The standard SEO advice on content refresh is “update old posts to keep them fresh.” It sounds harmless, but it’s produced two failure modes that are everywhere right now.
The first one is constant tinkering. Every post gets touched every quarter. Dates get bumped. A paragraph gets reshuffled. Nothing meaningful changes, but the “Updated” badge says it did.
The second one is benign neglect. Posts get written, posts get published, posts get forgotten unless they break. Two years later, half your top traffic posts are saying things that aren’t true anymore.
Both patterns underperform. Constant tinkering trains Google and AI engines to treat your dateModified signal as noise. Benign neglect lets content drift out of date until the post embarrasses the author who wrote it.
The discipline that actually works sits somewhere between these two extremes, and the right cadence depends on the post type. Let me walk you through the framework we use for our own blog and for client engagements.
Not all content decays at the same rate. The first move in any refresh discipline is figuring out which post is which type.
Evergreen reference content explains how something works, what something means, or what the canonical answer to a recurring question is. These posts age slowly. Written well once, they stay useful for years with minor edits.
Refresh cadence: 12-18 months. Update when underlying facts change (a new version of a tool, a new feature in a platform, a major industry shift). Otherwise leave them alone.
Example: a 2024 post explaining what schema markup is. Most of it is still correct in 2026. A refresh updates a couple of examples to reflect 2026 schema.org changes. The bulk of the post stays.
Time-sensitive list content is “best tools for X in 2026” or “the 10 most-cited brands in AI search this quarter.” These posts include a year or quarter in the title, and the underlying list changes faster than the title implies.
Refresh cadence: 3-6 months. The list needs to be current. If the title says 2026, the data better be 2026, not 2024 with the title swapped.
When you refresh, actually re-run the underlying analysis. Don’t just update the title and add an “Updated for 2026” badge. AI engines and Google both detect superficial refreshes and de-weight them.
Opinion and field notes are posts that capture a specific observation at a specific time. “First seven days of inbound.” “What four fit calls taught us.” Time-anchored content.
Refresh cadence: never. These posts are anchored in time, and the time anchor is the value. Refreshing breaks them. If the observation needs to be updated, publish a new post. Don’t edit the old one.
How-to and tactical content walks through a specific procedure. “How to set up FAQPage schema.” “How to audit a Bing index.”
Refresh cadence: when the procedure changes. Schema syntax updates, Bing dashboard redesigns, API version bumps. The refresh trigger is the procedure changing, not the calendar.
The post ranks in positions 4-10 on Google but doesn’t get clicks. It’s candidate-worthy but losing on title, meta description, or first-screen content. Refresh the title and meta description, rewrite the first 100 words to be answer-first. Don’t refresh the body unless it’s also weak.
The post used to rank, dropped, and the cause is identifiable. A competitor published a better version, the search intent shifted, a Google update changed what gets ranked. The fix is targeted. Improve the part that became uncompetitive.
The post gets cited inconsistently by AI engines. Some queries cite it, similar queries don’t. The cause is usually that the post almost works but has structural gaps. A 100-word answer block at the top, FAQPage schema, named sources in the body. The refresh adds the missing structure rather than rewriting the substance.
The post contains a factual claim that’s now wrong. A tool was renamed, a feature was removed, a price changed, a statistic got updated by the source. Fix it immediately. This isn’t a discretionary refresh. It’s correction maintenance.
The post earned external links and those links are now driving traffic. Lean in. Add a featured CTA, improve the conversion path, link from the post to your money pages with better anchor text. The post is doing the most work it’ll ever do. Treat it accordingly.
You re-read the post and the writing makes you wince. Voice has improved. The post sounds dated or thin. Refresh for voice, even if the substance is fine. This is a quality investment, not a ranking move. Both audiences (humans and AI engines) reward better writing.
Five categories of edit, ranked by impact on both classic SEO and AI citation.
The highest-impact edit is the first 100 words. AI engines extract heavily from openings. Google’s featured snippets pull from there. If the opening is weak, the post under-performs regardless of how good the body is. Refreshing the first 100 words is often the single highest-impact edit you can make.
H2 structure is the next-highest impact. Sub-headings tell engines what the post covers. Refreshing weak H2s into more specific, more answerable ones lifts both ranking and citation candidacy. Convert “Why this matters” into “Why entity binding matters for AI citation.” Convert “Conclusion” into “What to do this week.”
Factual claims and statistics matter for trust as much as for ranking. Outdated stats produce wrong AI citations. Wrong citations train the engines that your source is unreliable. Refresh stats with current numbers, and cite the source explicitly so the engine can verify.
Meta description and title affect Google CTR more than ranking. AI engines mostly ignore them in favor of the page body. Refresh when CTR is weak.
Word count is the lowest-impact lever and the one most agencies mistakenly use. Adding 500 words of filler to “make the post more comprehensive” doesn’t improve ranking or citation. Word count is a side effect of substance, not a lever. Most refreshes that pad word count make the post worse.
Three patterns that look like refresh but actually degrade content.
Bumping dateModified without substantive edit. Some SEO advice still says to update the modified date every few months to signal freshness. AI engines and Google both detect this and de-weight pages that update the modified date without changing the content. The “freshness signal” is meant to indicate real updates. Faking it trains engines to distrust the signal.
Adding “Updated for 2026” badges without updating the substance. Same problem. The reader sees the badge and expects current information. If the substance isn’t current, the gap between badge and substance trains the reader (and the engines tracking reader behavior via bounce rate) that the brand is dishonest.
Refreshing posts that should be republished. If a post is 4 years old, the underlying topic has changed substantially, and a refresh would amount to a rewrite, just republish as a new post and 301 the old one. The republish gets a fresh date, fresh schema, and a clean URL. The refresh forces you to engineer around an old URL that was indexed under different assumptions.
We refresh on a slow cadence. The blog is 30 days old so most posts haven’t aged into refresh windows yet. But the framework we’ll use is:
We also use one rule that doesn’t bend: every refresh substantively changes at least 300 words of the body content. If the proposed refresh doesn’t pass that bar, it isn’t a refresh. It’s a fake-date update and it doesn’t get shipped.
The right move this week isn’t to start refreshing everything. It’s to audit which posts are which type, identify the ones currently in refresh windows, and refresh only those.
For most B2B blogs we audit, the breakdown is roughly:
That means at any given month, about 10-15 percent of the blog is due for a refresh check. Most posts should be left alone.
If you want us to build a content refresh cadence for your blog (categorize the existing posts, identify the ones due for refresh, ship the refreshes that move the needle), the fit call covers it. The first cadence ships in week two.
Related reading:
– Topic clusters in the AI search era (do they still matter?)
– The ‘Sources’ section is the most underrated AI citation move
– Original research as the highest-impact content shape in 2026
– Answer-first writing: before and after for B2B
– The new search: how AI is changing search behavior
