Sitemap index

Fifty thousand URLs per sitemap file, which is a hard cap in the protocol.

Fifty megabytes uncompressed per file, which is usually reached later than the URL count.

A sitemap index is itself subject to the same caps, so very large sites nest one level of indexes.

Compression is permitted and does not change the uncompressed size limit, which is the one that applies.

Most sites are nowhere near either, which is why the decision is rarely about capacity.

Search console tooling reports discovery and indexing per submitted sitemap.

Splitting by section therefore tells you that one section is indexed and another is not, which is a signal a single file cannot produce.

A single file reports one number for everything, which hides exactly the comparison worth having.

Against that: opening a single sitemap shows a reader the site's URLs, while opening an index shows them a list of other files.

Both are valid. The first optimises for diagnosis, the second for anyone who opens the URL directly.

Canonical, indexable URLs that return 200. Nothing else.

Not redirects, not pages carrying noindex, not URLs blocked from crawling — each of those is a contradiction between what the site advertises and what it serves.

Not parameterised variants whose canonical points elsewhere, which advertises a URL you have already said is not the one that counts.

Generating it from the same source that decides what is published is what keeps the two from drifting, and hand-maintained sitemaps drift immediately.

It should reflect a real content change, not a deployment or a file touch.

Search documentation is explicit that an inconsistent lastmod causes the field to be ignored entirely.

So a site stamping every URL with the deploy date gains nothing and loses the field's value for the pages where it was accurate.

Omitting it where no real date exists is worth strictly more than inventing one.

The sitemap advertised in the robots file and served at the conventional path is retrieved and parsed.

Whether the URLs it lists are indexable and canonical is checked against what the crawl found.

A URL in the sitemap carrying noindex is reported as the contradiction it is.

Whether the site uses a single file or an index is a structural observation rather than a finding, since both are correct.