The sitemap lists a different number of addresses than the audit found by following links

The size of the site is estimated from several independent sources: how many URLs the sitemap lists, and how many distinct addresses we discovered by following links. The finding is raised when those two measured numbers diverge by more than one and a half times (the larger over the smaller is 1.5 or more). The report gives both.

The 1.5 threshold lives in config/scope-estimate.json under consistency.divergence_ratio and it is deliberately tolerant. A small discrepancy between a sitemap and reality is normal; the finding should fire only where the gap is large.

What the check does not do:

• It does not say which source is right. The sitemap may be inflated with old URLs, or link-following may have hit a limit. Both look the same.
• It does not list the specific URLs that differ, only counts. Specific mismatches are reported by ACC-06 and SEO-07.
• It works from measurements, never from an estimate. Our estimate of the whole site can be higher than the number of addresses we discovered (we add one conservative extrapolation step) - that computed number enters neither this finding nor the cross-check between sources. Until 19 September 2026 it did, and the finding then claimed we had found more addresses than we had actually seen.
• It does not count addresses that lead to a file. Images, video, audio, documents, archives, stylesheets, scripts and fonts we recognise from the extension in the address or from the content type the server reported, and we leave them out of the discovered pages. Otherwise a gallery where clicking a thumbnail opens the photo itself would inflate the count by hundreds of addresses; on one audited site that was two thirds of all the “discovered pages”. The list of extensions lives in config/file-url-extensions.json and is deliberately closed: it holds only formats where the answer is unambiguous, so .php and .html are not among them.
• It does not conclude that your sitemap is stale. The difference may simply be pagination and parameterised addresses we discovered, which do not belong in a sitemap.
• It does not run when there is only one source of the estimate (a site with no sitemap).

The finding is site-level; severity is warning.

For visibility in AI answers there is no documented effect, and this finding is not really about your site in the usual sense. It is a finding about the reliability of the picture we formed of your site.

Why we tell you rather than keep it to ourselves: because it has consequences for the rest of the report, and it is more honest to state them than to hide them.

When the estimates diverge, we do not know for certain how large your site is. The audit walks a sample, and some checks switch themselves off when they cannot claim anything about the site as a whole: SIT-01 (orphaned pages), SEO-12 (unlinked pages) and part of ACC-06 stay quiet when the crawl was not complete. A large spread in the estimate therefore means you may have received fewer findings than a complete pass would have produced.

The second consequence is commercial and concerns money: the scope of the crawl feeds into the price of the audit. When a sitemap lists ten times more pages than link-following found, the estimate is easy to get wrong.

The practical conclusion: this is not a defect of your site that you must fix. It is the information “do not fully trust this number”, and a pointer to where to look when the scope of the report does not match what you know about your site.

First find out which of the two sources is off. Two questions decide it:

Then, accordingly:

• An inflated sitemap → regenerate it from existing, indexable URLs only. The specific URLs we could not find are listed by SEO-07.
• An incomplete sitemap → check what generates it. It is usually a plugin that knows only one content type.
• Content reachable only from the sitemap → add links to it from the site. That is also SIT-01 and SEO-12, and it matters more than the mismatch of numbers.
• A sitemap not generated automatically → that is the root of most of these cases. A hand-made sitemap goes stale before anyone notices.

And one thing you may run into on the way that is worth more than this finding: when link-following finds markedly fewer pages than you actually have, the problem is navigation, not the sitemap. Content with no route from the site itself is harder for people and search engines alike to find; the sitemap merely masks it.