Answer Cloud generates discovery resources from published website content. Keep the authoritative content in your Astro project and CMS entries, then use the normal publishing workflow to update the generated layer.
Know what is generated#
| Resource | Purpose |
|---|---|
llms.txt |
A Markdown overview linking to machine-readable content. |
sitemap.xml |
A list of public page URLs for discovery. |
robots.txt |
Crawl preferences and a sitemap reference. |
| Markdown mirrors | Readable text versions of ingested or published pages. |
| Canonical and discovery links | Connect a page to its public HTML and text representation. |
| JSON-LD | Machine-readable context associated with the rendered page. |
A Markdown mirror is another representation of a page. Use the canonical HTML URL when linking a reader to the page or attributing its content. Use the mirror advertised by the page or llms.txt when a tool needs text; paths can differ between static pages and migrated CMS collections.
The llms.txt format is a content-guide proposal. It is not a permission system or a promise that any AI provider will read or cite your content. The current publishing workflow does not require a customer-authored llms.json file.
Update through publishing#
- Update an Astro page or the intended CMS entry.
- Build and deploy source changes, or preview and publish the CMS change.
- Inspect the deployment result and act on skipped pages or path collisions.
- Open the canonical page and its advertised text mirror.
- Inspect
llms.txtandsitemap.xmlon the same public hostname.
Do not add hand-written discovery files, widget scripts, or analytics snippets to a hosted source project to duplicate this layer. CMS publication refreshes the entry, collection index, and related discovery resources. Promotion also materializes generated resources for the current production origin.
Verify consistency#
Choose a page containing a distinctive factual sentence. Confirm that its live HTML, text mirror, and CMS published payload agree. Follow a few discovery-file links and verify they return the intended content rather than a 404 page or unrelated redirect. Confirm that production resources use the production hostname.
If a page is absent, check whether it was published, skipped by an indexing allowance, archived, or shadowed by a static artifact. A successful deployment does not guarantee every page was indexed. Avoid changing a working page's slug merely to repair a discovery discrepancy.
See deployment troubleshooting, CMS rendering troubleshooting, and crawler access for the next checks.