Track: Fundamentals · Format: text lesson · Est. read: 12 minutes
By the end of this lesson you'll be able to explain, in your own words, why a brand shows up (or doesn't) when someone asks ChatGPT, Perplexity, or Google's AI Overview a question, and what that has to do with the website you already own.
No jargon is assumed. The few terms that matter link out to the Knowledgebase so you can go deeper later.
Start with what changed
For twenty years, getting found online meant one thing: rank on Google. Someone typed a query, Google returned ten blue links, and your job was to be one of them, ideally near the top. The click was the prize. You won it by ranking well.
That path still exists. A second one now runs beside it.
Ask ChatGPT "what's a good tool for tracking B2B pipeline" and you don't get ten links. You get a written answer, in sentences, naming a few products. Ask Perplexity the same thing and you get an answer with little numbered citations next to it. Run a Google search for it and, above the old blue links, you'll often see an AI Overview: a paragraph Google wrote for you, pulled from a handful of pages, that tries to settle the question before you scroll.
These are answer engines. They don't hand you a list and send you off to read. They read for you and hand back a conclusion. That single shift is what this whole field is about.
Here's why it matters to you. When the engine writes the answer, the only brands the reader sees are the ones the engine chose to mention. Ranking eleventh used to mean "page two, but still in the game." In an answer engine there is no page two. You're in the answer or you're invisible.
How an engine decides what to say
It helps to picture what happens in the second after someone hits enter. It's less magic than it looks.
The model behind ChatGPT was trained on a huge slice of the internet, but that training is a snapshot, frozen months before you asked your question. It knows nothing about your new pricing, your product launch last week, or the comparison page you published yesterday. If it answered from memory alone, it would be confidently out of date.
So most answer engines don't rely on memory alone. When your question comes in, the system first goes and fetches documents that look relevant to it, right then, from the live web or a search index. Then it hands those fresh passages to the model and says, in effect, "answer using these." The technique has a name: retrieval-augmented generation, or RAG. Those citations stacked under a Perplexity answer are literally the documents the retrieval step pulled.
Two things have to go right for your page to end up in that answer.
First, the engine has to be able to find and read your page when it goes looking. Not rank it on Google. Retrieve it, in the moment, and parse what it says. A page hidden behind a login, locked inside a PDF, or built so that the important text only appears after a script runs might as well not exist to the retrieval step.
Second, once your page is in the candidate pile, the engine has to choose it over the others. This is where trust comes in. Given several pages that all seem to answer the question, the model leans toward the one that looks most credible: a clear author, real data, a site with a track record on the topic. That quality has a name too, source authority, and it's often what breaks the tie between two pages that say roughly the same thing.
Findable, readable, credible. Miss any one and you're not in the answer.
Why most sites are invisible to it
Plenty of sites that rank fine on Google never show up in AI answers. The reasons are usually dull and fixable.
The answer is buried. A page might contain exactly the right information, three paragraphs down, after a founder's origin story and a mission statement. A human skims past that. A retrieval system, trying to lift a clean passage that answers "how much does this cost," often can't isolate it and reaches for a competitor who put the number in the first line.
The content is thin or vague. "Contact us for pricing" answers nothing. If the concrete fact the buyer asked for isn't on the page in plain words, the engine has nothing to quote, so it quotes someone else.
The page can't be read by a machine. Key details trapped in a PDF, an image, or a form the crawler can't submit are invisible to retrieval. So is anything gated behind "enter your email to see the details." If a person needs to fill out a form to reach the fact, the engine never reaches it at all.
Nothing tells the engine what the page is. Adding structured data, a bit of behind-the-scenes markup that labels a price as a price and an FAQ as an FAQ, makes it far easier for an engine to pull the right passage cleanly instead of guessing.
None of these are exotic. They're the same content problems that always cost you, showing up in a new place with higher stakes. The difference now is that a buried answer doesn't just rank lower. It gets skipped entirely while the engine quotes the page that led with the answer.
What this does to the buyer's journey
Step back from the mechanics and look at the buyer.
The old top of the funnel was a search box and a page of links. A buyer with a vague problem would search, click three or four results, open a dozen tabs, and slowly build a mental shortlist by reading. You could earn a place on that shortlist by ranking and writing something worth the click.
More of that first stretch now happens inside a conversation with an AI. The buyer describes the problem in their own words, asks follow-ups, and gets a shortlist handed to them, already narrowed, often before they visit a single vendor site. Many of those questions get answered without a click leaving the engine at all, a pattern called zero-click search.
Sit with what that means. A chunk of the research that used to happen on your website now happens somewhere you don't control, and you may never see the visit in your analytics. By the time the buyer does land on your site, the engine has already decided whether to put you on the list. If your brand wasn't in the answer, you were cut before the buyer knew your name.
This is why "AI visibility" is worth a course of its own. It's the top-of-funnel question for a world where the funnel starts inside a model's answer.
What being AI-visible actually looks like
Make it concrete. Imagine two companies that sell the same B2B analytics product, and a buyer who asks an engine: "how much does an analytics tool for a 20-person sales team cost, and how does it compare to the alternatives?"
Company A has a pricing page that opens with a founder's letter about their journey. The actual numbers sit in a comparison table further down, rendered by a script, with no labels an engine can parse. Their competitor comparison lives in a gated PDF you download after a demo call. On Google they rank well for "sales analytics pricing." None of that helps here. The retrieval step can't cleanly lift a price or a comparison, so the engine builds its answer from pages it can read.
Company B has a pricing page that opens with a plain sentence: "Plans start at $40 per seat per month for teams up to 25, with annual billing at $32." Right below it, a short table labeled as pricing data. They also publish an ungated "[Company B] vs. the alternatives" page that names competitors and states honest tradeoffs. When the engine assembles its answer, Company B's number and comparison are sitting there, easy to quote and easy to trust.
The engine writes its answer. It quotes Company B's pricing, mentions their comparison, and sends the buyer a citation. Company A isn't in the answer at all, despite ranking higher on Google.
Now notice the second half of the win. The buyer clicks Company B's citation. If that click lands on a generic homepage that says nothing about pricing, the momentum is gone and the buyer starts the hunt over. If it lands on the exact page that answered their question, with an obvious next step, the visit turns into a conversation. Earning the citation and capturing the warm visitor it sends are two separate jobs, and you have to win both. This course is about the first. Keep the second in mind, because a citation you can't convert is a compliment, not a customer.
Company B didn't outspend anyone. They wrote the answer where a machine could find it, read it, and trust it. That's the whole game, and it's mostly work you can do on the site you already have.
The one idea to take away
A brand shows up in an AI answer when the engine can find its page, read the relevant fact off it, and trust it enough to quote over the alternatives. A brand stays invisible when any of those three breaks, and most of the breaks are ordinary content problems: the answer buried too deep, locked in a PDF, gated behind a form, or never stated plainly at all.
Your website is still the asset. The audience reading it just changed. Part of it is now a model deciding what to tell a buyer who never typed your name.
What's next
You now have the mental model. The next courses turn it into a practice:
- AEO Fundamentals goes deep on the first job from this lesson, structuring your content so engines quote it. It builds on answer engine optimization.
- GEO vs. SEO is for marketers who already know search and want the precise delta, framed around generative engine optimization.
Together with this lesson, they build toward the AI Visibility Analyst certification, which tests whether you can read an AI visibility report and act on it.
Before you move on, take the quiz in quiz.md. It's short, and it's the fastest way to find out whether the model actually stuck.