Content best practices for AI Search

AI Search finds and reads your content in a particular way. A few habits in how content is built and linked make a real difference to whether it turns up in an answer. None of this requires new tools or a redesign, it's mostly about how existing pages are linked and written.

AI Search starts from your sitemap, then follows the links it finds on each page, the same way a person clicking around your site would. A page that isn't in your sitemap and isn't linked to from anywhere else won't be found, even if it's published and live.

This is the single most common reason a page is missing from AI Search. If a page can currently only be reached by typing its exact address, make sure it's in your sitemap, or add a normal link to it from somewhere relevant, a related page, a listing page, a menu.

Some page-to-page navigation doesn't use real links, for example a "load more" button that fetches more items without changing the page address, or a filter that updates results without a link for each one. AI Search can't follow these, so anything only reachable this way won't be discovered.

The fix is usually small: a "view all" link, a handful of extra links to key items, or a plain listing page alongside the interactive version.

Since AI Search uses your sitemap as its starting point, it's worth treating it as something to actively maintain rather than a one-off technical setup: remove pages that no longer exist, make sure new pages are added to it promptly, and keep it a plain, flat list of pages, not one that lists other sitemaps, since that structure isn't supported.

A sitemap that's stale or includes dead links can bring inaccurate or outdated pages into AI Search just as easily as it brings in the right ones.

If a long list of items (news articles, staff, courses) is split across several pages using "next page" buttons rather than links to each page, only the first page will be found unless every page is also included in your sitemap. Adding direct links to a few more pages, or making sure your sitemap covers them, brings the rest in.

Pages that are mostly a list of links, a department landing page, a document index, a hub with little paragraph text of its own, don't give AI Search much to work with. It can end up matching on navigation text rather than anything meaningful.

A short paragraph of genuine, descriptive text on these pages, even two or three sentences, makes a real difference to how well they're matched to a relevant question.

AI Search can read the contents of PDFs and Word documents, not just the web pages linking to them, but this isn't switched on by default. If it isn't enabled on your site, AI Search can still link to a document in an answer, but won't be able to search what's inside it.

If policies, forms, or reference details commonly live in documents on your site, it's worth checking whether document content search is turned on. If it isn't and you'd like it enabled, contact your account manager or raise a support ticket.

AI Search doesn't automatically know how old a page is, it only goes on whatever date is written in the page itself. An old news article or outdated policy page can be picked up just as readily as a current one.

It's worth periodically reviewing older pages that reference something time-sensitive, names, figures, ongoing projects, and updating or unpublishing anything that's no longer accurate.