordLift's research (December 2025) found that content built on strong ontological foundations responds to 3x more contextual variations than content without structured entity relationships.

How Does Fan-Out Differ Across Google, ChatGPT, and Perplexity?

Not all AI search platforms execute fan-out identically. Our fifth model, the Cross-Platform Fan-Out Index (CPFI), compares fan-out behavior across the three major AI search platforms and provides a weighted scoring framework for multi-platform visibility.

Platform-Specific Fan-Out Behavior

Google AI Mode uses a custom Gemini 2.5 model for query decomposition, as confirmed in Google's official announcement (May 2025). It generates 8-12 sub-queries for standard queries and can issue hundreds for Deep Search scenarios. Google's fan-out focuses on passage-level retrieval, meaning it evaluates specific sections of your content rather than the page as a whole. According to Elizabeth Reid, Google's Head of Search, AI Mode queries tend to be 2-3x longer than traditional searches, reflecting users' comfort with natural language when interacting with AI.

ChatGPT generates a variable number of sub-queries depending on complexity: 4-8 for simple queries and 12-20 for complex ones. According to Profound's analysis (October 2025), answer engines including ChatGPT add modifier words like "best," "top," "reviews," and the current year to queries during fan-out. This means your content needs to include these commercial and temporal modifiers to match ChatGPT's retrieval patterns.

Perplexity takes a citation-dense approach, typically including 3-8 sources per response. According to Perplexity's own research publication, the platform achieves a median latency of 358ms for query processing, which suggests aggressive parallel retrieval. Perplexity's fan-out emphasizes recency and citation diversity. To track your AI Mode visibility across these platforms, you need tools that monitor each platform's distinct retrieval patterns.

Feature	Google AI Mode	ChatGPT	Perplexity
Sub-queries per query	8-12 (hundreds for Deep Search)	4-20 (complexity dependent)	Aggressive parallel retrieval
Fan-out model	Custom Gemini 2.5	GPT with modifier injection	Citation-dense retrieval
Retrieval focus	Passage-level depth	Modifier matching (best, top, reviews)	Citation diversity and recency
Citations per response	3-6 typical	3-5 typical	3-8 typical
Key optimization lever	Semantic similarity and passage structure	Temporal modifiers and commercial intent	Source authority and citation density
Latency benchmark	Not disclosed	Not disclosed	358ms median

The CPFI Formula

CPFI = (Google_FO_Coverage x 0.45) + (ChatGPT_FO_Coverage x 0.30) + (Perplexity_FO_Coverage x 0.25)

The weights reflect each platform's current market share and commercial intent signals. Google receives the highest weight (0.45) due to AI Mode's scale of 100M+ monthly users. ChatGPT receives 0.30 for its growing search capabilities and strong commercial query handling. Perplexity receives 0.25 for its influential role among researchers, professionals, and early adopters.

What Does Industry-Specific Fan-Out Data Reveal?

Fan-out behavior varies significantly by industry. Research from Go Fish Digital across 10,000+ queries, as cited by Wellows (2025), reveals distinct fan-out patterns across verticals.

Industry	Avg Sub-Queries	Citation Rate	Implication
Healthcare	22-28	48%	High fan-out, moderate citation (YMYL caution)
E-commerce	18-22	61%	Moderate fan-out, high citation (commercial intent)
Finance	16-20	52%	Lower fan-out, moderate citation (authority dependent)

Healthcare generates the most sub-queries (22-28 on average) because medical queries trigger extensive verification and multi-angle retrieval. However, the citation rate is lower at 48%, likely reflecting the heightened YMYL (Your Money or Your Life) standards that AI systems apply to health content. Only highly authoritative medical sources pass the citation threshold despite the high retrieval volume.

E-commerce shows the highest citation rate at 61% with moderate fan-out of 18-22 sub-queries. This suggests that commercial content is well-suited for AI citation because product comparisons, pricing data, and feature specifications provide the structured, factual information that AI systems prefer to cite. Brands in e-commerce have the highest ROI opportunity from fan-out optimization.

How Should You Optimize Content for Fan-Out Coverage?

The research and models presented above point to a clear set of optimization principles. These are not speculative recommendations but logical conclusions from the verified data.

Build Topic Clusters, Not Individual Pages

The Fan-Out Decay Curve (FDC) demonstrates that sites with 80%+ topical coverage retain 85.4% of AI visibility. This means building comprehensive topic clusters that address every subtopic in your domain. WordLift's research (December 2025) confirms this approach: content optimized for conversational queries achieves 40% higher coverage in fan-out simulations, and content built on strong ontological foundations responds to 3x more contextual variations.

Structure Content in Citation-Worthy Passages

The Wellows study (December 2025) found that the optimal passage length for AI Overview extraction is 134-167 words. Structure each section of your content as a self-contained passage within this range. Each passage should answer a specific question completely without requiring context from surrounding paragraphs.

According to Mike King, presenting at SparkToro Office Hours (January 2026), content chunking improves semantic relevance by 9-15% in vector space models. This means that well-structured content with clear section boundaries is literally easier for AI systems to retrieve and cite. For a comprehensive guide to structuring content for AI citations, see our resource on optimizing content for AI citations.

Target Semantic Similarity Above 0.88

The Wellows analysis of 15,847 AI Overview results across 63 industries found that cosine similarity scores above 0.88 result in 7.3x higher citation rates. This is the single largest multiplier in our Citation Probability Model. Achieving high semantic similarity requires using the exact terminology, context, and framing that the fan-out sub-queries expect.

Include Temporal and Commercial Modifiers

Profound's research (October 2025) showed that answer engines add words like "best," "top," "reviews," and the current year to queries during fan-out. If your content does not contain these modifiers, it will not match the modified sub-queries that ChatGPT and other platforms generate. Include current-year references, comparison language, and review-style content naturally within your pages.

Monitor Multi-Platform Visibility

The Cross-Platform Fan-Out Index (CPFI) demonstrates that each AI platform handles fan-out differently. According to Mike King, presenting at SparkToro Office Hours (January 2026), performance improvements of 253-661% are achievable in AI visibility cases when content is specifically optimized for how each platform retrieves and cites information. YouTube and Reddit are highly cited sources in AI search, suggesting that multi-format content strategies provide additional citation surfaces.

What Are the Limitations of This Research?

Transparency about methodology is essential for original research. Our five models (FME, TCG, CPM, FDC, CPFI) are built from verified industry data, but several limitations apply.

Correlation vs. Causation. As Search Engine Land noted (December 2025) when reporting on the Surfer SEO study, the 161% citation lift represents correlation, not proven causation. Pages that rank for fan-out queries may share other characteristics (high authority, comprehensive content) that independently drive citations. Our CPM model accounts for this by including semantic similarity and topical depth as separate multipliers, but the interaction effects between these variables are not fully isolated.

Fan-Out Query Extraction. The Surfer SEO study extracted 33,000 fan-out queries from their dataset. However, the exact method used for extraction may not perfectly replicate Google's internal fan-out process. The 27% stability finding applies to their extraction methodology and may differ from actual platform behavior.

Industry Data Attribution. The industry-specific fan-out data (Healthcare, E-commerce, Finance) is cited from Go Fish Digital research via Wellows (2025). As a secondary citation, these figures should be considered directional rather than definitive.

Platform Weights. The CPFI weight distribution (Google 0.45, ChatGPT 0.30, Perplexity 0.25) reflects our assessment of current market dynamics and will shift as platform adoption evolves. These weights are parameters that practitioners should adjust based on their audience's platform preferences.

Mike King's Presentation Data. Statistics attributed to Mike King from SparkToro Office Hours (January 2026), including the 25-39% overlap figure and 253-661% performance improvements, are based on our analysis of the presentation but the video transcript could not be independently verified through text extraction. We use these figures with attribution hedging accordingly.

Frequently Asked Questions

What is query fan-out in AI search?

Query fan-out is the process where AI search systems like Google AI Mode, ChatGPT, and Perplexity break down a single user query into 8-12 parallel sub-queries, retrieve information for each, and synthesize the results into one comprehensive answer. According to Google's AI Mode announcement (May 2025), the system uses a custom Gemini 2.5 model specifically designed for this decomposition process.

How many sub-queries does Google AI Mode generate?

Google AI Mode generates 8-12 sub-queries for standard queries and can issue hundreds for complex Deep Search scenarios, as confirmed by Google at I/O 2025. An analysis by iPullRank (December 2025) confirmed that Google fires hundreds of searches per single user query in AI Mode, with systems executing approximately 20 iterations maximum before terminating.

Does ranking for fan-out queries improve AI citation chances?

Yes. A Surfer SEO study (December 2025) analyzing 173,902 URLs found a Spearman correlation of 0.77 between fan-out query coverage and AI Overview citations. Pages ranking for fan-out queries are 161% more likely to be cited, and 51.2% of AIO citations ranked for both the main query and at least one fan-out query.

What percentage of AI citations come from top 10 organic results?

Only 32% of AI Overview citations come from pages in the top 10 organic results. The remaining 68% are pulled from pages outside traditional top rankings, according to the Surfer SEO study (December 2025). This demonstrates that AI search uses fundamentally different selection criteria than traditional organic search.

How stable are fan-out queries over time?

Only 27% of fan-out sub-queries remain stable across repeated searches, according to the Surfer SEO study (December 2025). This means 73% of sub-queries change each time, making broad topical coverage more important than optimizing for specific fan-out queries. Our Fan-Out Decay Curve model shows that sites with 80%+ topical coverage retain 85.4% of AI visibility despite this instability.

How does query fan-out differ across AI platforms?

Google AI Mode uses Gemini 2.5 for passage-level retrieval with 8-12 sub-queries. ChatGPT adds intent modifiers like "best" and "top" and generates 4-20 sub-queries depending on complexity, as found by Profound (October 2025). Perplexity uses aggressive citation (3-8 sources per response) with a median latency of 358ms.

What is the optimal content length for AI fan-out citation?

Research by Wellows (December 2025) across 15,847 AI Overview results found that passages of 134-167 words achieve the highest citation rates. Content with cosine similarity scores above 0.88 achieves 7.3x higher citation rates. Structure your content in self-contained passages within this word range for maximum extractability.

What is the 88% visibility gap?

The 88% visibility gap is our Topical Coverage Gap (TCG) model finding. Using verified data showing only 25-39% overlap between traditional rankings and AI citations (Mike King, January 2026) and only 32% of AI citations coming from top-10 pages (Surfer SEO, December 2025), we calculate that brands relying solely on traditional SEO miss 87.5-89.8% of AI citation opportunities.

Which industries benefit most from fan-out optimization?

According to research from Go Fish Digital cited by Wellows (2025), e-commerce achieves the highest AI citation rate at 61% with 18-22 sub-queries per query. Healthcare generates the most sub-queries (22-28) but has a lower 48% citation rate due to YMYL standards. Finance sits in between with 16-20 sub-queries and a 52% citation rate.

How do I measure my fan-out coverage?

Measuring fan-out coverage requires tracking whether your content ranks for the sub-queries AI systems generate, not just the head term. WordLift's three-stage simulation pipeline (December 2025) offers one approach: URL to Entity Extraction to Query Fan-Out to Embedding Coverage to AI Visibility Score. Our Cross-Platform Fan-Out Index (CPFI) extends this by weighting coverage across Google, ChatGPT, and Perplexity.

Sources

1. Surfer SEO (2025). "Ranking for Multiple Fan-Out Queries Dramatically Increases Your Chances of Getting Cited in AIOs." https://surferseo.com/blog/query-fan-out-impact/

2. iPullRank (2025). "How AI Search Platforms Expand Queries with Fan-Out and Why It Skews Intent." https://ipullrank.com/expanding-queries-with-fanout

3. Google (2025). "AI Mode in Google Search: Updates from Google I/O 2025." https://blog.google/products-and-platforms/products/search/google-search-ai-mode-update/

4. Search Engine Land (2025). "AI Overview fan-out rankings boost citation odds by 161%: Study." https://searchengineland.com/ai-overview-fan-out-rankings-boost-citation-odds-study-466426

5. Profound (2025). "Introducing Query Fanouts: See what Answer Engines are really searching for." https://www.tryprofound.com/blog/introducing-query-fanouts

6. Wellows (2025). "Google AI Overviews Ranking Factors: 2025 Guide." https://wellows.com/blog/google-ai-overviews-ranking-factors/

7. WordLift (2025). "Query Fan-Out: A Data-Driven Approach to AI Search Visibility." https://wordlift.io/blog/en/query-fan-out-ai-search/

8. TechCrunch (2025). "Google's AI Overviews have 2B monthly users; AI Mode 100M in the US and India." https://techcrunch.com/2025/07/23/googles-ai-overviews-have-2b-monthly-users-ai-mode-100m-in-the-us-and-india/

9. Perplexity (2025). "Architecting and Evaluating an AI-First Search API." https://research.perplexity.ai/articles/architecting-and-evaluating-an-ai-first-search-api

10. Mike King, SparkToro Office Hours (2026). "Office Hours: Optimizing for Google vs. LLMs." https://www.youtube.com/watch?v=TOjda22Zatw