Content Update Cadence: What Freshness Actually Means to Google

By Matt Griffin, Founder, Formative Digital · Published 2026-07-19

Somewhere on your site right now is a page that was accurate when you published it and is quietly wrong today. A price that moved. A statistic from a study that has since been rerun. A screenshot of an interface that no longer exists. Nobody decided to let it rot; the calendar just never came back around. That slow drift is what this page is about, because "how often should you update content" has a real answer, and it is not the vague "regularly" that most agency blogs settle for. It depends on what the page is, what the query expects, and whether the update is genuine. In our methodology this is Vector 8, Refresh: keeping content current so re-crawled pages carry honest freshness signals instead of cosmetic ones.

What freshness actually is in Google's systems

Freshness is not one signal. Google decides per query whether recency matters, a mechanism that dates back to the query-deserves-freshness approach: searches about news, prices, "best X" comparisons, and anything versioned get a strong recency preference, while searches about stable subjects do not. The publish date on a page is therefore not a ranking lever by itself. It is an input Google weighs only when the query type calls for it.

What has changed by 2026 is how wide the freshness-sensitive category has grown. Queries that once behaved as evergreen, like "how to" processes and tool comparisons, now carry recency expectations because the underlying subjects change faster than they used to. Search Engine Land's guide to byline dates makes the operating rule explicit: Google tries to surface the date a page's substance last changed, and it cross-checks the visible date against structured data and crawl history rather than trusting your byline. You do not declare freshness. Google infers it, and it has several independent ways to catch a declaration that does not match reality.

The two clocks: query decay and content decay

Every page sits under two clocks running at different speeds. The first clock is the query's: how quickly does the searcher's expectation of currency expire? A search for "GST filing deadline" resets annually. A search for "what is schema markup" barely ticks. The second clock is the page's: how quickly do its specific claims stop being true? A guide full of tool screenshots decays in months even if the underlying concept is stable for years.

Cadence planning means reading both clocks, because they disagree constantly. A conceptual explainer aimed at a stable query can still decay fast if you stuffed it with 2024 statistics. A seasonal page aimed at a fast query can survive untouched if it contains no perishable claims. The useful audit question is never "how old is this page?" It is "which sentences on this page have an expiry date, and has any of them passed?" Count those sentences and the page tells you its own refresh interval.

What a real update is: facts, dates, sources

A genuine content update changes at least one of three things: the facts the page asserts, the evidence behind them, or the coverage of the question. In practice that looks like replacing superseded statistics with current ones, re-verifying every external source and swapping dead or outdated citations, correcting recommendations that experience has overturned, refreshing screenshots and examples, and adding the sub-question readers started asking after you first published. When SEOTesting documented its own refresh programme in April 2026, one refreshed article, revised 51 times since first publication, ultimately drove a 524% increase in organic clicks. The revisions were substantive every time; that is why the compounding worked.

A useful internal test: could you write one honest sentence describing what changed? "Updated July 2026: replaced 2024 adoption figures with the current Stanford AI Index data and rewrote the tooling section" is a real update. If the changelog sentence would read "changed the year in the title," you have not updated anything.

Date-stamp fraud: the pattern, named

The inverse of a real update has an official name. Google's Publisher Center guidelines prohibit what they call artificially freshening content: changing the visible date without significant change to the page. The SEO industry knows the same move as date bumping or republishing-in-place, and it remains one of the most commonly sold "optimization" services precisely because it is cheap to perform and briefly looks like work.

John Mueller's guidance, reported by Search Engine Roundtable, is unambiguous: do not adjust the date unless you write something new or change something significantly. The tactic fails on three separate layers. Google compares the newly claimed date against its own crawl snapshots and sees an unchanged document. Readers who click expecting new information find the article they already read, which is a trust cost you pay in behavioural signals. And the 2026 generation of quality models, which evaluate what changed rather than what the byline claims, downweight cosmetic year-swaps specifically. An agency that bumps your dates every month is not maintaining your content. It is spending your credibility to pad a report.

Honest cadences by content type

With the two clocks and the real-update definition in place, cadence becomes a table rather than a debate. These are review intervals, meaning the date you reread the page against reality; you only edit when the reread finds drift.

Two caveats keep the table honest. First, ranking trouble overrides the calendar: a page sliding in Search Console gets pulled forward regardless of its slot. Second, no interval obliges you to edit. A reference page that passes its six-month reread untouched is a healthy page, and forcing token edits onto it to feel productive drifts toward the artificial-freshening pattern you are trying to avoid. Cadence discipline of this kind is one of the quieter reasons content programmes compound; the louder reasons live in our guide to building topical authority.

How staleness affects AI-engine citation

AI answer engines have raised the price of staleness. Content refresh studies published through 2026, including Slate's refresh analysis, converge on a striking window: roughly half of the sources cited by AI assistants were published or meaningfully updated within about the previous 13 weeks, and across large URL samples the pages these engines cite average significantly fresher than the organic results for the same queries. Retrieval systems assembling an answer prefer the document least likely to embarrass them, and recency of verification is one of their cheapest proxies for that.

The mechanism matters for planning. A traditional ranking can coast on accumulated authority for a long time after the content goes stale; an AI citation is re-decided at answer time, every time. That makes the quarterly refresh slot on the calendar below more than hygiene: for the statistics-heavy pages that answer engines quote most, it is the difference between being the cited source and being the archived one. This shift is part of the broader 2026 picture we mapped in our SEO trends for 2026 analysis, and it rewards exactly the honest-update discipline Google already asked for. Google's own AI features documentation points the same direction: reliably current, people-first pages are what its generative surfaces are built to draw on.

Signalling a genuine update: dateModified done right

When you have made a real revision, tell the machines about it in the one place they check first. The Article schema on the page should keep the original datePublished and advance dateModified to the revision date. Never overwrite the publish date; the pair together is what communicates "established page, recently verified," which is the strongest freshness posture a URL can hold. Here is the pattern, valid JSON-LD exactly as written once unescaped:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Your Article Title",
  "datePublished": "2024-03-12",
  "dateModified": "2026-07-19",
  "author": {
    "@type": "Person",
    "name": "Your Named Author"
  }
}
</script>

Match the visible page to the markup: a short "Last updated" line near the byline, and ideally a one-sentence note of what changed. Quality raters are explicitly instructed to weigh whether a page's dating is honest and its maintenance visible, a standard we unpack in our breakdown of the Quality Rater Guidelines. The rater lens is a good self-check here: would a careful human, comparing your date claims to your content, conclude you maintain this page or that you decorate it?

A maintenance calendar a small business can actually run

Most refresh advice fails small teams because it assumes someone's whole job is content. The calendar below assumes the opposite: one owner or marketer, a spreadsheet listing every URL with its type and last-verified date, and a few hours a month.

• Monthly (about 1 hour): Open Search Console, sort pages by click decline over 90 days, and reread the top three decliners against reality. Fix what drifted. Skip pages that are declining for reasons other than staleness; not every drop is a freshness problem.
• Quarterly (about half a day): Work through every statistics-heavy page on the list. Re-verify each cited figure at its source, replace what has been superseded, and advance dateModified only where something actually changed.
• Twice a year (about 2 hours): Read every service page aloud, ideally with the person who delivers the service in the room. Correct prices, offerings, credentials, and service areas the moment they are said out loud wrong.
• Annually (about half a day): Full inventory pass. Decide for each old post: refresh, consolidate into a stronger page, or retire with a redirect. Carrying dead weight costs crawl attention that your maintained pages could be using.

That is roughly 30 hours a year to keep a 50-to-150-page site verifiably current, which is less time than most owners spend producing two new posts that duplicate what they already have. If even that budget is unrealistic, structured maintenance on this exact rhythm is what our content engine service exists to carry, with every revision logged so you can see precisely what changed and when. Results depend on your industry, competition, and existing digital presence; the calendar guarantees currency, not rankings.

When not to touch a page

Refresh discipline includes restraint. Leave a page alone when it is ranking well, its claims still verify, and the query intent has not moved; editing a winner to satisfy a calendar introduces risk with no offsetting gain. Leave the URL and publish date alone even during heavy revisions, because history attaches to both. And resist merging every ageing post into one mega-page during the annual pass: consolidation is for genuine duplicates, not for any two pages that share a keyword. The goal of the whole system is a site where every claim a machine or a human finds is one you would still sign today. Some pages get there by being rewritten every quarter. Others get there by being correct the first time and left in peace.