Google ships SERP changes every 90 days. Most marketing teams do not.

Three Google platform shifts in ten weeks: the March 2026 core update, the I/O 2026 SERP redesign on 19 May, and the May 2026 core update closing on 2 June. The platform is now rebuilding itself every 90 days. Most marketing teams still run annual plans against a quarterly-shifting target. Here is the operating rhythm that survives it.

The cadence is real

The dates are easy to find. Google's March 2026 core update and spam update ran 27 March to 8 April. Google I/O 2026 on 19 May announced the biggest Search redesign in twenty-five years: Gemini 3.5 Flash as the default AI model, AI Mode hitting one billion monthly active users, and Preferred Sources expanding from approximately 90,000 to over 345,000 domains. The May 2026 core update rolled 21 May to 2 June.

Three platform-level shifts inside a ten-week window. The pattern is not new. Google has been running three to four core updates a year since at least 2023, plus interleaved spam updates and quieter SERP feature changes. What is new in 2026 is the size of each shift. AI Overviews and AI Mode are not refinements. They are structural rebuilds of how the SERP is constructed. Each one rewrites the rules visitors arrive on your site under. I wrote about what that is doing to B2B SaaS traffic in a previous piece. This one is about the operating model that survives it.

The half-life of a marketing plan against Google has gone from twelve months to twelve weeks.

This is not a complaint about Google. They have earned the right to ship as often as they need to. The complaint is about marketing teams who have not updated the operating model to match. Annual planning offsites. Quarterly steering committees that mostly re-confirm last quarter's strategy. SEO frameworks written in 2018 and never re-versioned. Process docs in Confluence that nobody has touched since the last reorg. All of it built on the quiet assumption that the rules will hold long enough to follow them.

The rules will not hold.

Change management was built for a different platform

The frameworks most B2B marketing teams use for change management were written for stability. ADKAR. Kotter's eight steps. Prosci's three phases. All of them dating from the 1990s and early 2000s, when platforms had migration windows and steady states between them. They assume change is the event and steady state is the default. Form the steering committee. Run discovery. Propose the change. Communicate. Train. Deploy. Bed it in.

By the time you have finished proposing the change, Google has shipped two core updates.

These frameworks are not wrong. They were right for the platforms that produced them: enterprise software in the late 1990s, ERP rollouts in the 2000s, SaaS migrations in the 2010s. Platforms with deprecation notices and twelve-month migration windows. The marketing world ran on the same rhythm: annual planning offsites, quarterly reviews against fixed targets, a CMO change every eighteen months.

The platform we are marketing on now does not work like that. Treating it like it does is the operating equivalent of bringing a steering committee to a fire drill.

What breaks when change is the state

Three things, in roughly this order.

Decision-making lag. The standard marketing operating model has too many decision layers for the speed Google is moving at. By the time a recommendation has gone from analyst to manager to director to CMO to steering committee, three weeks have passed. Three weeks is a third of the time you have before the next core update. Most "monthly marketing review" meetings are now decision sinks, not decision tools.

Process documentation as monument. Most marketing teams treat process docs as deliverables: a Confluence page, an SOP, a "playbook" that someone spent two weeks writing and another week being proud of. They go stale faster than they get read. The team learns to ignore them. Tribal knowledge replaces written process. Tribal knowledge is worse than no documentation because it is invisible and load-bearing at the same time.

Channel specialisation. Most B2B SaaS marketing teams are still organised around channel depth: SEO Lead, Paid Lead, Content Lead, Lifecycle Lead. Each role optimises for fluency in a channel that is being restructured every quarter. The SEO Lead who spent five years getting good at technical SEO is now staring at AI Overviews, AI Mode, Preferred Sources, and a buying journey that increasingly skips the SERP entirely. Their expertise has not vanished. It has become narrower than the platform now demands.

The 90-day operating rhythm

This is what works. Each quarter is one cycle. The structure is the same every cycle and the contents are different every cycle. The symmetry is the point.

Decision rights collapse to two layers

At the start of every 90-day cycle, the head of marketing picks two or three opinionated bets. Written down. Specific. Killable. "We will move organic traffic into AI Overview citations on our top twenty commercial queries" beats "we will improve our SEO".

Inside the 90 days, tactical decisions are made by the closest individual contributor with the platform context. The channel owner decides what to test in their channel. No steering committee. No mid-quarter strategic pivots unless something has obviously broken. The split eliminates the meeting where six people half-decide something and then re-discuss it three weeks later.

Process docs become versioned and disposable

Every process document in the marketing team gets a version number and a kill date. The default kill date is twelve weeks from creation. At twelve weeks, the doc either gets re-versioned (still useful, update the date and ship it again) or archived (no longer useful, move on). Nothing in marketing process documentation gets to be more than one cycle old without earning its place.

The hidden benefit: the team stops treating process docs as monuments and starts treating them as tools. Tools are easy to throw away.

Team structures shift from channel depth to platform fluency

The job title "SEO Lead" needs to die or become much wider. Either rename to "Organic Discovery Lead" (covering SEO, AI search, AI Overviews, Preferred Sources) or accept that the role will be specialist in a shrinking surface. The same logic applies to "Paid Lead" once Performance Max and AI-assisted bidding consume more of the day.

T-shapes win when the platform changes faster than career-deep specialisation can re-tool. The vertical bar of the T, your deep craft, needs to be something that travels across channels: analysis, writing, technical implementation, system design. The horizontal bar is platform fluency: staying current on where the surfaces are moving even when you are not the one tuning them this quarter.

The hiring implication is uncomfortable. Most marketing job specs are still written for the channel specialist who did not exist in this form before 2018 and may not exist in this form by 2028. The SEO Lead you hire in Q1 2026 is doing a job that will be forty percent different by Q1 2027. Either they re-tool quickly, or the role gets re-shaped under them. Worth rewriting the spec before you post it.

Insight production becomes the core skill

The 90-day cycle dies without fast, honest insight. But the role of insight inside the quarter is different from the role of decisions inside the quarter. Most marketing teams confuse the two.

The bets are set at the start of the 90 days. Two or three opinionated calls, written down, killable. They do not get re-decided weekly. What gets done weekly is reading the data to see how the bets are progressing, what the platform is doing under them, and what trends are starting to show.

"Are organic clicks on our top twenty commercial queries moving the way we predicted?" is the right weekly question. "Should we change our SEO strategy this week?" is not. The first is monitoring. The second is mid-quarter panic dressed up as analysis.

Honest monitoring still requires the team to go from question to answer in days, not weeks. "Did the AI Overviews citation pattern shift this week?" should not require a three-week Jira ticket to BI. It should be: someone in the team pulls the query, plots the chart, brings it to the next stand-up. Fast insight is the precondition for honest tracking. Slow insight forces the team to either guess or pivot the bet, and the bet is not supposed to move.

Three capabilities make this possible. Most B2B marketing teams have not invested in any of them.

Direct data access. The marketing team has read access to the warehouse, the GA4 BigQuery export, the CRM tables, and the CDP if you have one. Not "marketing requests data from BI". Marketing pulls data from the same place BI does. SQL fluency on the team is a hiring criterion, not a "nice to have".

A small, owned set of dashboards. Three to five dashboards, owned by the team, that answer the questions the team actually asks. Not the 47 dashboards inherited from a consultant in 2021 that nobody opens. Each owned dashboard has a named owner, a refresh cadence, and a kill date on the same clock as a process doc.

A weekly insight rhythm. One named person on the team writes a one-page insight summary every Monday morning. What moved. What did not. Which trends are showing up against the bets. Not a status report. A monitoring log that the whole team reads before the weekly stand-up. The log accumulates across the quarter and becomes the input to the 90-day review.

What this looks like when it is working: a channel owner notices on Tuesday that organic clicks on the target queries dropped 30% over the weekend. They run the SQL, confirm the pattern is correlated with the latest core update rolling out. They log it in the weekly insight summary. The bet keeps running. The note goes into the file that gets read at the next 90-day boundary. No mid-cycle pivot. No emergency meeting. Just a clean record of what the platform did and what the bet did under it.

The team stops arguing about which dashboard to trust and starts noticing what the data is saying. The decision about what to do with that noticing waits for the quarter review. That is where this rhythm earns its keep.

Measurement re-baselined every quarter

Stop comparing Q3 2026 to Q3 2025. The platform that produced last year's numbers no longer exists. Compare Q3 2026 to Q2 2026, with a one-paragraph note on what shifted in the platform between them. Year-over-year is a vanity metric when the platform itself was not comparable across the year. Quarter-over-quarter, with annotated context, is honest.

Fast on decisions, slow on conclusions

A 90-day cycle is not permission to thrash. The fastest way to lose to AI search is to react to every weekly wobble like it is structural. Most are not. SEO has always been a discipline of patience, and that has not changed. What has changed is the cadence of the platform, not the cadence of reliable evidence.

Three guardrails sit on top of the rhythm above. Without them, the 90-day cycle becomes a panic loop.

Wait for the update to finish before you act. Google's core updates take seven to fourteen days to roll out fully. Decisions made on day three of a rollout are based on partial data and unstable rankings. The rule: no structural change to content, technical setup, or paid spend until the update has been declared complete and the data has settled for at least seven days after that. Three weeks of deliberate inaction during a rollout window is the right answer more often than action is.

Require corroboration before declaring causation. A single dashboard moving is not a finding. A finding is the same pattern showing up in two or more independent measurements: organic clicks down AND impressions down AND average position drop AND a shift in branded versus non-branded mix. If only one of those moved, the story is probably wrong. Park the hypothesis and re-check next week. Most of the worst SEO decisions of the last decade were made by teams who saw one chart move and rewrote a content strategy that week.

Protect the evergreen work. Some marketing investments do not move on a 90-day clock. Core content that defines your category. Technical SEO foundations like crawlability, schema, internal linking. Brand authority signals built over years. These do not get torn up because of one bad quarter. The 90-day rhythm operates on top of an evergreen base, not in place of it. If a proposed change touches the evergreen layer, the bar is much higher: corroborated evidence across two quarters, not two weeks.

The discipline is the one most marketing teams find hardest to hold: fast on decisions about what to test, slow on conclusions about what to keep. The team that moves quickly on hypotheses and slowly on permanent changes survives both the 90-day cadence and its own occasional bad weeks. The team that does the opposite ends up running its own personal core update every Monday, while Google ships one every quarter.

What this looks like on a Monday morning

The head of marketing walks into the quarterly planning session with a one-page document. Three bets, written down. A list of what was true about the platform last quarter that is no longer true this quarter. A list of process docs that hit their kill date during the quarter. The team walks out with their channel responsibilities and a calendar of three planned re-baselines. No ninety-page strategy deck. No annual budget reset. No change management committee.

This is harder than it sounds. It requires three things:

1. A head of marketing who can make opinionated calls quickly and live with being wrong fast.
2. A team that does not need consensus to ship.
3. A board that accepts the marketing budget being re-negotiated quarterly against a re-baselined target.

If you cannot get those three, you do not have a marketing function on a 90-day clock. You have a content marketing operation running an annual plan against a quarterly platform. The gap between those two is where your organic traffic is currently going.

Closing thought

The platform is now faster than the planning cycle most marketing teams run. That is the structural problem behind every "our rankings are flat but our clicks are down" chart. The marketing teams that survive the next two years are the ones that figure out how to run a 90-day rhythm. The rest will be writing change-management decks while their organic traffic halves again.

The fix is structural, not tactical. There is no AI search hack, no Preferred Sources cheat code, no "just optimise for AI Overviews" workshop that lets you keep running an annual plan. You have to change how you decide, document, and re-baseline. Every 90 days. Forever.