How pages are found

Every count you see and every page the run touches comes from a Confluence search. The assistant runs one search per app to produce the counts, then the same search again to drive the run.

That has three consequences worth knowing: the counts are pages rather than macros, the scope is exactly what your filter says, and a rejected filter looks identical to a clean site.


#What the search returns

Unit

Pages, not macros. One page with forty diagrams counts once.

Filtered

Your search filter is applied to every count and to the run itself.

Order

Ascending page id, which broadly means oldest content first.

Space scope

The whole site, or one space if you opened the assistant from space settings.


#Wider than it needs to be, on purpose

The search is deliberately broader than the conversion. It errs towards reading a page that turns out to need nothing, rather than missing one that did. That is why the "pages checked" number dwarfs "pages changed", and why a large Unsupported count is usually normal rather than alarming.


#Pages that drop out

A small line under the progress bar reads "N pages skipped, no space metadata returned by search". These are pages the search returned that could not be resolved to a space. They are dropped entirely and never appear in the log at all, so they cannot be reviewed or re-run individually.


#A few things that catch people out

  • There is no page picker. A specific list of pages can only be reached through a search filter that happens to select them.

  • On a very large site it is not established whether Confluence itself caps the reachable result set. Verify a big run's page count against your own expectations.

  • Pages you cannot read are not returned at all, so they never appear as failures. Pages you can read but not edit do appear, and fail on permission.



Everything starts with the search.