Google's AI Overviews now appear in 82% of B2B technology searches, up from 36% a year earlier. That single stat should have changed every B2B content team's publishing calendar. For most teams, it hasn't.
The standard playbook still treats Google rankings as the only scoreboard. Publish a blog post, track its keyword positions, celebrate when it climbs, accept gravity when it falls. But 2026 split organic discovery into two distinct surfaces, each with its own freshness requirements and its own decay curve. Content teams running a static publishing cadence are optimizing for one surface while letting the other erode underneath them, and the cost gap between the two approaches gets wider every quarter.
We've modeled what this looks like in dollar terms for a two-person team at 6, 12, and 18 months. The numbers tell a clear story.
The Overlap Between Rankings and AI Citations Collapsed
In July 2025, 76% of AI Overview citations came from pages ranking in the top 10. Ranking well meant getting cited. These were, for practical purposes, the same game.
By March 2026, that overlap had fallen to 38%. A 50% relative decline in eight months. The two surfaces decoupled.
This is not an incremental shift. It means a page ranking #3 for a competitive B2B keyword has roughly a coin flip's chance of appearing in the AI Overview for the same query. And a page that isn't ranking at all might still get cited if it matches the freshness and structural signals AI citation pools reward. The 5W Research data makes this hard to argue with.
So what does that mean operationally? It means you're publishing into two separate systems with two separate reward structures. And the maintenance window for each is different.
Two Decay Curves, One Editorial Calendar
Here's where the problem gets expensive. Traditional organic rankings still behave the way content marketers expect: a well-written post on a solid domain climbs over 3 to 6 months, plateaus, then gradually decays. An evergreen B2B post used to hold stable rankings for 18 to 24 months. In 2026, competitive terms shrink that to 6 to 8 months, but the curve is predictable.
AI citations follow a different clock entirely.
New content enters AI citation pools within 3 to 5 days, which is dramatically faster than the months-long crawl-to-rank cycle. But the decay is equally fast. AI citations erode around the 13-week mark as facts, entities, and source freshness signals age out of trust. A post that gets cited in April's AI Overviews can disappear from them by July, even if it's still ranking on page one.
This creates an asymmetry most editorial calendars don't account for. Your organic rankings decay slowly. Your AI citations decay quickly. A static cadence treats both surfaces the same, which means it's always wrong for at least one of them.
The CTR Split in Real Numbers
The traffic economics confirm why this matters. Pages ranking #1 experience an average CTR drop of 34.5% when an AI Overview appears for their target keywords. With AI Overviews present, organic CTR falls to 0.61%, compared to 1.76% without them.
But there's a counterweight. Brands cited within AI Overviews earn 35% more organic clicks and 91% more paid clicks than brands not cited for the same queries. And AI search visits grew 42.8% year over year, reaching 27.4 billion in Q1 2026.
So the math isn't "AI Overviews steal your traffic." It's "AI Overviews redistribute traffic toward cited sources and away from uncited ones." If you're cited, you win more than you lost. If you're not, you lose on both surfaces.
This distinction changes everything about how you should allocate production versus refresh budget.
Modeling the Cost: Static vs. Dynamic Cadence
We built a simple model for a two-person B2B content team spending $4,000 per month on content (a mix of writer time, tools, and freelance support). Here's what the two approaches look like at 6, 12, and 18 months.
The Static Cadence
Publish 8 posts per month. No systematic refresh. Monitor keyword rankings. Accept AI citation decay as a cost of doing business.
At 6 months: 48 posts live. Roughly 30 indexed and ranking for at least one keyword. AI citation count peaks around month 3 and declines. Estimated cost per indexed page: $133. Cost per page with active AI citation: $400+ (because only ~10 of those 48 pages maintain fresh enough signals).
At 12 months: 96 posts live. Rankings compound, but the oldest 30% of content is decaying on both surfaces. No refresh means those early posts generate diminishing returns. Cost per indexed page drops to $100, but cost per AI-cited page climbs to $500+ because the citation pool churns faster than new posts replace lost citations.
At 18 months: 144 posts. The portfolio is large, but hollow. Maybe 40 posts still drive meaningful traffic. AI citations cluster around the most recent 30 posts. Effective cost per productive page: $180. The remaining 100+ posts are digital dead weight.
The Dynamic Cadence
Publish 5 new posts per month. Refresh 3 existing posts per month based on freshness signals (AI citation drops, competitive content updates, stat staleness). Run quarterly portfolio audits.
At 6 months: 30 new posts plus 18 refreshed posts. Fewer total URLs, but higher citation density. Cost per indexed page: $160 (higher than static because fewer new pages). Cost per AI-cited page: $240 (significantly lower because refreshes maintain citation eligibility).
At 12 months: 60 new posts, 36 refreshed. The refreshed posts re-enter AI citation pools, creating a compounding effect. Cost per indexed page: $100. Cost per AI-cited page: $150. This is where the math flips. The dynamic cadence's total productive page count overtakes the static cadence despite fewer total posts.
At 18 months: 90 new posts, 54 refreshed. Portfolio health is high because no post goes more than 90 days without a freshness check. Cost per indexed page: $83. Cost per AI-cited page: $120. The static cadence at this point has a cost per AI-cited page nearly 4x higher.
The crossover point sits right around month 10 for most B2B verticals. Before that, static cadences look cheaper because you're producing more URLs. After that, the maintenance deficit catches up.
Why the Crossover Happens
The static model assumes every new post is equally valuable. It isn't. A quarterly review of competitive pages with a 90-to-120-day refresh cycle prevents the decay that silently kills older content. Without it, your portfolio develops a long tail of zero-value pages that still count against your crawl budget and dilute your topical authority signals.
The dynamic model accepts a truth that's uncomfortable for editorial teams: producing fewer new posts and maintaining existing ones generates more total visibility per dollar spent, once your portfolio crosses roughly 50 indexed pages.
What Triggers a Refresh vs. New Production
This is where most frameworks get hand-wavy. We think the decision should be mechanical, not editorial.
Trigger a refresh when: The post still ranks for its primary keyword but has lost AI citations in the past 30 days. Or when a competitor publishes updated content on the same topic. Or when the post's core statistics are more than 90 days old. Content decay is one of the most common silent causes of lost AI Overview visibility as facts and entities age out of trust.
Trigger new production when: No existing post covers the target query. Or when the topic requires a fundamentally different angle (not just updated stats). Or when AI Overviews show citation patterns that favor new sources over incumbent ones.
The ratio isn't fixed. Some months you'll refresh 5 posts and write 3 new ones. Other months you'll write 6 new posts and refresh 1. The point is that the decision is data-driven, not calendar-driven.
A tangent worth mentioning: we've seen teams resist the refresh model because it feels like admitting their original content wasn't good enough. It's not about quality. A post that was excellent 6 months ago with 2025 statistics is legitimately less useful now. Freshness isn't a judgment on the writer.
The Structure Problem Hiding Inside the Cadence Problem
One thing the cadence model doesn't capture is structural optimization. Pages optimized for AI visibility see up to 40% higher inclusion in AI-generated summaries, and a clear answer in the first 150 words is the single biggest factor.
This means refreshes aren't just stat updates. They're structural rewrites. Moving the core answer above the fold. Adding explicit entity markup. Ensuring the first paragraph directly answers the query rather than building to it.
For a two-person team, this is genuinely hard. A structural refresh takes nearly as long as writing a new post. But it produces a page with existing authority signals (backlinks, age, crawl history) plus fresh content signals. That combination is what both surfaces reward.
Benchmarking Against Citation Frequency, Not Rankings
The model we're proposing replaces the standard "keyword position" KPI with two metrics: cost per AI-cited page and citation retention rate (how many of your AI-cited pages maintain citations after 90 days).
Rankings still matter. They drive direct traffic and they correlate, loosely, with AI citation eligibility. But a ranking without a citation is worth measurably less in 2026 than it was in 2024. And a citation without a ranking still drives 35% more clicks than no citation at all.
The two-person team that tracks citation retention rate will make better production decisions than the team tracking keyword positions. We're confident about this, though we'll admit the tooling for tracking AI citations at scale is still immature. Semrush and Ahrefs are building toward it, but as of mid-2026, most teams are doing it semi-manually.
That's a solvable problem. The cadence mistake is the expensive one.
Where This Gets Genuinely Messy
We've presented this as a clean model, but the reality has rough edges. AI Overview algorithms change without warning. Google's March 2026 core update followed a spam update in the same month and a December 2025 core update that itself rewrote visibility for thousands of sites. Any cadence model built on current citation behavior could be obsolete in 6 months.
The honest answer: we don't know exactly how long the 13-week citation decay window will hold. It could compress further. It could vary by industry. What we do know is that the static cadence model is already wrong, and the dynamic one is at least directionally correct.
The teams that will perform best over the next 18 months aren't the ones with the perfect model. They're the ones that built the operational muscle to adjust their production and refresh ratios quarterly, based on actual citation data rather than publishing targets they set in January and forgot to revisit.
References
- AI Overviews and Organic Traffic: What the 2026 Data Actually Shows - Contently
- Content Strategy for AI Overviews: Post-I/O 2026 Guide - Digital Applied
- What Has Changed After Google's Core Update in March 2026? - Medium
- Content Refresh Strategy 2026: How to Win Visibility in AI Search - Ten Speed
- New 5W Research: Overlap Between Top Google Rankings and AI-Cited Sources Has Collapsed - PR Newswire
