Content refresh: the cheapest traffic you are not buying

Content decay is not a mystery and not a penalty. Four ordinary forces act on every published page: search intent shifts as a market matures, competitors publish something better, the facts age, and the links that once pointed at the page stop being renewed because nobody talks about it any more. Across a typical library, the aggregate effect is a loss of fifteen to thirty per cent of organic traffic a year without maintenance.

The decay is uneven, which is what makes it manageable. Evergreen explainers on stable subjects fade slowly. Anything with a year in the title, a price, a screenshot of an interface, or a reference to a product version decays fast. News-adjacent content decays almost immediately. Knowing which of your pages sit in which bucket is most of the prioritisation work.

The practical implication reframes the whole content budget. A programme that publishes four new pieces a month and never revisits anything is running up an escalator: after two years, roughly a third of the library is producing meaningfully less than it did, and the new pieces are partly replacing losses rather than adding growth. The content programme simulator makes this visible by ageing each publishing cohort on its own curve.

The highest-return refresh candidates are not the pages that lost the most traffic. They are the pages with substantial impressions sitting at an average position between four and fifteen: demand exists, the page is eligible, and it is losing the click to something the reader preferred. Improving one of those moves a page that already has history, links and impressions — far cheaper than earning all three from scratch on a new URL.

Build the list in Search Console rather than in an analytics tool. Filter to the last three months, sort pages by impressions, and look for the ones with a position in that band and a click-through rate below the norm for that position. Cross-reference with a year-on-year comparison to catch the pages that used to do better, and with your own commercial map to drop the ones that would never produce revenue however well they ranked.

Then triage into three buckets, because they need different work. Pages that are close and simply outdated need a genuine update. Pages competing with another page of yours need consolidation, not editing. Pages with impressions on queries the page never intended to answer need re-scoping — the demand it attracted is not the demand it was built for, and that is often the most valuable discovery in the whole exercise.

Updating the publication date and adding a paragraph is not a refresh; it is a cosmetic change that search engines compare against the previous version and correctly ignore. A refresh that moves a page does at least three of the following: it answers questions the page did not answer, it replaces facts that have changed, it removes sections the market no longer cares about, it restructures around how the query is actually phrased now, and it adds evidence the original lacked.

Start from the SERP, not from the document. Read the pages currently outranking yours and note what they cover that you do not, what format the results share — comparison table, step list, calculator, video — and which sub-questions appear in People Also Ask. Most under-performing pages are not badly written; they are answering a slightly different question from the one the market is asking today.

Two mechanical additions repay their cost on almost every refresh. First, internal links: point three relevant existing pages at the refreshed one, and check its own outgoing links still resolve. Second, the extractable structure that decides featured snippets and AI citations — a direct answer in the first hundred words, question-shaped headings, real tables for comparisons. Our guide on being cited by AI covers why that structure now matters as much as the prose.

When two of your pages compete for the same intent, editing either one is wasted effort. The signals are split, the engine alternates between them, and both sit lower than one strong page would. The fix is consolidation: choose the page with the better history and links, fold the useful parts of the other into it, redirect the weaker URL, and update every internal link that pointed at the retired version.

Choosing which page survives is not a matter of which is better written. Pick the one with more referring domains, longer history and a URL that better matches the intent — those are the assets you cannot recreate. Then genuinely merge: a consolidation that redirects a page without absorbing what made it useful throws away the reason it had traffic.

The same logic extends to pruning. Pages with no traffic, no links and no commercial purpose consume crawl budget and dilute a site's apparent focus. Remove them — but check external links first, and redirect rather than 404 where anything points at them. Deleting is a legitimate content strategy, and it is the one nobody puts on a roadmap.

Refreshing works when it is a standing allocation rather than an occasional clean-up. Reserve a quarter to a third of content capacity for it permanently: if the team ships four pieces a month, that is one refresh a month, or a batch of five every quarter. The allocation matters more than the cadence, because what kills refresh programmes is always the same thing — new content is more visible internally, so it wins every scheduling argument.

Measure per page rather than in aggregate. Record impressions, clicks and average position for the four weeks before the change, then compare at four and twelve weeks after. Some refreshes do nothing, and knowing which kinds do nothing on your site is how the programme gets better: on most libraries, updating facts and structure works, while rewriting prose that was already correct does not.

Finally, write the refresh date honestly. A visible "last updated" date is useful to readers and to engines when it reflects a real change, and corrosive when it is bumped automatically on every deployment — the signal stops meaning anything and crawlers learn to ignore it. Honest dating is the cheapest trust signal a content library has, and one of the few E-E-A-T factors entirely within your control.