Make the sitemap follow the actual publishing state.
A manually maintained URL list can fall behind changing content. ImbaHub can implement sitemap generation from an agreed publishing source and test both additions and removals. The scope concerns selection and technical delivery; search engine indexing is monitored separately.
- Publishing source
Intended status, URL and language are defined.
- Generator
Selection is written into a checked artifact.
- Delivered XML
The active environment serves the approved selection.
Correct source data does not prove the correct sitemap is already being served.
The CMS page is unpublished but the CDN sitemap still lists it. Is the task complete?
No. Trace build and delivery versions and inspect the actual list after an agreed correction.
Identify the source of publishing decisions
Establish where status, intended URL and indexability are determined. A CMS record is not automatically a published route. Connect generation to the right source and exclude private, unpublished and other records that should not be advertised in a sitemap.
A fictional course library might have a draft in one language and a published version in another. Selection should follow each version's approved state. Inventing a translated address to complete a group does not create a real equivalent.
Generate consistent URLs and valid XML
Decide host, protocol, path and canonical before connecting the generator. The sitemap should list intended addresses without accidentally advertising staging, redirected variants or filtered utility routes. Internal links and canonical need to reflect the same publishing decision.
XML needs correct encoding and escaping. If modification dates are included, derive them from defensible content information rather than every build timestamp. Split files according to actual size and maintenance needs, rather than an attempt to manufacture search signals.
| Part | Decision | Test |
|---|---|---|
| Source | Which publishing state is authoritative? | Draft excluded and approved page included |
| Address | Which URL is intended? | Sitemap and canonical agree |
| Language | Which equivalents actually exist? | No invented language entries |
| Modification date | Which change can be established? | Rebuild alone does not claim a content update |
| Delivery | Which version reaches visitors? | Actual XML response inspected |
Connect publishing and removal to one lifecycle
Both adding and removing a page should change the delivered list. Define what happens when a publishing job fails: which version remains, how drift is detected and who owns the next action. Deleting a source record does not automatically clear an old CDN file.
Unpublishing also needs a separate decision about the page response, links and any replacement. Removing a sitemap row neither closes access nor reliably removes a known URL from search results. Sensitive material requires actual access controls.
Hand over a generator with lifecycle evidence
An initial scope can cover one content source, defined page types and one delivery environment. Test draft, publish, URL change, unpublish and republish using a safe selection. Each case compares source, build artifact and actual delivered list.
The delivery includes selection rules, code, test evidence and failure ownership. Search tool access and monitoring are scoped separately. We propose a tailored implementation after review, without promising indexing or AI citations for listed pages.
Sources and further reading
Questions, answered
Should every route appear in the sitemap?
No. Selection should follow intended public, indexable addresses rather than everything the technology can serve.
Is a passing generator unit test enough?
No. Inspect the built file and the response actually delivered by the agreed environment too.