The FAQ trap: writing answers that never get read

Somebody tells a business owner that AI search rewards answering questions. So the owner sits down, thinks up ten questions a customer might ask, writes ten answers, and puts them on a page.

Nothing happens.

There are two reasons for that, and they’re both worth understanding, because a lot of money is currently being spent on this exact mistake.

Reason one: the questions were invented

Here’s the thing about the ten questions somebody thinks up at their desk. They’re not the questions being asked.

They’re the questions the owner imagines being asked, which is a different set entirely, filtered through everything that person already knows about their own business. Experts are famously bad at guessing what beginners want to know.

Real questions can be found. They’re in search data, in the AI systems themselves, in the emails you get, in what people say on the phone before they book. There’s actual research to be done here, and it takes an afternoon, not a lifetime.

What comes back is usually a surprise. Half of what you assumed, nobody asks. Things you’d consider obvious turn out to be the most common question you get.

Reason two, and this is the one people miss: not all real questions are worth answering

Say you do the research and come back with sixty genuine questions. Now what?

Now you weight them. And the weighting depends almost entirely on who is asking.

Is it your customers? These are gold. Somebody in your market, with the problem you solve, wanting to know something before they hire anybody. Answer these first, answer them well.

Is it your professional community? Careful here. Questions your peers ask are interesting, and they’re flattering to answer, and they will not ring your phone. Writing for other practitioners feels like thought leadership and often is. It just isn’t the same job as being found by a customer.

Is it people who will never buy from you? Students, the merely curious, somebody three states away. Real questions, genuinely asked, of no commercial use to you whatsoever.

So the question isn’t just “is anybody asking this.” It’s “who is asking, and are they somebody I want?”

That’s the difference between an FAQ section that reads as busywork and one that quietly answers exactly what a prospect was wondering about ninety seconds before they call.

And now the bigger problem

Let’s say you do all of the above properly. Real questions, weighted correctly, answered well.

It still might do nothing.

Because writing answers is gate three work, and gate three is the last of the three gates an AI runs your business through. If gate one and gate two aren’t handled, nothing you write at gate three gets used.

To recap quickly. Gate one asks whether a machine can read your site at all, cheaply enough to be worth the effort. Gate two asks whether you’re who you claim to be, backed by reviews and corroboration and community. Gate three asks whether you actually answer anything.

They only open in order.

So the owner who spends three months building out beautiful answer content while their site remains unreadable and their reviews sit at 3.8 has produced something nobody will ever quote. The pages exist. They’re probably helpful to the handful of humans who find them. They will not ring the phone.

That’s the real FAQ trap. Not bad answers. Good answers, published behind a closed gate.

The part almost nobody mentions

Here’s something else about your FAQ content, and it points back at gate one again.

Those answers need to be baked into the schema coding on your site.

Not just written on the page where a human can read them. Structured in the markup, in a form that tells a machine: this is a question, and this is the answer to it. That’s not writing work. That’s the invisible technical layer underneath the page, and it lives squarely at gate one.

Which means your gate three content depends on your gate one work twice over. Once for whether the site is readable at all, and again for whether your answers are marked up in a way that gets them recognized as answers.

Schema is a whole subject of its own, and it belongs in a discussion about what a language model actually reads when it reads your site. That one is covered separately. For now, just know that writing the answer is only half of the job.

What to do instead

If you want this to work, do it in this order.

Find out whether your site can be read. Gate one. Nothing else matters until this is true.

Get your reviews and corroboration in order. Gate two. This is slower and it’s worth starting immediately.

Research the questions actually being asked. Not the ones you’d guess.

Weight them by who’s asking. Customers first. Peers later, if at all.

Answer them properly, and get those answers into the schema, not just onto the page.

Do it in that sequence and the content you produce actually gets picked up and credited to you.

Do it in the order most people do it, content first and foundation never, and you’ll have a lovely FAQ page and a quiet phone.

---

If you’re a Chicagoland business and you’d like to know which gate is actually stopping you before you spend another dollar on content, that’s a good fifteen-minute conversation.