How do websites show up in AI engines like ChatGPT?

A website shows up in an AI engine when the engine can read the page as plain text and finds a direct answer on it. That’s most of it. The rest is making the answer easy to find and easy to quote.

Here’s what the engines actually read. They fetch the page and read the plain HTML that comes back. They also read the structured data, a small block on the page that says in a fixed format what the page is, who runs the business, which town it serves, and which question each section answers. There’s also a file called llms.txt, a short plain-text summary of the site written for machines. It’s a new convention, and it costs nothing to have.

What most of them don’t do is run the page’s JavaScript. If your page is blank until a script builds it, the engine sees a blank page. That’s true of more sites than you’d think, and the owner never knows because it looks fine in a browser.

Then there’s what gets quoted. When someone asks an assistant whether a plumber in Gastonia needs a website, the assistant looks for a page where that question is a heading and the first sentence under it answers it. A page that circles the question for four paragraphs and answers it at the end gets skipped for one that answers it in the first line.

A situation. A bakery has a page titled Our Story with photos and a paragraph about the family. Nowhere does it say in plain words that it makes custom wedding cakes and delivers within thirty miles. An assistant asked about wedding cake delivery near Gastonia has nothing to quote. One heading and one sentence would fix that.

We build every site this way: plain HTML, structured data, an llms.txt, and the answer in the first sentence. We use AI too, and this is how we’d want to be read by it.