Your XML sitemap and your /llms.txt are two maps of your site — when they disagree about what matters, you publish a mixed signal about what's canonical. Make them agree.
Two files describe your site to machines: the XML sitemap (every indexable URL, for completeness — an established standard search engines and AI crawlers use) and /llms.txt (a curated shortlist of your best knowledge — an emerging convention, Phase 1.3, not yet confirmed consumed by the major AI systems). The sitemap answers "what exists?"; llms.txt is meant to answer "what's canonical?". Keeping them consistent costs little and removes any contradiction from the signals you control.
The fix is alignment, not duplication. The same retrieval-grade pages should be present in the sitemap, surfaced in llms.txt, and reachable from the homepage within about three clicks. Burying canonical content behind deep navigation undercuts both maps — if a person needs five clicks to reach it, a crawler may never prioritise it.
You can also check the groupings against what you are actually grounded on. Bing Webmaster Tools’ Topics Portfolio beta clusters the grounding queries where your site appears and shows demand and competitive intensity per topic (Bing Webmaster Blog — Topics Portfolio beta in AI Performance (Oct 2026)). Use it to confirm your llms.txt content-type groupings reflect the themes you are being grounded on, not only the ones you intended to own.
One consistent set of priority pages, surfaced the same way across both files and shallow in the navigation:
sitemap.xml and /llms.txt.robots.txt (Phase 1.1) and returns valid XML.sitemap.xml and verify every page on your shortlist appears as a <url><loc> entry with an accurate <lastmod> (the date of the last material change). Don’t rely on <priority> — Google ignores it. Drop dead URLs and redirects — a sitemap full of 404s erodes crawl trust./llms.txt, grouped by content type and annotated (Phase 1.3). If a page is on your shortlist, it belongs in llms.txt too.llms.txt together" part of your publish checklist (Phase 6.4) so they never drift.A single canonical page — the framework explainer — represented consistently across the sitemap and llms.txt so the crawler sees one coherent signal.
<!-- sitemap.xml -->
<url>
<loc>https://aiknowledgesignal.io/framework-guides/</loc>
<lastmod>2026-10-13</lastmod>
</url>
# llms.txt
## Concepts
- [Framework How-To Guides](https://aiknowledgesignal.io/framework-guides/): all 30 subdomains of the 6-phase method.Same site, same pages. The difference is whether the two maps agree — and whether the AI gets a whole map or half of one.
llms.txt but missing from the sitemap; the pricing page tops the sitemap but isn't in llms.txt. The canonical research is buried four clicks deep. The crawler gets conflicting priorities and may skip the best content.
llms.txt. One consistent authority map.
| Mistake | Fix |
|---|---|
Canonical page in llms.txt but absent from the sitemap (or vice versa) | Promote it into both, or remove from both — the two maps must agree. |
Treating llms.txt as a copy of the sitemap | Sitemap = everything; llms.txt = curated best. Different jobs, same priority pages. |
| High-signal page buried 4–5 clicks deep | Link it from a hub page so it's reachable within three clicks (Phase 4.2). |
Sitemap not declared in robots.txt | Add the Sitemap: line so crawlers find it at all (Phase 1.1). |
Using <priority> to signal importance | Google ignores it. Signal importance with internal links and llms.txt placement. |
| Changes take weeks to be picked up by Bing/Copilot | Ping IndexNow on publish, and submit your sitemap in Bing Webmaster Tools as well as Google Search Console. |
| Updating one file and forgetting the other | Add "update both" to the publish checklist so they never drift (Phase 6.4). |
sitemap.xml with a current <lastmod>/llms.txtrobots.txt and returning valid XML