Which of two competing pages wins
When two of your pages target one search, merging, redirecting, rewriting and leaving alone are all defensible. Picking wrongly costs traffic you already have.
WPCWP20 — until 31 October 2026
A post that already ranks somewhere has something the next new post does not: evidence that search engines consider it relevant. Improving it starts from that position. A new post starts from nothing.
The catch is that updating published work can lose ground as easily as gain it, so the process below is built around being able to reverse any step.
Pull your pages from Search Console and sort by impressions where average position sits roughly between 5 and 20. Those are pages search engines already show and users mostly do not click: the gap between "shown" and "chosen" is where an update pays.
Ignore age as a signal on its own. An article from 2019 that answers a question that has not changed does not need rewriting; an article from last year about something that has changed does.
Before generating anything, name the problem for each page. The usual ones: the intro answers a different question than the title promises; facts or figures have gone stale; the page is thinner than what currently ranks; the metadata does not match the content; there is no internal link pointing at it.
Each of those has a different fix. "Run it through AI" is not a fix; it is how a page that had a clear problem becomes a page with a vaguer one.
Metadata and schema first. Cheapest to change, easiest to reverse, and often the whole problem. Then structure — headings, an FAQ block, the opening paragraph. Then the body, section by section rather than a wholesale rewrite. Keep the URL. Changing it turns an update into a migration, with redirects to maintain and equity to lose.
Doing these separately means that when something moves, you know which change moved it.
WordPress revisions are the obvious rollback, and they are fine until a site has WP_POST_REVISIONS set to false or capped low — which plenty of managed hosts do for database size. A restore that is only sometimes available is not a restore.
Doing this by hand, copy the post content somewhere before you touch it. In WP Content Writer Pro the previous state of each post is recorded against the run before any job edits it, and if that recording fails the edit does not happen. Restoring writes back through WordPress, so the restore itself creates a revision — undoing a change never destroys it.
Give it time before judging. Compare the same page against itself — impressions, average position and clicks for that URL — rather than site-wide totals, which move for reasons that have nothing to do with your update.
If it went the wrong way, put it back and try a different layer.
When two of your pages target one search, merging, redirecting, rewriting and leaving alone are all defensible. Picking wrongly costs traffic you already have.
A hidden page still getting clicks means something is wrong, but which thing depends on why it was hidden.
Repointing the link and removing the sentence are different edits with different meanings.
Rewriting until a check passes does not make a claim true.
Rank by opportunity rather than by age. A post on page two for a term with real volume is worth more attention than an older post nobody searches for. Search Console's position and impression data is the fastest way to see that.
It can, which is why the order matters: change one thing at a time, keep the URL, and keep a copy of what was there before. Rewriting a page that is already working is the most common way to lose traffic you had.
Only if the content genuinely changed. Changing the date on an untouched post to look fresh is the kind of thing that works until it does not.
Put it back. If you are doing this by hand, that means a copy of the original before you start. In WP Content Writer Pro the previous state of each post is recorded before the run touches it, and you can restore one post, a selection, or everything a run changed.
WP Content Writer Pro runs this process as a proposal you approve, with the previous version of every post recorded first.