Find and refresh outdated content

SEOIntermediate~30 min

Before you start

A connected site with content reads on and the content write ability enabled. Take a database backup before a large run.

Old posts are an asset you already paid for. A guide from three years ago might still pull steady traffic, but it name-drops a tool that shut down, quotes a stat from 2022, and says “this year” about a year that is long gone. A light refresh often lifts a page that is already ranking further than a brand new post would.

This gives the agent the archive dig, finding the oldest posts and reading them for stale signals, and keeps the editorial calls with you. It does not rewrite the post. It targets the specific bits that went stale and proposes a precise edit for each, so your voice and structure survive intact.

The prompt
I want to refresh my outdated content. I have content writes turned on.
First pass, change nothing. Count my posts so I know the size of the archive,
then list the oldest ones by date and read the [twenty] oldest. For each, flag
the stale bits: outdated years, old statistics, references to tools or products
that no longer exist, "current" claims that have aged. Propose a specific
find-and-replace edit for each stale bit, and leave anything still accurate
alone.
I will tell you which edits to keep. Only after I approve, apply them with a
literal find-and-replace, work in small batches, and show me each batch before
it saves. Do not rewrite whole posts and do not touch post status. At the end,
list every post you refreshed and what changed on each.

Why this is safe to run

The whole discovery pass is reads. Counting the posts, listing the oldest by date, and reading them writes nothing, so every stale bit and its proposed edit reach you as a list before a single save.

The write half is a literal find-and-replace on the exact stale text, which edits the post body only and never touches status, so nothing gets republished or unpublished behind your back. The content write ability stays off until you enable it, every call is checked against the capability of the user you bound the agent to, and every edit lands in the Activity Log with the post it changed. Because the edits go through revisions, any change you want to undo is one restore away, and anything still accurate is simply left as it was.

Already connected? Skip the paste. Ask your agent for the “find and refresh outdated content” recipe, or just point it at your archive, and it runs the same flag-then-approve flow on your own site.

How it works

  1. Count the posts so you know the size of the archive you are working through.
  2. List the oldest posts by date and read the ones you want to review.
  3. Flag the stale bits per post: outdated years, old stats, discontinued references, and propose a specific edit for each, then wait for your approval.
  4. Apply the approved edits with a literal find-and-replace, in small batches, pausing to show each batch.
  5. Return the list of every post it refreshed and what changed on each.

Questions

Will it rewrite my whole posts?
No. It targets the stale bits, an old year, a dead stat, a reference to something discontinued, and edits just those with a literal find-and-replace. The rest of your post stays exactly as written, and every edit is reversible through revisions.
How does it decide what counts as outdated?
It surfaces the oldest posts by date and reads them, then flags concrete signals of staleness and proposes a specific edit for each. You see every proposed change before anything saves, so a fact that is still correct simply gets left alone.