A prospect opens ChatGPT and types something simple: “What’s the best tool for onboarding remote engineering teams?” You’ve spent two years making sure your product ranks for exactly that phrase. Your homepage says it. Your ads say it. You feel prepared.
But that single sentence is not the search the model actually runs. Before it writes a word back to the buyer, it fans the question out into a spray of smaller ones you never see. What’s the best onboarding tool for a fifteen-person startup? For a company with contractors in four time zones? Which ones integrate with GitHub? Which have the fewest complaints about setup time? What do people on Reddit say versus what the vendors claim? Each of those is a separate retrieval. Each pulls back its own set of sources. And the final answer the buyer reads is stitched together from whichever brands happened to win those hidden rounds.
You optimized for the question. The model was answering twelve.
This is query fan-out, and it is the quiet reason so many teams feel invisible in AI search despite doing everything the old playbook told them to. The playbook assumed one query, one results page, one shot at the top. You picked your keyword, you fought for the ranking, and if you won it you were seen. Fan-out breaks that arithmetic. The visible question is now just the doorway. The decision happens in a back room full of follow-ups, and most brands have never even read the guest list.
Here’s what makes it uncomfortable. The sub-questions a model generates are not the ones you would have chosen. They come from how buyers actually talk about the problem, from the objections real people raise, from the edge cases and the “yes, but” conditions that never make it into your marketing copy. A model asks about the boring parts—migration pain, pricing surprises, whether support answers on weekends—because those are the parts humans argue about online. If your entire presence is built around the flattering headline version of your product, you will win the doorway question and lose every room behind it.
I watched this happen with a team I know that sells project management software. They ranked beautifully for their category term. They were genuinely proud of it. But when we traced how a model actually assembled an answer about tools for their niche, their name appeared once, in the opening line, and then vanished. The follow-ups—does it handle dependencies across teams, how steep is the learning curve, what happens when you outgrow the free tier—were answered entirely by competitors and by third parties who had written honestly about those exact frictions. The company had a beautiful front door and nothing behind it that a machine could pick up and quote.
So what do you do about a set of questions you can’t see?
You stop writing for the question and start writing for the conversation around it. The fan-out isn’t random; it mirrors the way people genuinely interrogate a purchase. Sit with your best customers and your churned ones and write down the actual sentences they used—the doubts, the comparisons, the “we almost didn’t buy because.” Those sentences are, roughly, the sub-queries. When your content answers the awkward follow-up as plainly as it answers the headline, you start appearing in rooms your competitors thought were private.
The second move is to make your material easy to lift in fragments. Fan-out means a model rarely quotes your whole argument. It grabs the one clean sentence that resolves one sub-question and moves on. Long, throat-clearing paragraphs that bury the answer three sentences deep don’t survive that process. State the thing. “Setup takes about a day for a team of ten.” “It integrates natively with GitHub and Linear.” “There’s no per-seat charge below twenty users.” Each of those is a self-contained answer to a hidden question, and each can travel without dragging the rest of your page along.
The third is harder and slower: earn a presence in the places the model checks that you don’t control. When a model runs its consistency check—comparing what you say against what everyone else says—your absence from the forums, the comparison sites, and the honest third-party write-ups reads as a gap. You can’t manufacture those, but you can be worth mentioning in them, and you can stop treating the sites that review your category as noise beneath your brand.
The deeper shift here is one of humility. For twenty years, search rewarded the brand that controlled its message most tightly—one page, one phrase, one polished claim, defended to the top of the rankings. Fan-out rewards the opposite. It rewards the brand that has answered the most questions honestly, in the most places, in language plain enough to be quoted out of context. Control is worth less than coverage now. The buyer’s question was never really the question. It was the first of twelve, and you were only ever in the room for one of them.
Go find the other eleven. They’re where the sale actually lives.