Two things drove that. Fixes — internal links, redirects, localisation, duplicate and empty pages. And crawl-scope cleanup — I excluded subdomains I have no access to and which nothing links to, plus URLs that return 403 to crawlers only. The cleanup didn't improve the site; it made the score measure the part of the site we can actually influence, instead of counting noise from tracking subdomains and third-party endpoints.
share of the original count now cleared. Most of these came from one root cause — links set up incorrectly in menus, posts and pages. Fixed at the source rather than papered over with redirects. The same pattern exists on the other sites.
1,001 broken images remain, almost all in /blogs/onlyfans/. The URLs were never real
— example.com placeholders, invalid Unsplash paths, literal [IMAGE URL] text left in
from automated content creation. Nothing can be restored. Clearing this would take the Health Score
to roughly 95; it's the largest single item left. Removing the images is fast, sourcing real
replacements is considerably slower.
Important, but a very large piece of work. It touches styles, marketing tools and apps at once, so it can't be done in isolation without risk to any of them — and on a codebase this old, any up-front estimate would be fiction.
Is this a priority now?Something substitutes the desktop header menu and I haven't traced what controls it yet. Part of the problem is links written incorrectly; part is custom code that was apparently added to fix these same issues.
I take it on, or hand it to the MB developer?Pairs of pages covering the same ground, splitting their own rankings. It may be intentional on the marketing side, which is why it isn't mine to close.
I'll prepare a list — who reviews it on the marketing side?Most are purely technical — tracking, API, QR — and nothing links to them. A few, affiliates in particular, would benefit from work I currently can't do.
Grant access, or treat them as out of scope?Worth setting expectations: the quick wins are gone. Everything that could be closed in an hour has been, which is what the 45 → 87 jump represents. What remains either needs diagnosis before it can be estimated, or has no end state and runs in the background. The next crawl won't move like this one did.
Some things stay open on purpose and shouldn't be read as unfinished: external broken links (third-party sites changing or going down), third-party app scripts (not ours to change without removing the app), and technical subdomains with nothing linking to them. One exception worth watching — a dead URL that still has inbound links is real lost equity, and those get a redirect rather than an exclusion.