The “How Do I” Query: Becoming the Process AI Walks People Through

Before anyone asks AI who to hire, they ask it how the thing is done. “How do I file a mechanic's lien.” “How do I stage a house to sell.” “How do I set up a trust for my kids.” These how-to queries sit earlier in the journey than the who-should-I-hire queries everyone obsesses over — and whoever's process the model narrates becomes the invisible expert standing in the room when the buying question finally gets asked. If you want to be recommended, start by being the one who explained how it works.

I've spent a long time watching how answer engines assemble responses, and the thing that surprised me most is how much weight the "how" questions carry. We think about AI visibility as a popularity contest — who gets named when someone asks for a recommendation. But the recommendation almost never happens cold. It happens after a string of informational turns where the person is quietly learning the shape of the problem. The source that teaches them the shape sets the terms for everything after it.

Why the process answer comes first — and why that matters

Think about how you actually hire someone for something you don't understand. You don't start by asking who's best. You start by trying to figure out what you're even dealing with. How does this work? What are the steps? What should I watch out for? Only once you have a rough map do you start asking who can do it for you.

AI has slotted itself directly into that first phase. People now ask a chatbot the questions they used to be too embarrassed to ask a professional, or didn't know to ask at all. And the answer they get doesn't just inform them — it frames them. If the model walks someone through a five-step process, those five steps become the person's mental model. When they later talk to you, they're measuring you against the framework they already absorbed. If that framework was yours, you start the conversation already sounding right. If it was a competitor's, you spend the first ten minutes undoing someone else's narration.

This is the part that's easy to miss: the process answer shapes the expectations your sales conversation inherits. I've watched businesses lose deals not because they were worse, but because the buyer arrived with a mental model built by a competitor's content, and everything the business said sounded like a deviation from "how it's supposed to go." That's an expensive place to start.

How AI decides whose process to narrate

When someone asks a how-to question, the model isn't ranking ten links. It's trying to produce one coherent, ordered, trustworthy answer. To do that, it leans toward sources that already look like a coherent, ordered, trustworthy answer. That's the whole game, and it's more winnable than it sounds.

From everything I've seen, a few things make a process the one the model reaches for. The first is completeness in sequence — a method that goes from start to finish without obvious gaps, in an order that makes sense. Models are assembling a step-by-step, so a source that already is a clean step-by-step is doing the model's work for it. The second is legible structure: named steps, short declarative sentences, each step doing one thing. A wall of prose where the process is buried in storytelling is hard to extract; a labeled sequence is easy. The third, and the one most people underrate, is evidence that you've actually done this — the small honest asides about where people go wrong, what the exception is, what you'd skip if you were in a hurry. That texture reads as earned, and earned reads as trustworthy.

Notice what's not on that list: being the biggest brand. The how-to query is one of the places where a focused specialist genuinely out-punches a national name, because the national brand usually publishes thin, legally-vetted, say-nothing content, while the specialist publishes the actual method. My read, not a guarantee — but it's a pattern I see constantly. The clearest process wins, and clarity is not a budget line item.

Writing a process answer AI will actually lift

Here's the practical core. If you want to be the source the model paraphrases, you have to write the thing it can paraphrase cleanly. Most businesses never do, because writing down your method feels either too obvious ("everyone knows this") or too valuable to give away. Both instincts are usually wrong.

Start by picking the real questions. Not keyword-tool questions — the actual "how do I" that a confused version of your customer types at 11pm. You already know them; they're the questions you answer on every first call. Write those down verbatim. Those are your targets.

Then write the process the way you'd explain it to a smart friend who's never done it. Lead with a direct, plain-language answer to the question in the first couple of sentences, before any preamble — because that opening is what the model is most likely to keep. Then lay out the steps in order, each one labeled, each one short. Don't bury the method in a origin story. The story can come after; the method comes first.

Attach your vocabulary. If you have a name for your approach, a specific sequence you follow, a way you frame the decision — use it, consistently, and define it plainly. This is how your framework, not just generic advice, becomes the thing the model repeats. When AI starts explaining your category using your words, you've won something that's very hard to take back.

And be honest about the hard parts. Say where people screw it up. Say what the exception is. Say "most guides skip this, but here's what actually matters." Counterintuitively, this makes you more citable, not less — because it makes the content sound like it came from someone who's done the work a hundred times, which is exactly the signal the model is trying to find. Sanitized content that refuses to admit any difficulty reads as marketing, and marketing is the thing models are trying to see past.

The fear that stops most people — and why it's mostly wrong

Almost every time I suggest this, I get the same pushback: won't I just be training my competitors, or teaching people to do it themselves instead of paying me? I take the concern seriously, and I'll give you my honest answer rather than a reassuring one.

On competitors: the steps are almost never the moat. If your entire advantage is that nobody knows the sequence, you have a fragile business regardless of AI. What actually protects you is judgment, execution, relationships, and trust — none of which a published checklist gives away. If you truly have a proprietary mechanism, describe its shape and outcome without handing over the recipe. Everything surrounding it is safe to publish and helps you.

On DIY cannibalization: the people who read a how-to and decide to do it themselves were, in my experience, never your customers. What more often happens is the opposite — someone reads your clear, honest process, realizes how much judgment and risk is actually involved, and concludes they want the person who obviously knows how to do it. A good method is a qualifier. It filters out the tire-kickers and pre-sells the serious ones on your competence. That's not a leak; that's sales enablement that happens to also feed the AI.

What this looks like in practice

Let me make it concrete without pretending I can quote you a precise lift, because I can't and won't invent one. Picture a bookkeeper whose customers constantly ask some version of "how do I get my books ready before I hand them to an accountant." For years that answer lived only in her head and in repeated first calls. She writes it down once — a clean, ordered, honest walkthrough: what to gather, the order to reconcile things in, the two places everyone creates a mess, and the point where it's genuinely worth stopping and calling a pro. She gives the sequence a plain name and uses it consistently.

Now when someone types that exact question into a chatbot, the model has a complete, well-structured, experienced-sounding version of the process to draw on — hers. Whether it cites her by name or simply echoes her structure, the person arrives at their eventual hiring decision already thinking about the problem in her terms. She didn't win by being the biggest bookkeeping brand. She won by being the one who wrote down how it's actually done, in a form a machine could lift. That's the move, and it's available to almost any specialist willing to put their method on the page.

How to tell if it's working

I'll be straight with you: you cannot fully measure this. There's no dashboard that tells you "the model narrated your process to 4,000 people this week." But you're not blind, either. The move is to run the "how do I" prompts in your category across ChatGPT, Gemini, and Perplexity, and read the answers like a critic. Does the sequence sound like yours? Is your vocabulary showing up? Are you named as a source, or at least echoed in the structure? Re-run them over weeks, because these answers drift, and one reading is a snapshot, not a trend.

When you start recognizing your own framework coming back out of the machine — your steps, your words, your way of framing the decision — that's the signal. It means that long before anyone asks AI who to hire, the model is already teaching your future customers how the thing is done, in your voice. That's a quieter kind of visibility than being named in a recommendation, but it's often the one that decides who gets recommended in the first place.

Write down how you do the work. Do it clearly, in order, honestly, in your own words. It's the cheapest authority asset most businesses never build — and in an era where the first conversation your customer has about their problem is with a machine, it might be the most important one.

About the author

Scott Tischler is the Founder & Chairman of AIrecommend.ai and a practitioner-authority on AI search and Answer Engine Optimization. With 20+ years in marketing technology — including American Express, MetLife, and UBS — and executive and professional study at Wharton, Harvard, and Oxford, he helps businesses become the ones AI recommends.