Finding orphan pages that Google has known about for ages

An overview page is rebuilt, three links fall away in the process, and nobody notices. The pages concerned keep answering 200, keep sitting in the sitemap and keep collecting impressions in Search Console — it is just that no path through your own navigation leads to them any more, which is why Google fetches them less often and withdraws the internal link equity that used to carry them. It gets expensive at the relaunch. What appears in no structure appears in no redirect table.

How to go about it in JMX

1. Settings → Crawl: leave “Read sitemaps” switched on — it is by default — and set no URL limit for this run. An orphan page is by definition unlinked; without the sitemap as a second source the crawler does not find it. Then let it run in the Crawl tab until “Crawl finished”.
2. Results → orphan pages: all internal HTML pages without a single incoming hyperlink, with status, indexability, title, depth and the In sitemap column. That column separates the two cases: listed but unlinked — or not reachable through the site at all.
3. Issues: the finding SEO.STRUCTURE.ORPHAN, “orphan page with no incoming internal links”, preset as a warning, with recommendation and occurrences.
4. Search insights → “Reach without foundation”: the cross-analysis holds pages with Search Console impressions against crawl, sitemap and server log. Four kinds of finding, at most 200 rows — shown by Google and never found in the crawl; in the crawl but unlinked; linked but never fetched by a bot in the log analysis; reachable but in no sitemap.
5. Structure → crawl diagram: the force layout pulls hubs together and lets orphans drift to the edge — the picture for the meeting where somebody decides what happens to them.
6. Tasks: the rule carries effort “small” and reach “per page”. Forty orphan pages are forty links, not one template change.

What to watch out for

The two lists are not the same list. The Results tab counts every internal HTML page with zero incoming hyperlinks — including those with noindex, those with 404 and those reachable only via a linked 301. The rule in the Issues tab is narrower: it takes only indexable pages and follows reachability across up to five redirect steps before it judges. The result list is therefore longer than the number of findings — the difference between “unlinked” and “unreachable”, not an error.

The limit is in the headline: JMX only finds what some source names — a link, the sitemap, a canonical, an hreflang, a redirect. A page that nothing links to and that appears in no sitemap is invisible to the crawler too, while Google still knows it via an old backlink. No crawl closes that gap. You close it via the cross-analysis and via the Migration tab, whose inventory collection also knows the site: query and the Wayback Machine.

Two special cases are worth knowing. The rule goes quiet as soon as the crawl was cancelled, failed or hit its URL limit — a finding that argues from an absence needs a complete run; without a license every crawl stops at 25 pages and therefore counts as capped. And where the navigation only comes into being via JavaScript, the crawl takes half the site for orphaned. The Rendering tab shows which links only exist in the browser; they do not flow back into the crawl’s link graph.

As long as a page is reachable only via the sitemap, you are asking Google to take up something your own site recommends to nobody.