Search visibility for AI answers: what to publish first

The problem

A page can rank well in a search engine and still be invisible to an assistant that answers questions directly. Search engines crawl, index and rank pages. Assistants and agents fetch a few documents, often at the moment a person asks, and quote or act on what they find. They favour sources they can read quickly and attribute clearly.

That does not call for a separate "AI SEO" strategy. It calls for a site that is easy to fetch, easy to parse and honest about what it offers.

An order of work

1. Make every important page fetchable. Check robots.txt names the crawlers you allow, including AI crawlers you are happy to serve, and that it points to your sitemap. A page blocked by robots, or rendered only by script, is a page an agent may never see.
2. Keep the sitemap current. List every page you want found, with real last-modified dates. Remove pages that redirect or no longer exist.
3. Give each page one clear purpose. One title, one description, one main heading. Assistants quote the sentence that answers the question; write that sentence near the top.
4. Add structured data where it is true. Organisation, article, product and service markup (JSON-LD) tells a machine who publishes the page and what it offers. Only describe what is on the page.
5. Offer a plain-text route. An llms.txt file listing your key resources, and Markdown versions of articles, are cheap to publish. llms.txt is a proposal, not a standard, so treat it as a convenience rather than a ranking factor.
6. Publish machine-readable facts. Prices, terms and status in JSON let agents answer "how much" and "is it up" without guessing.

What this does not do

None of these steps guarantees a citation or a ranking. They remove reasons to be skipped. Visibility still depends on whether the page answers a real question better than the alternatives.

Checking your own site

Start with the files in steps 1, 2 and 5: they take minutes to check and are the most common gaps.