Showing Up in ChatGPT and AI Answers: How Businesses Get Cited by AI Search

What you'll take away

• Answer engine optimization is an extension of SEO, not a replacement for it — the same authority signals feed both.
• AI systems lift self-contained passages, so the answer has to sit at the top of the page, not nine hundred words down.
• Four things appear to drive source selection: a complete passage, topical depth, consistency with credible sources, and entity consistency across the web.
• Traffic down with leads flat usually means the visits you lost were never going to convert. Watch leads per session, not sessions.
• FAQ schema is a dead end. Put the questions and answers in visible page copy instead.
• Realistic timeline: three to six months for a site with existing authority.

Transcript

Eden: This is The Visibility Room. Today's episode: showing up in ChatGPT and AI answers — the new front page of search. Your host is Sam, who leads SEO and answer engine optimization at Mediaforce. Joining him is Claire, who runs a mortgage brokerage in Vancouver, and has been in the business fourteen years. Here they are.

Sam: Thanks, Eden. Claire, thanks for coming in.

Claire: Happy to be here.

Sam: So you said something to me before we started recording that I want to open with. You said your clients are showing up already knowing things.

Claire: Already knowing things, already having opinions. Yeah. It used to be somebody would call and say, "I'm looking at a place in Burnaby, what can I get?" Now they call and say, "I ran the numbers, I think I qualify for about seven hundred and eighty thousand, is that right?"

Sam: And where did that number come from?

Claire: That's the thing. A year ago I would have assumed a bank's calculator. Now, when I ask, more and more it's — "I asked ChatGPT." Or the answer just appeared at the top of Google and they never clicked anything.

Sam: Right.

Claire: And sometimes the answer's decent! Sometimes it's badly out of date on the stress test. But either way the conversation starts three steps in, and I had nothing to do with it.

What is answer engine optimization, and how is it different from SEO?

Sam: That's the shift, in one story. So let's actually define it. What is answer engine optimization, and how is it different from regular SEO?

Claire: Yeah, because people are using those two terms like they're the same thing and I don't think they are.

Sam: They're not the same, but they're not separate either. Answer engine optimization is the work of getting your content used as a source inside an AI generated answer — in ChatGPT, in Google's AI results, in Perplexity, in Gemini. Traditional SEO is about earning a position in a list of links. AEO is about earning a mention inside the answer itself, where there is no list.

Claire: So instead of being result number three—

Sam: You're the sentence. Or you're not there at all.

Claire: That's a bit stark.

Sam: It is, and it's the honest version. A ranked list has ten slots. A synthesized answer usually pulls from three to five sources. The competition for a citation is tighter than the competition for a ranking.

Did AI search kill SEO?

Claire: Okay, so — and I've heard people say this — does that mean SEO is dead? Should I stop paying for it?

Sam: No, and this is the misconception I'd most want to correct. The myth is that AI search killed SEO and you now need some separate, exotic AI strategy. The reality is that AI systems are largely reading the same web, using the same trust signals, and leaning on the same search infrastructure underneath. What earns a ranking is broadly what earns a citation — clear content, on a site that has authority for the topic, that actually answers the question.

Claire: So it's an extension, not a replacement.

Sam: It's an extension. What genuinely changes is where the answer has to sit on the page, and how the answer is written. That part is real work.

How should content be written so AI can quote it?

Claire: Explain that. Because when you say "how it's written," I picture the same blog posts we've always done.

Sam: So here's the mechanic. When one of these systems builds an answer, it isn't reading your page like a person with a coffee. It's pulling passages. It wants a self-contained chunk of text that answers the question without needing the four paragraphs above it for context.

Claire: Mm.

Sam: Most business websites bury the answer. You've got a page called "Mortgage Renewals," and the first three hundred words are, "Renewing your mortgage is an important decision, and at our firm we believe—"

Claire: Oh, I've written that page.

Sam: Everyone's written that page. And then the actual answer — what happens at renewal, what you can change, what it costs to break early — is nine hundred words down, split across three sections.

Claire: And the model doesn't get that far.

Sam: It might, but it has no clean passage to lift. Whereas a page that says, in plain language, right under the heading, "At renewal you can switch lenders without re-qualifying under the stress test in most cases — here's what that means" — that's a liftable block. Forty to sixty words. Complete on its own.

Claire: So the answer goes at the top.

Sam: Answer first, context after. It's the opposite of how most people were taught to write for the web.

How do AI models decide which sources to quote?

Claire: Okay. So how does the model decide whose forty words to use? Because there must be a thousand mortgage sites saying roughly the same thing.

Sam: There are. And nobody outside those companies knows the exact selection logic — I want to be straight about that. But the observable pattern is fairly consistent. Four things seem to matter.

Claire: Go.

Sam: One — is there a passage that directly and completely answers the question. Two — does the site have topical depth, meaning you've covered the whole subject area, not one lonely post. Three — is the information consistent with what other credible sources say, because a claim that contradicts everything else gets treated as an outlier.

Claire: And four?

Sam: Four is entity consistency. Is it clear who you are, where you are, and what you do — and does that story match everywhere it appears? Your website, your Google Business Profile, your industry listings, your regulator listing. If you're a licensed brokerage in British Columbia, the BC Financial Services Authority record and your site should agree on the basics.

Claire: That's not something I'd have connected to AI at all.

Sam: It's the least glamorous item on the list and it's probably the highest leverage one for a local business. These systems are trying to work out whether you're a real, identifiable organization. Contradictions make that harder.

Why is my website traffic down but my leads are the same?

Claire: Can I ask about something that's been bugging me?

Sam: Please.

Claire: Our website traffic is down. Not off a cliff, but down — maybe fifteen, twenty per cent over the last while. And our call volume is basically flat. Which I can't make sense of.

Sam: That's one of the most common patterns right now, and it usually isn't a problem.

Claire: It feels like a problem.

Sam: It feels like one. Here's the arithmetic, roughly. Say a hundred people search something like "how much down payment do I need in BC." Before, maybe forty of them clicked through to a page somewhere to find out. Now the answer is generated on the screen, so maybe twenty-five click. You've lost fifteen visits.

Claire: Right.

Sam: But those fifteen weren't going to call you. They wanted one number. The people who click now are the ones who read the answer, decided they need an actual human, and went looking. Lower volume, higher intent.

Claire: So the traffic that disappeared was the traffic that was never going to convert.

Sam: Largely, yes. Which is why the number to watch stopped being sessions. It's leads per session, and cost per qualified lead. If sessions drop twenty per cent and leads hold, your site just got more efficient.

Claire: Okay, that reframes about six months of me being quietly anxious. [Sam] It's a real shift. The caution is you have to actually check the second number, not assume it.

What do I actually change on my website?

Claire: Fair. So — practically. What do I change on the site?

Sam: Three things, in order.

Claire: Love an order.

Sam: First: write down the ten questions you answer on the phone every single week. Not keywords. Actual questions, in the words your clients use. "What happens if my mortgage renews at a higher rate." "Can I get approved if I'm self-employed in BC." "How long does pre-approval take."

Claire: I could give you those right now without thinking.

Sam: Everybody can, and almost nobody has them on the site in that form. Second: each of those questions becomes a heading on your site, phrased as the question. And directly beneath it, a short, complete, plain-language answer — forty to sixty words — before any elaboration.

Claire: The liftable block.

Sam: The liftable block. Third: keep it current, and date it. Anything with a number in it — rates, thresholds, program limits — goes stale fast, and stale content gets dropped as a source. A page that says "as of this month" and is actually accurate beats a beautiful page from two years ago.

Does FAQ schema help with AI citations?

Claire: What about all the technical stuff — the schema, the markup. My web guy keeps mentioning FAQ markup.

Sam: I'd skip the FAQ markup specifically.

Claire: Really.

Sam: Really. The rich result it used to produce in Google was rolled back for most sites, so you don't get the visual benefit anymore. And there's no good evidence it moves AI citations — the models read the visible text on the page, not that hidden block. So you're paying a developer to maintain code that isn't doing anything.

Claire: So the questions and answers should just be — on the page. Visible.

Sam: On the page, visible, in the copy. Which is where they should have been anyway. Structured data is still worth doing for the things that identify you — organization details, location, articles, reviews. That's the entity work we talked about. It's the FAQ flavour specifically that's a dead end.

How do I know if AI optimization is working?

Claire: Okay. Bigger question. How would I know if any of this is working? Because with Google I can at least see rankings.

Sam: This is where I have to be honest with you: the measurement is genuinely worse than what you're used to, and anyone telling you they have precise AI attribution is overselling it.

Claire: Great. [Sam] But there are four things you can actually look at. The first is referral traffic. When somebody clicks a link inside an AI answer, that visit often shows up in your analytics as a referral from that AI service. So go into GA4, look at your referral sources, and see whether those names are in the list. Two years ago they weren't there at all for most businesses.

Claire: That's checkable this week.

Sam: It is. Second — go ask. Genuinely. Take your ten questions, put them into the two or three AI tools your clients actually use, and see who gets named. That's not sophisticated, but it's the most direct signal there is, and most people never do it.

Claire: And if I'm not named?

Sam: Then you look at who is, and you look at what their page does that yours doesn't. Usually it's the same answer: their answer is nearer the top and written more plainly.

Claire: What's three?

Sam: Third, Search Console. Not for the AI part specifically — that's largely folded into your overall numbers and you can't cleanly separate it. Watch it for query breadth instead. Are you showing up for more distinct questions than you were six months ago? That's the early signal that your topical coverage is widening.

Claire: And four?

Sam: Four is the one that matters most and it's off-platform. Ask new clients how they found you, and write the answer down. Not a dropdown on a form — an actual question, asked by a human. When people start saying "I asked an AI and your name came up," you'll hear it long before any dashboard shows it.

Claire: We ask that already. Badly. It's a checkbox.

Sam: Most are. The free-text version is worth ten dashboards.

How long does it take to start showing up in AI answers?

Claire: Alright — how long? Because you know what I'm going to ask.

Sam: I do. And the answer is the same unsatisfying answer as SEO, because it's the same machinery. You're not going to publish a page on Tuesday and get cited on Wednesday. Realistically you're looking at a few months before content is crawled, trusted, and starts getting pulled into answers. Three to six is the honest window for a site with some existing authority. Longer if you're starting cold.

Claire: And in the meantime I just... wait.

Sam: You watch the leading indicators. Query breadth going up. Referral traffic appearing where there was none. Your name starting to turn up when you test the questions yourself. Those move before anything else does.

Claire: Can I say the thing I actually find reassuring about all this?

Sam: Go ahead.

Claire: It's that the answer isn't a trick. I was braced for you to tell me there's some new format I have to learn, or some tool I have to buy.

Sam: No.

Claire: It's — answer people's questions clearly, put the answer where they can find it, keep it accurate, and be consistently who you say you are. That's just... doing the job well.

Sam: That's the whole thing. And I'd say it's actually good news for a fourteen-year-old brokerage, because you have something a content farm doesn't. You know what people actually ask, in the words they actually use, and you know the answers that are specific to this province. That's the raw material. It's just sitting in your phone calls instead of on your website.

Claire: Which is a very polite way of telling me to go write some pages.

Sam: It is. So if you take one thing from this — this week, write down the ten questions you answer most often. Just the list. Then look at your website and find how many of them are answered on the page, in plain words, in the first hundred words of anything. My guess is one or two.

Claire: That's probably generous.

Sam: Then that's the work. It's not a rebuild. It's ten answers, written the way you'd say them out loud.

Claire: That I can do.

Sam: Claire, thank you. This was great.

Claire: Thanks for having me.

Eden: That was The Visibility Room. Thanks for spending some time with us — we'll see you next episode.