Claude Code · Anthropic
Audit and rewrite a whole blog with a quality gate
A developer and blogger used Claude Code with parallel agents to audit and rewrite posts on his own site after search traffic collapsed. He reports quality scores rose and indexing recovered, but is candid that he cannot say what caused the recovery.
What they did
A Claude Code session read the production database in read-only mode and produced per-post reports, then ten parallel research agents scored posts against target queries. Rewrite agents each owned a fixed set of post IDs and had to cite named evidence, and a Python verifier checked headings, internal links against a slug index, banned phrases and minimum scores before a script backed up each row and updated one post at a time. The 84-post job ran in seven batches, with a manifest file so work could resume when sessions hit usage limits.
How it went
The poster says average internal quality scores rose from about 75 to 86 and indexing recovered over the following weeks after indexed pages had fallen from about 7,100 to about 1,700.
Worth knowing
Automated content had caused real damage first: in one news section most cited sources were dead links and all those articles were pulled. Technical fixes overlapped with the rewrites, so the cause of recovery is unclear, and it was his own site rather than a client's.
Try it with Claude Code
Connect read-only to <blog database> and produce a report per post: target query, heading structure, internal links and obvious quality problems. Score each post against <target query list> and propose rewrites only for the bottom <number>. Before anything is saved, run a check that every internal link exists in <slug list> and every cited URL loads, and back up each post to <backup folder>.
We wrote this prompt from the case. Replace the parts in angle brackets with your own details.
What this label means: User reports
Someone using it at work posted about it. Any numbers are theirs, so we write "the poster says".
Real cases of AI agents at work, from every major agent, each linked to where it was first reported. Rewritten in our own words, most with a prompt you can try. These are not Wize customers.