How Often Should You Publish? What the Freshness Data Says

Publishers producing three or more pieces per week earn 2.7× more AI citations than less frequent publishers, according to Whitehat's analysis of AI search data. But 76.4% of those cited pages had been updated within the previous 30 days, and Google's August 2026 spam update hit sites that chased volume without value. The answer to "how often should you publish" is not a number. It is a trade-off between frequency, value per page, and the hours you have to review.

What the freshness data actually shows

Ahrefs' analysis of 17 million citations, reported by Whitehat, found that 76.4% of AI-cited pages had been updated within the previous 30 days. Pages with substantive updates earned 3.8× more citations than timestamp-only refreshes, and adding a visible 'Last Updated' date lifted citation rates by 47%.

An Ahrefs study of ~17 million citations across 7 AI platforms found AI-cited URLs are 25.7% fresher than top-10 organic results on average. ChatGPT shows the strongest preference for fresh content, while Google's AI Overviews cite content 16 days older than organic results on average, the only platform not favoring fresher content.

Other research narrows the window further. Amsive found roughly 50% of AI citations come from content less than 13 weeks old. Seer Interactive found about 50% of Perplexity citations come from current-year content. AirOps tracked that 35.2% of cited pages were updated in the last 3 months and 53.4% within six months.

Recency matters, but the benefit comes from genuine revision, not manufactured freshness.

The risk side of frequency

Google's August 2026 spam update rolled out from August 18 to August 21, taking 2 days and 16 hours, with no new policies announced [1][2]. SE Ranking found 16.71% of top-10 URLs fell beyond position 100 during the update, versus 9.2% in a quiet week, an 82% increase in catastrophic drops [1].

The targets were specific. One site lost over 200,000 queries due to programmatic plus AI-generated content in an ultra-YMYL niche. A thin-affiliate site with 97%+ probability AI-generated text lost over 14,000 queries. Recovery typically takes months; one documented case took five months [1].

Google's spam policies explicitly cover attempts to manipulate generative AI responses, not just traditional rankings, and list using generative AI to produce many pages without adding value as an example of scaled content abuse [3]. Frequency without editorial judgment is now a direct liability.

The trade-off framework

The three variables are frequency, value per page, and review capacity. Raising one without the others creates a predictable failure mode: high frequency with low review capacity produces thin or unchecked pages; high value per page with low frequency misses citation windows; high review capacity with no frequency target wastes momentum.

The Ahrefs study author cautions that most brands will see better results from creating new high-quality content than from extremely frequent updating, and warns that low-quality content updated daily will not gain visibility from freshness alone.

The research threshold for a freshness benefit is at least 20% substantive revision. Google has explicitly identified artificially inflated modification dates as manipulative. Wikipedia remains heavily cited despite years without major rewrites, showing quality can outweigh freshness.

A starting cadence by team size

One-person team: One new post per week, with a standing block every other week to refresh the best-performing existing page with at least 20% substantive revision. This yields 4 new posts and 2 substantive refreshes per month.

Small agency or multi-client freelancer: 2–3 new posts per week across projects, with one dedicated refresh day per month per client. Prioritize pages that already rank but sit below the top positions; they have the most to gain from a recency boost.

Add visible 'Last Updated' dates to refreshed pages. Whitehat found this alone lifted AI citation rates by 47%.

How drafting tools fit into the cadence decision

A drafting tool with approval gates lets you set a steady pace without committing to full autopilot. ContentRails offers per-project pace controls including about 3 per week and about 5 per month [4], with autopilot levels that stop at draft generation unless you explicitly enable publishing [4]. Limits are hard stops: when a project hits its monthly cap, autopilot pauses and emails you, nothing is lost [4], and failed runs do not count against limits on Pro [4]. The tool re-reads your site every two days by default [4], keeping the index fresh for accurate refreshes, and you can send a published post back with notes for a rewrite that preserves what works [4].

This maps to the framework: set the pace your review capacity can sustain, keep the approval gate closed, and let the tool handle steady output without risking a spam-policy violation from unchecked volume. Pro users get 20 posts and 20 publishes per month at $100 [4], with extra posts available one at a time if needed [4].

Start with one quality post per week and a standing refresh block every other week. If your review time is the bottleneck, use a drafting tool that leaves the publish gate closed until you open it.

Frequently asked questions

Does changing the published date count as a content update?

No. Pages with substantive content updates earned 3.8× more AI citations than those with timestamp-only refreshes. Google has explicitly identified artificially inflated modification dates as manipulative, and John Mueller has warned against superficial date changes. The research threshold for any freshness benefit is at least 20% substantive revision.

Is it better to publish new posts or update old ones?

Most brands will likely see better results from creating new high-quality content than from extremely frequent updating, according to the Ahrefs study author. That said, publishers producing three or more pieces per week earn 2.7× more AI citations than less frequent publishers. The ideal mix is steady new publication plus selective refreshes of existing pages that already rank.

Can AI-generated content get penalized by Google?

Yes, if it violates spam policies. Google's scaled content abuse policy targets many pages generated primarily to manipulate rankings, regardless of how the content is created, and lists using generative AI to produce many pages without adding value as an example [3]. Sites that violate these policies may rank lower or not appear at all [3]. The August 2026 spam update specifically hit sites with programmatic plus AI-generated content in sensitive niches.

How does ContentRails handle publishing limits if I go over my plan?

When a project hits a limit, autopilot pauses and emails you, and nothing is lost [4]. You choose the pace per project, with options including about 3 per week and about 5 per month [4]. The Write drafts autopilot level generates posts without publishing them, so you review before anything goes live [4]. Pro users can buy extra posts one at a time if needed [4].

How does ContentRails keep content fresh after it's published?

ContentRails re-reads your site every two days by default, so the index stays current for accurate refreshes [4]. You can send a published post back with a note and ContentRails will rewrite it while keeping what already works [4]. This supports the refresh cadence in the framework without requiring you to track which pages need updates manually.

Sources

1. Google August 2026 Spam Update Complete: 2 Days, 16 Hours (seo-kreativ.de)
2. Google Finishes Rolling Out The August 2026 Spam Update (searchenginejournal.com)
3. Spam Policies for Google Web Search (developers.google.com)
4. contentrails.ai