The 2003 Framework That Was Already Doing Query Fan-Out

I used to be doing query fan-out by hand for twenty years earlier than anybody referred to as it that. I simply referred to as it one thing else: “Russian nesting dolls.”

In August, MJ Cachón revealed a dataset research. She ran 189 branded prompts by means of ChatGPT, watched the mannequin fireplace off 1,797 sub-queries no person typed. Studying it, I acknowledged the form of one thing I’d been constructing into press releases since early 2003, lengthy earlier than “fan-out” or “GEO” existed as phrases.

After I used to optimize a launch, I seemed for a four-word phrase with a three-word phrase nested inside it, the way in which a Matryoshka doll holds a smaller doll inside an even bigger one. Say the smaller phrase was “airfare to Philadelphia.” I wouldn’t cease there. I’d discover the four-word model, one thing like “low-cost airfare to Philadelphia,” and construct the discharge round that as an alternative. Use the four-word phrase, and the discharge could be discovered by somebody looking the shorter time period or the longer one. Use solely the three-word time period, and also you’re invisible to anybody who sorts the longer phrase, as a result of it’s lacking a phrase your web page by no means mentions.

That’s not a intelligent trick. It’s simply constructed for the slice of search that almost all SEOs undervalue: The queries no person has typed but.

The 15% No one Can Goal Straight

Google first put a determine on it in its 2019 introduction to BERT, writing that 15% of the queries it sees on a given day are ones it has never encountered before. At Search Central Dwell NYC in March 2025, John Mueller revisited the number and appeared nearly bothered by how cussed it’s. “I’d have thought in some unspecified time in the future a lot of the searches would have been made; individuals simply ask the identical factor time and again,” he stated. “However after we recalculate these metrics, it’s all the time round 15%.” He’d anticipated massive language fashions to push the determine greater. They haven’t. It simply sits there, a set fraction of a rising pile of searches, decade after decade.

Multiply that 15% in opposition to the billions of queries Google handles day by day, and also you get a whole bunch of hundreds of thousands of searches each single day for phrasing that has by no means existed earlier than. Most of it isn’t random. It’s the offspring of breaking information, a brand new product title, a coverage that simply modified, a phrase a journalist coined on deadline {that a} thousand individuals then typed right into a search field inside the hour. Information creates language, and language creates search quantity that didn’t exist yesterday.

Press releases had been uniquely positioned to catch that visitors, and I don’t assume most SEOs, then or now, understood why. A press launch is without doubt one of the solely content material codecs written and revealed on the identical day the information occurs. A weblog submit takes a information cycle to catch up. A press launch IS the information cycle. If the phrasing in that launch occurred to nest the precise phrases somebody would later sort into Google, in both the three-word or four-word model, the discharge might rank for a question that didn’t exist when the author sat down to jot down it.

See additionally: An Easy Digital PR Strategy For AI SEO

What Cachón’s Information Provides That My Outdated Trick By no means Had

Right here’s the place the nesting-doll intuition will get an actual improve. Cachón discovered that branded ChatGPT fan-out doesn’t behave randomly both. The primary sub-query in a run tends to be plain, conversational language. From there, the mannequin narrows with a “website:” operator, then begins pulling precise quoted phrases to test whether or not a supply really says what it thinks it says. Throughout her dataset, quote utilization climbed 25-fold between a run’s first search and its final.

That’s the nesting-doll precept operating in reverse. I used to construct outward from a small phrase to an even bigger one so a reader’s question, no matter size it occurred to be, would nonetheless land on my web page. The AI systems Cachón studied are narrowing inward, beginning broad, and drilling all the way down to a literal phrase it may possibly confirm phrase for phrase. Both course, the underlying requirement is identical: your content material has to comprise the precise wording, at multiple size, otherwise you disappear from a part of the funnel.

I’ve already reported that the average AI Mode query in the U.S. now runs triple the length of a traditional search query, based mostly on Google’s personal Might 2026 utilization information. Line that up subsequent to Cachón’s quantity, {that a} single branded immediate followers out into sub-queries averaging seven phrases every. Size isn’t a aspect impact of AI search. It’s the terrain now, and it has been constructing towards this for longer than a lot of the business observed.

See additionally: Data Shows AI Citation Patterns Reveal Strategic SEO Opportunities

My Take

I feel the business spent the 2010s optimizing for the improper finish of the doll. Head phrases acquired all of the technique conferences and all of the finances, whereas long-tail phrasing acquired handled as an afterthought that Search Console would floor if you happen to had been fortunate. That was backward even earlier than generative search existed and it’s extra backward now that the programs doing the looking, not simply the people, are those fanning a single immediate out right into a dozen particular phrasings.

How To Apply This To Your Content material Technique

You don’t want Cachón’s API entry to make use of any of this. You want three habits.

Discover the nested phrase, not simply the seed phrase. No matter three-word core time period you’re concentrating on, write down the 2 or three four-word, and five-word phrases that naturally comprise it, then construct your opening paragraph or H2 across the longer model. Test Google Search Console for queries with excessive impressions and low clicks first. These are normally the longer variants already knocking in your door.

Publish on the pace of the information, not the pace of the content material calendar. The 15% of queries which can be model new are disproportionately tied to one thing that simply occurred. In case your group has a same-day publishing channel, whether or not that’s a press launch, an organization weblog submit, or a rapid-response web page, that’s your greatest shot at proudly owning language earlier than a competitor even is aware of the phrase exists.

Write the literal reply as a standalone, quotable sentence. Cachón’s information exhibits AI programs more and more confirm claims by trying to find an actual quoted phrase from your personal content material. If the sentence that solutions the query can’t be lifted complete and nonetheless make sense, rewrite it till it may possibly.

None of this replaces the basics. It simply means the basics had been pointed on the lengthy tail earlier than most of us had a reputation for it.

Extra Assets:

Featured Picture: Rawpixel.com/Shutterstock