WTF is ‘query fan-out’ in Google’s AI mode?

The way AI Mode retrieves that information means more searches are happening behind the scenes, making it tricky for publishers to determine how to optimize their editorial content strategy to appear in those search results. Four heads of SEO at publishers told Digiday they are still in preliminary tests with the feature.

We’ll break down how “query fan-out works” and its implications.

WTF is query fan-out?

When a user types a question into AI Mode, the AI model breaks down that query into multiple search queries around related subtopics. Query fan-out looks at the “subintents” behind a search query, according to Mike King, CEO and founder of SEO agency iPullRank.

For example, if someone typed “best sneakers for walking” into AI Mode, it could break that query up into subqueries like “best sneakers for men,” “best sneakers for walking in different seasons,” “sneakers for walking on a trail” and “best slip-on sneakers.” 

This all happens behind the scenes, in real time. The AI model pulls information from the web related to the original query, and generates a synthesized answer that includes all that information. 

“A user didn’t ask for it in the query but [the AI model is] already predicting all of this information is going to be useful and gives it in a direct answer,” said Adithya Hemanth, SEO lead at marketing agency Incubeta.

This mechanism (which is also the backbone of AI Overviews) allows AI Mode to answer longer, more complex queries — such as “book a vacation for a family of 5 around the U.S.” The query fan-out might entail searches like “family-friendly activities,” “road trip for families” and “family of 5 travel ideas.” 

“In the background, it’s extrapolating what are all the next steps that the user might care about when searching for this search,” said Lily Ray, vp of SEO strategy and research at performance marketing agency Amsive. 

The idea is query fan-out anticipates what people might ask next in search, likely with the goal of keeping users within the search experience, Ray added.

How is this different from the way traditional search works?

Traditional search is based on keywords. One search query surfaces one set of search results. With AI Mode, one search query sparks multiple, different queries that surface many sets of results, according to Olaf Kopp, co-founder and head of SEO at German online marketing agency Aufgesang.

In addition, traditional search looks at the entirety of a webpage to give results. But in query fan-out, relevant “chunks” — or passages — from pages can be fed into the generated answers, he explained.

Google also doesn’t share which queries are being processed in the query fan-out from an original query.

“This is new in the visualization sense, but logically it’s what we should’ve all been doing anyway,” said Mollie Ellerton, head of SEO at digital optimisation agency Hookflash. “And now we’re dealing a lot more with the unknown. We’re optimizing for the unknown.”

What does this mean for publishers?

In AI Mode, ranking for a single keyword is no longer the goal behind search optimization. It’s now about having useful information around many sub-queries, the heads of SEO interviewed for this story told Digiday. 

Ultimately, this also means fewer clicks to publishers’ sites. Instead of a user typing in all those additional searches, giving publishers the opportunity to get surfaced in search results and the chance for a user to click a blue link to their site — all those searches are happening for that user, without them even seeing it, with the information given in a summarized answer. 

“I play with AI Mode every day. It’s hard to get an external click right now. It’s clearly designed in a way to discourage any external linking,” Ray said.

Hemanth said some users will still want deeper analysis or information and click through to a website from AI Mode. “But the majority of the intent would be captured by the AI Mode response, which obviously implies to publishers that users would be less likely to click on their evergreen stories,” he said.

The outlook is bleak for publishers, according to those in the SEO community interviewed for this story. The reality is that search is going to send less referral traffic to their sites. 

“SEO is typically a performance channel but now it’s more about brand performance,” Ellerton said.

Can publishers optimize for this?

Because Google isn’t sharing any data around query fan-out searches, heads of SEO are turning to Google search features like PeopleAlsoAsk, as well as third-party tools, such as AlsoAsked and Profound, they told Digiday. The focus is shifting from measuring click-throughs to using tools to measure visibility in AI search engines, according to King.

But there aren’t many tools available yet.

“AI Mode’s response is a function of synthesis — not what ranks for this one query,” King said. “In SEO we don’t have tools to support this… [We’re] not prepared in the software that’s available. A lot of people have to build custom tooling to be able to support this sort of work. SEO as we know it isn’t enough for this.”

King built a tool called Qforia to replicate the query fan-out, based on prompts in Gemini. When you put in a search question, it shows 20-30 examples of other related queries. A few hundred people are using this tool, King said. One head of SEO at a large publisher said they’re using Qforia, and also developing their own internal tool.

Otherwise, the only SEO strategy in place right now is the basic one (such as E-E-A-T, or “Experience, Expertise, Authoritativeness and Trustworthiness,” Google’s framework to measure content quality), according to the head of SEO at a publisher, who requested to speak anonymously. 

“It’s a bit of everything we’ve already been doing but in a much more focused and structured manner,” Hemanth said.

So how do editorial teams need to think about this? 

Content has to be easy for AI models to understand — with clear, specific language, structure and bullet points. Passages in webpages should ideally be two to four sentences long, according to Kopp.

SEO editorial content strategy now needs to focus on topics and subtopics that answer questions likely to be part of a user’s search journey. Publishers need to identify potential keywords, look at their rankings for those keywords, and optimize passages that Google is most likely to use, according to King. That means splitting up paragraphs so that each one is focused on one specific, clear topic. 

It also helps if content showcases real-world experiences, such as videos of people trying on products in reviews, and includes original research, data and reporting, Hemanth noted.  

“Trying to compete for volume keywords is not happening anymore. We have to focus on the specific journey our customers might be on,” King said.

In other words, publishers need to anticipate sub-queries and different search intents. For example, if the keyword is “Denver travel,” content should also answer queries such as “Denver flights,” “how to get a hotel in Denver,” “how to get from the Denver airport to the city,” “best places to visit in Denver,” “business trip in Denver,” Kopp said.

“You need to be seen as a website that has useful information for the user. It requires a shift in mindset,” Hemanth said. “You can’t just be the best page but you need to have the best answers.”

Ellerton is taking another approach and looking at user-generated content platforms like Reddit and TikTok to figure out what people are searching for.

“We don’t know how people are searching [because] Google doesn’t share that. So we have to get this data from somewhere else,” she said.