GSC-driven rewrites

Every “AI rewrite” tool works on the one post you’re staring at. The actually hard part of SEO at scale isn’t rewriting a page — it’s knowing which pages to rewrite, and then improving them without destroying the keyword fit that’s already working. Search Console answers the first question precisely: it shows which pages are losing impressions, drifting in position, or pulling weak click-through despite a stable rank. That’s your rewrite queue. Hand an assistant the page body plus the queries it currently ranks for, and the rewrite strengthens the existing signal instead of accidentally killing it. This is the workhorse recipe — the one worth running every month.

What you need

• Specter synced to your blog
• An AI assistant you already use
• A Search Console export (last 90 days) with both Pages and Queries-per-page data

The recipe

1. Pull and drop in the data. Run a Specter pull and save your Pages and Queries-per-page exports alongside the posts.
2. Build the rewrite queue. Run a scoping pass that flags declining or under-performing pages and writes a queue CSV — each row with the page’s top queries and a chosen focus (title+intro, depth+schema, or freshness).
3. Review the queue. Drop any rows you don’t want touched — brand pages, intentionally shrinking topics.
4. Run the rewriter, dry-run, push. Read every diff carefully (these are your traffic pages), revert anything that drifts off-keyword, then push. Set a reminder to check Search Console in 30 days.

The prompt

You are rewriting underperforming pages from the rewrite queue. For each row,
read the post (matched by slug) and note its top 3 queries — any rewrite MUST
keep semantic coverage of all three. Then apply the row's focus:

- "title+intro": rewrite only the title (45–65 chars, leads with the top query,
  makes a specific promise) and the first two paragraphs (answer the top query
  in plain language up front). Touch nothing else.
- "depth+schema": insert one new ## section that directly addresses the top
  query — 200–400 words of specifics, examples, edge cases — and add Article
  schema if missing. Don't delete existing sections.
- "freshness": update dated references to the current year where still true, add
  an "Updated for <year>" line under the H1, and bump the modified date.

Never change the slug or URL. Never drop the top-3 query keywords. Log each
change as slug | focus | what changed | new word count.

Cost and time

Most of the wall-clock is reviewing diffs. Don’t skip it.

Pitfalls

• Keyword drift. The fastest way to torpedo a ranking page is to rewrite it without the target keyword. The top-three-query constraint is non-negotiable.
• Slug changes. Never let the rewriter touch the slug — URL changes need redirects, which are out of scope here.
• Volume. Don’t run more than about 20 pages a week. You want measurable signal before you scale.

Where to go next

Pair it on the same monthly cadence with the headline A/B loop for compounding gains, and feed it the rewrite rows straight out of the thin content auditor.