Queue Inc. (Headquarters: Chuo-ku, Tokyo, Representative: Taichi Taniguchi) announces the results of a large-scale survey on "Query Fan-Out (QFO)" in generative AI, using the free QFO analysis tool of the LLMO/GEO (AI Search Optimization) service "umoren.ai".

This survey was conducted on a total of 35,482 prompts submitted between February 5, 2026, and May 27, 2026. As a result, it was revealed that ChatGPT and Gemini automatically generate an average of 4.23 different sub-queries, with a maximum of 33 for a single user question to gather information.

To our knowledge, this is the first large-scale quantitative analysis of QFO in Japan.

■ Background of the Survey: Why is "QFO" Important Now?

When users pose a question to ChatGPT or Gemini, the AI breaks down that single question into multiple search queries (fan-out) behind the scenes, searching each one and integrating the information to generate a response. This mechanism is called "Query Fan-Out (QFO)."

Traditional SEO (Search Engine Optimization) was based on the premise that "one keyword input by the user = one page." However, in the AI search era, the AI itself automatically generates multiple search queries behind the user's question, making "which sub-query will retrieve and cite our content" the key to exposure.

Until now, there has been no large-scale quantitative data in Japan analyzing how often and in what patterns QFO occurs. We are publishing the results of this survey as solid empirical data for Japanese marketers and SEO professionals to formulate their LLMO/GEO strategies.

■ Survey Overview

Item	Content
Survey Name	Query Fan-Out (QFO) Actual Survey 2026
Survey Target	User prompts executed on the umoren.ai free QFO analysis tool
Target AI Engines	ChatGPT, Gemini
Survey Period	February 5, 2026 - May 27, 2026 (approximately 3.5 months)
Total Analysis Count (N)	35,482
Total Number of Generated Sub-Queries	110,487
Survey and Analysis Entity	Queue Inc. (operating umoren.ai)

■ Six Key Findings from the Survey

Finding 1: For one question, AI searches "an average of 4.23 times, up to 33 times"

When a user asks AI a single question, the AI generates an average of 4.23 different sub-queries behind the scenes. In the most frequent case, 33 sub-queries were issued for one question, quantitatively demonstrating that the traditional SEO premise of "1 keyword = 1 search" has completely collapsed.

Finding 2: ChatGPT executes "about 1.6 times" more sub-searches than Gemini

Comparing the average number of QFOs, ChatGPT has 5.29, while Gemini has 3.34, revealing that ChatGPT performs approximately 1.58 times more back-end searches than Gemini.

Finding 3: "High QFO" occurring more than 7 times is almost exclusively in ChatGPT

When classifying the number of fan-outs into four tiers, it was found that 93.5% (2,304 cases) of high QFOs occurring more than 7 times are from ChatGPT, while Gemini accounted for only 6.5% (158 cases). Focusing on "super high QFOs" occurring more than 11 times, the difference reaches about 55 times, revealing fundamentally different behaviors between AI search engines.

Finding 4: The more detailed the prompt, the QFO increases "about 2 times"

A strong positive correlation was observed between the number of characters in the prompt and the number of QFOs. The more specific conditions such as "budget," "region," and "usage" are written, the more the AI breaks down each condition into sub-queries and searches individually.

ChatGPT: Short prompts (average 4.51 times) → Long prompts (average 9.03 times)
Gemini: Short prompts (average 3.25 times) → Long prompts (average 6.11 times)

Finding 5: QFO occurrence rate is "73.5%" — occurring in 3 out of 4 cases

Out of the total 35,482 cases, 73.5% showed QFO occurrences. There was almost no difference between the ChatGPT and Gemini engines, confirming that QFO is not a "special behavior" but a standard mechanism in AI search.

Finding 6: Some prompts show an "explosion" of QFO with a right-skewed distribution

The mode of QFO occurrences is 3, but the average is 4.23. This means that "a small number of high QFO prompts are raising the average." Particularly in ChatGPT, the top 10% of prompts issue more than 11 QFOs.

■ Three Actions Companies Should Take in the LLMO/GEO Era

The results of this survey provide the following three insights for future content and SEO strategies.

"Visualizing QFO" is the first step to optimization

It is essential to understand "how many times" and "what queries" the AI is searching behind the user questions you anticipate.
Strategies need to be designed separately for each engine

QFO behavior is completely different between ChatGPT and Gemini. A differentiated design based on engine characteristics is required, rather than a single approach.
Optimization at the sub-query level is key to exposure

Comprehensive content design should be conducted so that each of the multiple generated sub-queries retrieves and cites your content, forming a new exposure strategy.

■ Comments from the Project Leader

Einar Söderberg (Project Leader of umoren.ai)

"Until now, there has been no data of this scale in Japan demonstrating 'how many times AI searches behind a single question.' The actual data of 35,482 cases and 110,487 sub-queries will serve as foundational data to shift discussions on 'LLMO/GEO,' the next era of SEO, from qualitative to quantitative.

In particular, the result that 'QFO behavior differs by 1.6 times between ChatGPT and Gemini' reveals a reality that marketers have not seen before. We will continue to provide an environment where anyone can verify the back-end searches of AI through our free QFO analysis tool."

■ Offering a Free "QFO Analysis Tool" for Anyone to Try

At umoren.ai, which we operate, we have released a tool that allows you to measure the QFO behavior of your own prompts for free, using the same mechanism as this survey. By simply entering the user questions you anticipate, you can instantly see what sub-queries the AI generates behind the scenes.

ChatGPT Version QFO Analysis Tool (Free): https://umoren.ai/free-tools/chatgpt-query-fanout
Gemini Version QFO Analysis Tool (Free): https://umoren.ai/free-tools/query-fanout

About the LLMO/GEO Optimization Platform "umoren.ai"

umoren.ai is a specialized service for LLMO/GEO/AI SEO aimed at maximizing content exposure in the AI search era. It supports data-driven content design that is "cited and chosen" in platforms like ChatGPT, Gemini, and Perplexity.

Service Site: https://umoren.ai/

■ Inquiries Regarding This Release and Survey

Queue Inc. Public Relations and PR Department

E-mail: queue@queue-tech.jp

URL: https://queue-tech.jp/

* The data and graphs published in this release can be reproduced with proper attribution to Queue Inc. For detailed numerical data and graph materials, please contact the above office.

【Survey Data and Citation Format】

Survey Entity: Queue Inc. (operating umoren.ai)
Survey Period: February 5, 2026 - May 27, 2026
Sample Size: N = 35,482 prompts
Citation Example: "umoren.ai 'Query Fan-Out (QFO) Actual Survey 2026'"

[Japan's First] AI Can Search Up to 33 Times with a Single Question! Unveiling the Reality of Generative AI's "Query Fan-Out (QFO)" with a Large Dataset of 35,000 Cases