New Research: What We Learned From Analyzing 60K+ Google Fan-Out Queries - Nectiv

So a couple of months ago, our team at Nectiv figured out how to extract ChatGPT search queries at scale. We then created a data study that analyzed trends on ChatGPT search and how it worked. We found insights such as common search terms, how often it searches, average number of fan-0uts.

Well now we wanted to replicate that same study but this time using Google’s data. After all, Google is still the dominance of the two players and probably has the more advanced fan-out functionality to date. So as a result, we thought it would be a good idea to recreate a similar study but give the SEO community insights with Google’s fan-out data.

Once again, we were able to extract data at scale to create a pretty large study across multiple industries on how Google is likely using fan-out queries. Let’s get into what we found.

P.S. If you’re interested in getting Google’s fan-out queries for your own query/prompt set, feel free to contact us.

Background

So I was recently an article from Dan Petrovic where he showcased a new tool he built. If you’re not following him on LinkedIn/Twitter, you’re doing yourself a massive disservice.

His Query Fan Out tool has a Google Grounding API tab. Add any prompt you want and the tool will extract Google’s own query fan out and show them to you.

When looking into it a bit more, he’s using the Gemini API to pull this data. In case you’ve missed it, Google has dedicated documentation to show you how to extract fan-out queries from Gemini.

One of the outputs from the response is the webSearchQueries which will show you the exact fan-out queries used. You can use Python/JS/REST to create a script to extract the queries. The result is that you’ll get a spreadsheet that lists Google’s fan-out queries along with the original prompt they were sourced from:

If you know anything about me, I like scale. I want to analyze the results for thousands of queries at a time to understand common themes and patterns. So using this tool inspired to perform the same research we did for ChatGPT search queries but using Google’s data instead.

It’s important to note that we don’t know for sure if this is the exact same way AI Mode. However, it’s been confirmed that AI Mode is built with Gemini 3 so it’s definitely possible that the fan-outs act in a similar manner.

As I was performing research for this article, I ran across this article from Nick Haigler of Seer Interactive that reviewed a similar analysis. In the article they reviewed 500 different prompts and analyzed insights from Gemini 3 fan out queries. They found some really interesting insights such as:

10.7 average fan-out queries per prompt
95% of fan-out queries had 0 MSV
6.7 average words per fan out query
21.3% contained a year

I was glad to see this as it created a useful benchmark for our study. We can get a gauge of how similar/different the data is by comparison. We will return to this data later.

Methodology

In Nectiv’s analysis of ChatGPT search queries, we already had a list of nearly 9,000 different prompts across different verticals such as Software, Commerce, Travel, Local and more. We figured it would make sense to use the same prompt set except running this through the Gemini API and using Gemini 3 as it’s the most up to date model.

Using Claude Code, we were able to run that same prompt set through the API to get the fan-out queries at scale.

Since all of the queries grounded through search, this resulted in 70K+ exported rows of data.

We then analyzed the Gemini 3 fan-out queries for different insights.

How Many Fan-Out Queries Does Google Use?

So this was extremely interesting. When looking at the data, we can see that Google can use A LOT of fan-out queries. In fact, the data showed use that Google uses on average 9.06 different fan-out queries per prompt. So it’s not just 2-3 queries that you have to account for, it’s many more.

Looking at the distribution was also fascinating. When breaking things down by “Mode”, we can see that “5 fan-out queries” was actually the most common bucket. All in all 59% of the results contained between 5-11 fan-out queries.

However the “long tail” was not irrelevant here. 24% of all prompts contained between 12-19 fan-out queries. The longest we found were prompts that had as many as 28 fan-out queries so that appears to be the maximum for now.

When comparing our data to the one from Nick H, we can see that the distribution looks almost identical:

In our data, we did see that it is possible to have only two fan-out queries. This tended to be more common in certain industries such as “Local”.

How Do Fan-Out Queries Change By Industry

As we know with a lot of these data studies, segmentation is also a great way to look at the data. We wanted to see how changing the industry impacted the number of fan-out queries. Are some industries susceptible to a larger number of fan-outs than others?

Turns out that this is indeed the case. When segmenting by industry we can see clear differences in how frequently Google uses fan-out queries in some vs others.

Some interesting insights include:

Software tends to have the most number of fan-outs, averaging a whopping 11.7 per prompt
Travel (10.8) and Careers (9.8) tended to have higher number of fan-outs
Local was by far the lowest, averaging only 3.79 fan-outs.

This graph also helps visualize how different each vertical was from the average:

How Many Words Do Google Fan-Out Queries Use?

This shouldn’t be a surprise but Google fan-out queries are generally long – much longer than someone would normally search. On average, the analysis found that fan-out queries are 6.7 words. This is the EXACT number the Seer study also found.

The distribution was also pretty interesting. 77% of the queries were between 5-8 words in length. The distribution actually showed that queries could be up to 19 words in length, but this is more rare.

Here are just a few examples of the types of queries Google will search in that 5-8 word count range:

reformation plus size pink dress reviews
The Glade Rehoboth Beach community features
best job boards for tech jobs 2025
best free home design software reviews 2024
benefits to look for in funeral service jobs

Does The Number Of Words In Fan-Out Queries Change By Industry

I’ll make this quick: Not really.

When breaking down each industry by fan-out query length, there aren’t any clear differences. The vast majority of fan-out queries tended to be in that 6-7 word range across most industries.

The Credit Card/Finance space did seem to the be the only outlier averaging 7.67 words per fan-out.

What Are The Most Common Search Terms Google Fan-Outs Use?

I’m going to talk about the most common search terms since this is one of the most actionable pieces of the data study. By understanding this, we’ll better know which content to create and what to to optimize in the AI search experience. We can find the most common terms by using N-grams as a proxy.

However I do want to caveat that this is going to be highly determined and influenced by the types of prompts used. In our study, we used more commercial/transactional types of prompts to learn what’s influencing product/company recommendations.

Across the dataset we found these were the most common Ngrams:

2024/2025 (6.26%)
Reviews (2.14%)
Vs. (1.41%)
Free (1.05%)
Top (1.05%)

Freshness Is Key

The Gemini fan-out queries are just the latest examples of AI engines prioritizing and looking for fresh content. In our previous study, we found the term “2025” was the most popular Ngram in ChatGPT. So content that’s optimized with the current year in the title and on-page content is going to be critical to be included in the consideration set for AI search.

One curious thing about the Gemini data is how many instances of “2024” there were in the data. At the time of me writing this we’re in November 2025. Every search should be 2025 or even 2026.

However, 45% of all instances of the year used 2024. When diving into this, one technique that Google seems to be doing is performing a search that includes both “2024 and “2025” in order to catch content associated with either year.

So one strategy we might see it employ is searching for both the current and previous year in order to find any content written in that range.

AI Wants Reviews

Also similar to ChatGPT is the fact that Google fan-out seem to be looking for reviews. The fan-out queries clearly aim to go deeper than simply looking at the search results for the products that are listed. They want to take another step and gather information on reviews on the actual best solutions out there.

For instance, in this example for “What is the best martial arts software”, the first fan-out looks for product recommendations with “top rated martial arts school management software”. However, it then proceeds to search a bunch of queries around reviews such as: best martial arts software 2024 2025 reviews, Kicksite pricing and user reviews 2025, Spark Membership software pricing and reviews.

Comparison Content Is Critical

One Ngram that is present in Google’s top 5 that ISN’T for ChatGPT is the term “vs”. In Google’s fan-out queries it’s often looking to identify products and then search for content on the web that directly compares each one.

Probably the most prominent industry we saw this behavior take place in is the Credit Card vertical. Google would often perform really long searches that compares 3-5 options against each other, not just 1-2.

For examples, for the prompt “What is the best credit card for kids?”, Google directly searches “Step vs Greenlight vs GoHenry vs Chase First Banking comparison”

A really specific article optimized for all of these options would likely put that article in the consideration set when this search is performed.

Conclusion

I know the era of AI search seems overwhelming so hopefully this article makes it feel a little more accessible. AI search engines like Google will use query fan-out to perform multiple searches instead of one. This will make things dramatically more complex as exponentially increases the number of queries that you’re trying to optimize for. However, the good news is it is possible to get the data to better understand the exact queries that Google is searching for.

If you’re looking to understand the Google fan-out queries for your prompt set, we’re happy to help.