You've inherited 340 articles. You know roughly 40 of them are wrong. You have Thursday.

The instinct is to start at the top of the list and improve things. It feels productive, it's the thing your job title implies, and it's almost always the wrong first move.

Improving is the most expensive option you have

Rewriting an article means reading it, working out what's changed in the product since, checking with someone who knows, rewriting, re-screenshotting, and publishing. Call it 2 hours if nothing surprises you.

Retiring it, once you've checked that nobody still needs it, takes minutes.

Both of those are legitimate outcomes. Only one of them is treated as the default, and that's why backlogs like yours never move.

The 4 fates

Every old article has exactly 4 possible futures, and naming them makes the decision much faster than staring at the text.

Keep it as is. It's still true and people still read it. Nothing to do. You'd be surprised how many land here once you check rather than assume.

Fix it. It's read often and it's wrong. This is the only group that earns your Thursday, and it's always smaller than you expect.

Merge it. It answers the same question as another article. Fold what's true into the stronger one, and redirect the other to it.

Retire it. Nobody reads it, nobody searches for it, support never gets asked about it, and no customer still has the feature. A feature that's gone for new customers is often still in use by existing ones, so check that first. Then redirect the URL to the nearest living thing so you keep the search equity, and let it go.

The bit people find hardest

Retiring feels like failure. It isn't. An article nobody reads is not neutral: it still gets indexed, still gets retrieved, and still gets quoted back to a customer by an AI that has no idea it's describing a feature you removed in March.

A smaller help center that is entirely true beats a large one that is mostly true. That's more obvious now than it used to be, because your readership now includes something that cannot tell the difference.

How to actually get through it

Do it one category at a time, not one article at a time. Sort by traffic, because that tells you where being wrong costs the most. And decide the fate before you open the article, not while you're reading it. Once you're inside the text, everything looks fixable.

If you just want to know how bad it is before you decide anything, the free health check takes about 10 minutes.