SEO for AI Search: How to Get Quoted, Not Just Ranked

You checked Search Console last week and the line was flat. Not down, which would at least be a fire to fight. Flat. Impressions holding, clicks sliding a few percent a month, the kind of slow leak that never triggers an alarm. Then you asked ChatGPT a question a customer would ask, something like "best med spa near me for microneedling," and it answered in three tidy sentences with two names in it. Neither was yours. It did not show ten blue links. It did not send anyone anywhere. It just answered, and the person who asked closed the tab and picked one of the two.

That is the whole problem in one screen. For fifteen years the game was rank in the top three and collect the click. The click is now optional. A large share of searches already end without one, and the AI Overview or the chatbot reply is the destination, not the doorway. If you have been treating your website like a fishing net that catches clicks, the water level just dropped. The question is no longer only where you rank. It is whether the machine quotes you when it writes the answer.

The click was never the point, the customer was

Here is the reframe that makes this manageable instead of terrifying. You never actually wanted clicks. You wanted the customer. Clicks were the delivery mechanism, and for a long time they were the only one. Now there is a second mechanism: the model reads your page, decides you are a credible source, and puts your name, your number, your specific claim into the answer it hands the buyer. Sometimes with a citation link, sometimes just as stated fact the buyer now believes because a machine said it.

So the new job is to be quotable. And quotable is a narrower target than rankable ever was. Google could rank a hedge. A page that said "we offer a range of solutions tailored to your needs" could still climb on backlinks and brand. A language model reading that same sentence extracts nothing. There is no fact in it. It cannot be quoted because it does not say anything. The model moves to the next source that states a price, a number, a named service, a verifiable claim, and quotes that one instead.

The machines read faster than your customers and they are far less forgiving of filler. A human skims past your throat-clearing paragraph. A model just deletes it.

That is the wry part and also the whole strategy. Everything you were tempted to pad, the mission-statement intro, the "we believe everyone deserves to feel their best" line, the paragraph that restates the heading, is now actively working against you. It dilutes the extractable facts. The filler is not neutral anymore. It is noise the model has to wade through to find the one sentence worth citing, and often it just does not bother.

What actually changed, and what did not

Three things changed, and they are all mechanical, which is good news because mechanical things have checklists.

First, answers get consumed without clicks, so being the quoted source is the new page-one. You are not optimizing for a blue-link position, you are optimizing to be the sentence the model chooses when it composes a reply. That means your competition is not only the other ten results. It is also the model's own confidence that it can answer without citing anyone. You beat that by being more specific than the generic answer the model would otherwise generate from memory.

Second, entity clarity got promoted from nice-to-have to load-bearing. Models build a mental map of who you are, an entity: this Organization, at this address, offering these services, run by this Person with these credentials. If your site makes that map easy to build, you get pulled into answers about your category and your area. If the model cannot tell whether "Skin and Self" is a business, a product, or a phrase, you are invisible to it. Structured data, the JSON-LD schema in your page head, is how you hand the model the map instead of making it guess. Organization schema, LocalBusiness, Person for the owner or practitioner, FAQPage for the common questions, Service for what you actually sell.

Third, welcoming AI crawlers is now a decision you have to make on purpose. Your robots.txt tells bots what they may read. A lot of sites, out of caution or copied config, quietly block the exact crawlers that feed the tools your customers use. If GPTBot, ClaudeBot, PerplexityBot, and Google-Extended cannot read your pages, you cannot be quoted by the systems they power. You opted out of the new channel without meaning to. We made the opposite call on our own site: robots.txt welcomes the AI crawlers by name, and every page ships structured data, because we would rather be readable than protected from being read. When we tell a client to do this, we have already done it to ourselves.

What did not change is the part that should relax you. Expertise still wins. Specificity still wins. Real numbers still win. The model is not looking for keyword density or a clever meta trick. It is looking for a source that states something true and precise that it could not have generated on its own. That is the same thing a good human researcher was always looking for. The bar did not move. It just got enforced by something that reads every page at once and never gets tired.

A dermatology client, 41 percent, and one FAQ block

Skin and Self is a med spa with 757 reviews sitting at 4.9 stars. Before we touched anything for AI search, that trust lived in one place: their Google Business Profile. Ask a chatbot about them and it knew almost nothing, because the reviews were on Google's property and the website itself hedged. The service pages read like a brochure. Lots of adjectives, few facts, no schema.

We did three concrete things. We added LocalBusiness and Service schema to every treatment page, so the model could see "this business, at this address, offers microneedling at this price range with this aftercare." We rewrote the top of each page to lead with a verifiable fact in the first sentence instead of a mission statement, because that first sentence is what gets extracted. And we built an FAQ block on each service page with FAQPage schema, answering the literal questions buyers type: how long does it take, does it hurt, how many sessions, what does it cost. Plain answers, specific numbers, no hedging.

Within the tracking window, AI-referred sessions, visits where the referrer was a chatbot or an AI answer surface, went from a rounding error to 41 percent of new organic-adjacent traffic. The FAQ pages did most of the lifting because a question with a direct, numeric answer is the single most quotable thing you can publish. The model was already going to answer "does microneedling hurt." We just made sure it answered using their words and their name.

If your traffic is flat and you are staring at the same problem, book a call and we will pull your robots.txt and your schema in the first ten minutes. Usually the fastest wins are sitting right there.

The checklist for a business that does not have a growth team

You do not need a platform or a subscription for most of this. You need an afternoon and a willingness to state facts plainly.

• Open your robots.txt and confirm you are not blocking GPTBot, ClaudeBot, PerplexityBot, Google-Extended, or CCBot. If you are, and you want the traffic, remove the blocks.
• Ship Organization or LocalBusiness schema on your home page, with your real name, address, phone, and hours. Add Person schema for the owner or lead practitioner with actual credentials.
• On every service or product page, add Service schema and lead the page with one verifiable fact: a price, a duration, a specific outcome, a named method.
• Build an FAQPage block on your money pages answering the exact questions buyers ask, with numbers in the answers. This is the highest-return single move.
• Delete the filler. The mission-statement intro, the "range of solutions" line, the paragraph that restates the heading. Every sentence should contain a fact or leave.
• State specifics you would stand behind: years in business, number of clients served, review count, response time. "757 reviews at 4.9 stars" is quotable. "Highly rated" is not.
• Keep your local fundamentals honest too, because AI answers still lean hard on Google Business Profile signals. Our local SEO playbook covers that layer.

None of this is a trick. It is the opposite of a trick. It is making your site say true, specific things in a structure a machine can read, then getting out of the way. The businesses that win the next few years of search are the ones that were always telling the truth in numbers and just needed to make it legible to a reader that never clicks and never forgets.

The line in your dashboard is not going to un-flatten on its own. The channel moved. If you want us to make your site quotable before your competitor's gets there first, book a call and we will start with the two files that matter most.