Finding decay with data, not instinct
Search Console's position history and a rank tracker show which pages slid and when. Comparing each page against the current top results shows why: usually the competitors are deeper, more specific, or more current. The audit produces a ranked list — recoverable value first — rather than a plan to touch everything. On most sites, a fifth of the pages hold most of the recoverable rankings.
What a real rewrite adds
Depth on the sub-questions the top results answer and yours doesn't. Specifics — numbers, examples, local texture — where the page was generic. Structure that lets a reader and a crawler find the answer. Updated facts, dates and prices. Readability fixed where sentences ran long. And internal links to and from the pages that matter. Length rises as a consequence of answering more, never as the goal.
Consolidation is often the biggest win
Sites that published for years accumulate near-duplicates: three posts on the same question, a service page and a blog post competing for one term, an old and a new version of the same guide. Each one splits the authority the topic could hold. Merging them into one strong page — with 301 redirects from the retired URLs — is the single most reliable ranking gain in a refresh, and it requires no new writing at all.
Verifying where Google reads it
A refresh that isn't measured is a rewrite. Each refreshed page gets re-indexing requested, then tracked in Search Console for impressions and position over the following weeks. Pages that recover are done. Pages that don't get diagnosed again — sometimes the answer is that the intent moved and the page should redirect to something else entirely.
What I found refreshing my own site
Thin pages at four hundred words tripped a text-ratio check and were deepened to seven hundred with genuinely new sections. A county page with a readability flag went from a Flesch score in the twenties to the fifties by shortening sentences, and lost two hundred words of padding in the process. Six pages carried a schema property I had wrongly added. Thirteen near-duplicate pairs became single pages with redirects. Every one of those findings came from a crawl and a reconciliation, not from rereading the site — which is the point of doing it with data.
