Why Content Decays Without a Refresh Schedule — and How to Fix It

The SERP a page was planned against is not the SERP it competes in months later: competitors publish, search features shift, the questions around a query fan out. The page itself ages — prices, regulations, dates, product names — while its facts sit frozen. Decay is the gap between a moving market and a static document.

What makes decay manageable is that every input is observable. A documented process logs the search evidence behind each map decision, a provenance record carries the receipt for every claim, and the internal-link plan is a stored artifact. Decay becomes a measurable diff.

Fifteen percent is not a growth target; it is a floor against entropy. Without a quota, refresh work behaves like any unowned backlog: everything is urgent, nothing is scheduled, and the pages that decay first are the ones nobody opens.

Make the quota trackable, not aspirational: a shared freshness log — a simple spreadsheet works — records which pages were touched, which were skipped, and how the backlog grows when the quota is ignored. What gets logged gets managed; a quota nobody can see is a slogan.

A refresh queue built on vibes re-freshens whatever the loudest stakeholder remembers. A queue built on signals re-freshens what the evidence says — and because a documented process stores its artifacts and re-reads them, the five decay patterns below are conditions a team can detect:

• Snippet loss: the extractive answer within the 30–45 word budget no longer matches what the SERP rewards.
• Competitor coverage added: a rival URL now answers subtopics the page never covered.
• Fact staleness: newer provenance-backed observations have arrived and beat the rendered claims.
• Internal-link orphans: link-graph changes left the page with too few inbound edges.
• Date-sensitive claims: prices, regulations, or time-windowed statements whose validity has quietly expired.

None of these signals requires a rank tracker — they are internal evidence. Snippet and coverage signals compare the page against current SERPs; staleness compares claims against your fact record; orphans against the internal-link map. A page can rank today and still accumulate decay a good process sees.

The dangerous way to refresh a site is page-by-page editing in a CMS: rewrite a paragraph, rename a heading, delete a section that carried an anchor — and the hub-and-spoke graph quietly rots. Treat internal linking as a directed acyclic graph — edges are explicit artifacts with equity direction: hub-to-spoke downflow, earned upflow, sibling ties — and a refresh may not mutate them as a side effect.

Two structural rules carry over unchanged. The anchor ceiling still applies corpus-wide: no more than three documents may share one identical anchor text, and refreshed pages must re-pass it. When a refresh reveals two pages overlapping enough to merge, the merge goes back through the same SERP-overlap simulation that governs new candidates, diff recorded before a URL changes.

Concretely, the cycle narrows the standing workflow to the flagged pages. The facts update first — new provenance-backed observations replace stale ones — then the affected sections are re-phrased around them, with writing assistance under human supervision. Nothing is hand-edited into the published output.

The refreshed documents face the same pre-publish audit as new ones. The audit re-reads structure, snippets, anchors and fact integrity, and re-verdicts every refreshed page against the same checklist; blockers hold publication as they would for a first run. A refreshed page earns publication; it does not inherit it.