Content Refresh: When to Update, Merge or Delete Old Posts

Every blog older than a year is quietly split into three populations: posts still earning their keep, posts that could earn again with an afternoon’s work, and posts actively dragging the site down. Most owners treat all three identically — publish, forget, add more — which is why “write more content” so often underperforms: the archive is leaking faster than the calendar can pour.

This is the decision guide for the archive: how to inventory what you have in an afternoon, the four-way call — update, merge, delete or leave alone — with the criteria for each, the mechanics of doing all three safely (redirects included), and why refresh discipline now matters twice over, because AI search engines weigh freshness when choosing what to cite.

Why good posts go bad

Content decays on three clocks at once. The facts age: prices change, tools rename themselves, screenshots drift out of resemblance, “in 2023” becomes an archaeology label. The competition moves: the query you once won attracts newer, deeper answers, and search engines notice engagement shifting to them. The signals stale: both classic ranking systems and the newer answer engines demonstrably prefer recently-maintained sources for anything time-sensitive — our breakdown of how AI Overviews choose sources found freshness among the recurring selection signals, which means an unmaintained archive is now invisible in two places at once.

Decay is not uniform, though — and that is the whole opportunity. Evergreen explainers rot slowly; anything with prices, tools or tactics rots fast; and a handful of posts (the ones ranking and converting) barely rot at all. Triage exists to spend your maintenance hours where decay meets value, which requires knowing what you actually have.

The afternoon inventory

Build one spreadsheet, one row per post, four columns of evidence: organic traffic (last 90 days, from analytics), impressions and average position (from Search Console — the leading indicator traffic hides: a post at position 9 with impressions is an update candidate screaming), inbound links (internal count, plus any external links worth respecting), and business relevance — does this topic still serve who you sell to? Sort by traffic, and the three populations appear on their own: a head of winners, a long middle of maybes, and a tail of zeros. The whole exercise fits in an afternoon for a site under a few hundred posts, and it converts the vague guilt of “we should do something about the blog” into a ranked worklist.

The four-way call

For each row in the inventory, one of four verdicts. The criteria, in the order to check them:

Two judgment notes the table cannot carry. First, links outrank traffic in the hierarchy of evidence: a zero-traffic post with genuine external links never gets deleted — it gets updated or merged so the equity flows somewhere useful. Second, relevance is a veto, not a vote: a post pulling traffic for a service you exited or an audience you never sell to is a delete-or-repurpose candidate regardless of its numbers, because irrelevant traffic flatters analytics while teaching both visitors and answer engines the wrong story about what your site is for — the same topical-focus logic behind tight internal linking.

Updating: what counts as substance

A real refresh changes what the page knows, not what it claims. The working checklist: correct every dated fact and price; replace tired examples with current ones; add the sections the query now expects (scan what currently ranks — the gaps are visible); retake screenshots; tighten the opening so the answer arrives first; add the FAQ the topic has accumulated; and re-run the internal links both ways — new relevant posts should point here, and this post should point to relevant pages that did not exist when it was written. Then, and only then, update the visible date — honestly, as “updated,” with the machinery (sitemap lastmod, article schema) agreeing.

What does not count: swapping the year in the title, re-dating an untouched page, or bolting three paragraphs of filler onto the end. Search systems compare versions; readers compare claims to reality; and a page marked fresh that is not fresh spends trust you cannot easily rebuy. The 20-percent rule of thumb keeps refreshes honest — if less than a fifth of the page changed, you have done maintenance, not an update, and the date should stay put.

Merging: one intent, one page

Overlap happens innocently — three years of “email marketing tips,” “email subject line guide” and “how to improve email open rates” — and its cost is structural: when several of your pages half-answer the same query, engines must guess which to show, the guess rotates, and none accumulates the engagement history a single strong page would. The merge sequence:

1. Pick the winner URL — usually the one with the most links and history, not necessarily the best prose. Age is equity.
2. Combine the substance: the winner absorbs every section, example and FAQ the others did better. The result should be the page you would write today if starting fresh — a real editorial pass, not concatenation.
3. 301-redirect the losers to the winner. Their link equity and any residual rankings follow.
4. Repoint internal links directly at the winner — redirects catch stragglers, but first-class links should not lean on them.
5. Update the winner’s date honestly — a merge is almost always a substantive update by any measure.

Deleting: the pruning cut

For the true dead weight — no traffic, no links, no relevance — deletion is hygiene, not loss. Thin, outdated pages dilute the site’s overall quality signal, and pruning them concentrates what remains. The mechanics: if the URL has any equity or plausible bookmarks, 301 it to the closest living relative — the related category, the nearest sibling post — and only to the homepage when nothing closer exists (blanket homepage redirects are treated as soft 404s anyway). If it truly has nothing, let it return 404/410 honestly; that is what those codes are for. Either way, remove it from the sitemap, strip internal links pointing at it, and keep a one-line log of what went where — the same discipline, in miniature, as a site migration’s redirect map.

The scary version of this advice — “delete half your blog” — earns its headlines from publishers with thousands of zombie pages. Small-site reality is gentler: a typical five-year-old small-business blog prunes ten to twenty percent, merges another ten, and updates the rest on rotation. The point is never volume of deletion; it is that everything remaining deserves its place.

The AI-search dividend

Refresh discipline now pays a second time. Answer engines — AI Overviews, ChatGPT with browsing, Perplexity — select sources per answer, and they demonstrably favour pages that are current, specific and internally consistent. A refreshed page states current facts with a current date and matching schema; a stale one contradicts newer sources and loses the citation. Keep the machinery aligned as you refresh: the llms.txt inventory should list what you actually maintain, the FAQ schema should mirror the questions the page now answers, and the “updated” date must mean it. For a small site, being the reliably-maintained source in a niche is an achievable moat — most competitors’ archives are rotting unattended.

The quarterly rhythm

Triage works as a rhythm, not a rescue. The sustainable shape for a small business: quarterly, one afternoon — re-pull the inventory numbers, action the two or three most urgent updates, execute any merge the data has been suggesting, prune the true zeros. Annually, one deeper pass — re-examine business relevance across the whole archive, because services shift and last year’s “leave alone” can be this year’s “delete.” Between passes, keep a running “decay notes” list: every time you notice a price change, a renamed tool or a dead link mentioned in a post, log it — the quarterly afternoon starts with that list instead of a blank page.

Sequencing within the afternoon follows leverage: position-4-to-15 posts with impressions first (small nudges, visible gains), then merges (structural wins), then facts-rot on still-trafficked pages (trust protection), then pruning (hygiene). New content still matters — the calendar and the archive are complementary — but when time is scarce, the refresh usually wins on effort-to-impact, because a refreshed page starts with age, links and history that a new URL spends months accumulating.

A worked example

A services firm’s five-year-old blog: 78 posts. The inventory shows 9 winners left untouched, 22 posts with impressions at positions 5–14 queued for updates over three quarters, three clusters of overlapping posts (11 pages) merged into three strong guides with 301s, and 14 zero-value posts pruned — nine redirected to relatives, five returned honest 404s. Net result: 61 pages where 78 stood, every one either performing or scheduled for its refresh — and the three merged guides, carrying the combined history of eleven URLs, become the archive’s strongest pages within a season. Nothing in that story required new content; it required deciding.

Measuring whether it worked

Refreshes are measurable if you note the date of each action: compare clicks, impressions and average position for the 90 days before versus after (Search Console makes this a two-filter exercise). Expect updates to move middling positions within weeks, merges to consolidate over a season, and pruning to show up — if at all — as a slow site-wide firming rather than a headline. Log every action in the same spreadsheet as the inventory; after a year you will know, with your own numbers, which interventions pay best in your niche — knowledge no generic guide can hand you.

Edge cases the framework meets

The seasonal post. “Holiday marketing checklist” should not be deleted in January — it should be refreshed each year on the same URL, accumulating equity annually. Seasonal content on stable URLs is the compounding version of the genre; new-URL-every-year is the wasteful one.

The news post. Announcements and event recaps age out honestly — they were never meant to rank forever. Leave them as archive (they cost little), or prune the truly thin ones. Do not waste refresh hours on content whose job ended.

The wrong-audience winner. A post pulling real traffic that never converts — hobbyists on a B2B site, DIYers on a done-for-you service — is a judgment call: keep it if it earns links that lift the domain, add a bridge (“want this done for you?”) to test whether any of the traffic converts with a better path, and only then decide whether the maintenance is worth it.

The competitor-copied topic. If a post only ever existed because a competitor had one — no traffic, no fit — the framework already answered: relevance veto, delete. Content strategy by imitation fills archives with exactly the pages this article exists to remove.

Repurpose before you prune

One checkpoint before any deletion: is there value in the page that belongs somewhere else? A dead post’s best paragraph often deserves a second life as a section in a surviving guide, an FAQ answer on a service page, or a week of social posts — the research was paid for once and can be spent twice. The practical habit: as you queue deletions during triage, skim each for its one salvageable asset (a worked example, a table, a hard-won explanation) and move it before the page goes. This is minutes per page, it softens the psychological resistance to pruning (“but I worked on that” becomes “the good part lives on”), and it means deletion only ever removes the packaging, never the knowledge. What has no salvageable asset was confirming the delete verdict all along.

The bottom line

An archive is a portfolio, and portfolios get rebalanced: winners left to run, fixables refreshed while their equity still warms, overlaps consolidated into single strong pages, dead weight pruned with redirects where equity exists. One afternoon a quarter keeps the whole estate earning — in classic search and in the answer engines now choosing their sources — and it is routinely the highest-yield content work a small business can buy, because it spends effort where history has already done the hard part.