When the customer asks an AIinstead of searching onGoogle

First things first: SEO isn't dead. Whoever says it is, is selling something.

Most people still type into Google, still click on results, still compare three tabs before choosing. Showing up well on Google is still the most profitable work a small website can do, and it still works for the same reasons as always.

What changed is that a second door appeared, and it doesn't work the same way.

Some of your customers already ask an AI directly: “which dentist near me takes insurance in the evening?”, “how much does a website for a small clinic cost?”. And they get a text answer, with two or three recommendations, without a list of ten links to compare.

Through this door, you aren't a result. You are cited — or you aren't.

What changes when the reader is going to summarize

It is worth understanding the mechanics, because they explain every practical recommendation that follows.

When someone asks an AI, it has to put together a short, assertive answer. To do that, it looks for content that is already in answer form.

If your page says “we serve with excellence and commitment, with years of experience in the market,” there is nothing there to repeat. It isn't false — it is empty of information. The AI can't turn that into a recommendation without inventing details, and inventing details is exactly what it avoids when it finds a better alternative.

If your page says “we are open Monday to Saturday, 8 a.m. to 8 p.m., downtown, we take insurance X and Y, and the first appointment costs R$ 180,” there is material. That sentence can be quoted in full, with no risk.

The difference between the two pages isn't writing quality. It is density of facts per paragraph.

A page that answers a question beats a page that talks about the company

The site structure almost every small business has is: Home, About, Services, Contact.

That structure was designed for people who already know you and want to browse. It wasn't designed for someone who asked a question.

The adjustment that pays off most is creating a page per real question. Not “Services,” but “How much it costs,” “How long it takes,” “Do you take insurance,” “What size of business is it for.”

That has a good side effect: those pages work through both doors. On Google, because the question the person types matches the page title. In AI, because the whole page is about one thing only, and it is easy to quote without ambiguity.

And it has an even better side effect: to write those pages, you are forced to decide the answers. Many companies discover at this point that they have no defined price — they have a range negotiated case by case, on the spot, off the top of their head.

State the price, the timeline and who it isn't for

The three pieces of information most often missing are the three that decide the most — and the three that are scariest to publish.

Price. The argument against is “it depends on the case.” It almost always depends, and it is almost always possible to publish a range stating what makes it vary. A range with criteria is information. Silence isn't.

Timeline. Same logic. “Between 15 and 30 days, depending on how many pages” answers the question and already filters out whoever is in too much of a hurry.

Who it isn't for. This is the rarest and the most powerful. A sentence saying “we don't take projects below X” or “we're not the best option if you need it yesterday” does two things at once: it saves pointless meetings and signals that there are criteria.

Whoever hides all three is betting the customer will message on WhatsApp to find out. Some do. Most go to the competitor who already answered.

Where this argument stops holding

Three limits, and the third is the one almost nobody talks about.

If your site doesn't show up on Google, this won't save you. AIs rely largely on what is already indexed. A new site, with no links pointing to it, doesn't become a recommendation because of formatting. The order is still: exist, be findable, then be quotable.

And don't rewrite everything for robots. People are the ones buying. A text written entirely like a spec sheet, without a single sentence that sounds human, answers the machine well and drives the reader away. The goal is to put facts inside text people want to read — not to trade one for the other.

The third limit is the most uncomfortable: this is hard to measure. There is no reliable report today saying “you were cited in 40 AI answers this month.” You don't see volume, you don't see competition, you don't see progress. So any promise of “AI optimization” with a guaranteed number is inventing the number.

What remains is what has always worked: publishing true, specific, dated information. If AIs change their criteria tomorrow, that content still serves — because it was made for people first.

There is no new shortcut here, and that is the best part.

A site that answers questions with specific facts works on Google, works when an AI summarizes it and works when a person reads it before deciding whether to get in touch.

The hard work isn't technical. It is deciding what you charge, how long you take to deliver and who you don't serve — and committing to it in writing.

Whoever has already done that part doesn't need to worry about the next algorithm change.