How AI search decides which pages to cite

When an assistant answers with sources, it has read pages, picked passages and linked back. You cannot control that choice, but you can stop making it hard.

It has to be readable without a browser

Many AI crawlers do not run JavaScript. A blog that renders its posts in the browser after load can look empty to them. Posts that arrive as finished HTML, the way a server-rendered page sends them, are readable by every crawler at once.

Answer the question near the top

A passage is easy to quote when it stands alone. Lead a post with the answer, then the detail. A heading that is a real question, followed by a paragraph that answers it, is exactly the shape an assistant can lift.

Be specific

Numbers, steps, names and examples survive being quoted. Vague claims do not, and there is little reason to cite a page that says what every other page says.

Keep the basics right

• One clear title per page, and a description that matches it.
• Stable addresses; if a post moves, redirect the old one.
• A sitemap that lists what exists, and nothing that does not.
• robots.txt that allows the crawlers you want, by name if you need to.

None of this is new. It is what made pages good search results, applied to a reader that summarises instead of listing ten links.