March vs May 2026: Pattern Analysis

Google today began rolling out the May 2026 core update — 43 days after the March 2026 update completed on April 8. That 6-week gap is the tightest cadence between distinct broad core updates since the November–December 2024 calibration pair, and the pattern data from March gives the clearest leading indicator available for what May is likely targeting.

March 2026 was, by SE Ranking's measures, the most volatile core update on record: 79.5% of top-3 URLs shifted positions (versus 66.8% in December 2025), 24.1% of top-10 pages dropped entirely out of the top 100, and YouTube alone reportedly lost 567 visibility points in Lily Ray's Amsive dataset. But the headline numbers obscure a methodology story that almost no coverage addresses — Sistrix UK and Amsive's analysis disagree sharply on whether Reddit and Instagram were winners or losers, and the reason why matters for how you interpret any winners-and-losers list, including the ones that will emerge for May.

This analysis covers the March 2026 pattern in full, explains the tool-divergence dynamic, maps the cadence acceleration against the 2019–2026 historical record, assesses early signals for May 2026, and identifies which March patterns are most likely to repeat. For the May 2026 launch-day breakdown, see the companion post. For a step-by-step recovery framework, see the 14-day core-update recovery playbook.

The arithmetic is simple and the implications are not. The March 2026 core update completed on April 8, 2026 — confirmed by Sistrix, Search Engine Land, and Search Engine Journal. The May 2026 core update began rolling out on May 21, 2026 at 08:40 PDT. That is 43 days, or approximately 6.1 weeks.

For context: the minimum gap between distinct broad core updates in 2025 was approximately 13 weeks (the March 2025 → June 2025 interval). The 2024 minimum was the November → December 2024 pair, which closed in roughly 30 days — but those updates are widely interpreted as a single Q4 calibration cycle rather than two independent broad core updates. Setting aside that anomaly, the next-shortest gap in 2023–2025 was approximately 70 days (August → November 2024). The March → May 2026 interval at 43 days is shorter than every comparable gap in the dataset.

What the compression means in practice: sites that were still in the volatile, unsettled phase of March recovery — rankings fluctuating, traffic counts still normalizing — now face fresh signal-level shifts before the previous update had fully resolved. Per PPC.land's May 21 coverage, sites that suffered 70–85% traffic declines in December 2025 and had not yet fully recovered through March were already absorbing overlapping signals when May launched.

The March 2026 core update (March 27 → April 8, 12 days 4 hours) ran concurrently with a March 2026 spam update, which complicated attribution for sites that experienced both ranking and manual-action signals simultaneously. The March 2026 impact analysis and recovery roadmap covers the spam-update interaction in full; this post focuses on the core update's content-quality and authority pattern.

The sector-level pattern per Lily Ray's Amsive analysis across 2,076 domains is the clearest picture available. In entertainment, streaming-service home sites (Netflix, Spotify, Apple Music pages) gained visibility while "what to watch" aggregators like JustWatch (-24%) and Rotten Tomatoes (-8.5%) declined. In travel, hotel chains and airlines improved while OTAs — Booking.com, Expedia — declined. In jobs, government job sites and employer career pages surged while aggregators (Indeed -18%, ZipRecruiter -21.6%) fell. In health, government and non-profit sources (NIH.gov, FDA.gov) gained while traditional publishers (WebMD, Mayo Clinic) declined. In finance, government and brand-owned financial sites rose while comparison platforms (NerdWallet, CreditKarma) fell.

The pattern is not subtle. In every vertical where the data is granular enough to see it, first-party destination brands gained and intermediary aggregator models lost. The degree of aggregation — how far a site sits from the original source of the information or transaction — appears to have been the primary risk signal, not any specific content tactic or technical element.

The dictionary and knowledge-database losses deserve a separate read. Wiktionary fell -14.96%, Collins Dictionary -12.87%, Britannica -9.00%, and Wikipedia -1.13% in Sistrix UK data. These are not low-quality sites by any traditional measure. The likely explanation — and this is an analytical inference, not a confirmed Google statement — is that a significant proportion of the informational queries those sites rank for are now being answered by AI Overviews above the blue links. The March update may have de-valued reference pages whose primary value to users is informational look-up, precisely the query type AI Overviews now satisfies without a click. See also the content-quality signals 2026 core updates reward for the E-E-A-T framing in depth.

The most underreported aspect of the March 2026 core update is the extent to which the two major public analysis datasets — Sistrix UK Radar and Lily Ray's Amsive analysis — contradict each other on specific domains. This is not a data error. It is a methodology story, and understanding it is essential for interpreting any winners-and-losers list, including the ones that will emerge for May 2026.

Sistrix UK Radar measures percentage visibility change for the UK market specifically, across Sistrix's keyword index. Lily Ray's analysis at Amsive measures absolute visibility-point changes across a global dataset of 2,076 domains. The two methodologies answer different questions: Sistrix tells you what happened in one geography as a percentage of baseline; Amsive tells you which sites lost the most absolute search presence globally in raw point terms.

The Reddit and Instagram rows are the clearest illustration of why winners-and-losers lists must always be attributed to a specific tool and methodology. When a post says "Reddit won in March 2026," it means "Reddit gained in UK percentage visibility in the Sistrix dataset." When a post says "Reddit lost in March 2026," it means "Reddit lost 64.2 absolute visibility points in Amsive's global 2,076-domain dataset." Both statements can be simultaneously true.

This divergence is not a new phenomenon — tool disagreements on core-update outcomes have been documented since at least 2019. What makes the March 2026 case particularly instructive is the magnitude of the disagreement: Reddit going from "+2.4%" (Sistrix) to "−64.2 points" (Amsive) is not a marginal discrepancy. It is a sign-flip. When you read May 2026 winners-and-losers coverage — which will begin emerging in approximately two weeks — look first for which tool the analysis cites, and treat any coverage that mixes tool sources without attribution as unreliable.

The most durable analytical frame for March 2026 — and the strongest leading indicator for May — is what we can call the first-party correction thesis. Three analysts, working from different datasets and methodologies, independently describe the same underlying pattern.

Aleyda Solis, quoted in Sistrix's March 2026 analysis, described it as "a shift away from intermediary and aggregator sites toward stronger destination brands and institutional sources." Lily Ray, in Amsive's analysis, framed it as a "first-party, official-source correction." The Sistrix team described their own data as reflecting Google's preference for "authority over interchangeability" — meaning sites where the content could only reasonably come from that specific entity, rather than sites that aggregate or republish content available from primary sources.

The three framings are functionally identical. Google's ranking signal appears to have shifted weight toward sites that are the original and authoritative source of the information or transaction, and away from sites that add a layer of aggregation, comparison, or curation over primary sources. The degree of intermediation — how many steps a site is from the entity that actually created or owns the underlying information — correlates with risk in this update cycle.

The "Google runs core updates every 3–4 months" rule of thumb was accurate for the 2019–2023 period. The 2024–2026 data invalidates it. Per Search Engine Land's confirmed-update library and Search Engine Journal's algorithm history, the actual cadence by year is as follows — note that the 2022 figure (two updates) is the outlier low, not the baseline.

The 2026 cadence, if maintained, puts the year on pace for five to seven broad core updates — double the 2022 figure and 25% more than the 2023–2024 peak. That is not a projection based on any Google statement (Google has not said anything about planned update frequency). It is a straight-line projection from the observed 43-day interval. If the second half of 2026 maintains the same 6-week cadence, five more updates would land before December 31.

The more important observation is the direction of the trend. From 2022 onward, the average gap between consecutive updates has been compressing. The 2022 inter-update gap averaged roughly 117 days; the 2023–2024 average was 65–70 days; the 2025 average was approximately 90 days (the three updates were spread unevenly through the year). The March → May 2026 gap at 43 days represents a continuation of that compression trend, not an aberration.

From a practical SEO perspective, the cadence compression means that the window between updates for recovery, measurement, and course-correction is shrinking. Sites that took three months to analyze what changed and implement fixes after December 2025 are now being overtaken by the next update before they could confirm whether the fixes worked. This is the operational argument for investing in continuous SEO analytics infrastructure rather than post-update triage cycles.

The May 2026 core update is hours old as of this writing (launched 08:40 PDT on May 21). No meaningful winners-and-losers data exists yet — ranking volatility tracking tools typically require 5–7 days of post-announcement data before the signal stabilizes enough to draw directional conclusions, and the full rollout is not projected to complete until June 2–4. Any coverage claiming to identify May 2026 winners on May 21 is speculating from tool noise, not measuring update effects.

What we do have is structural signal from the update's launch context. Google used identical boilerplate language for May: "Released the May 2026 core update. The rollout may take up to 2 weeks to complete." Google Search Central posted the same framing to X and LinkedIn: "This is a regular update designed to better surface relevant, satisfying content for searchers from all types of sites." This is verbatim the same language used for March 2026. No companion blog post was published for either update — the Search Status Dashboard one-liner is the entirety of official guidance for both.

Two contextual factors complicate May 2026 impact diagnosis in ways that were not present for March. First, Google's AI Mode reportedly surpassed one billion monthly users and the global Gemini 3.5 Flash deployment completed around May 19–21 — creating overlapping ranking signal shifts and user-behavior changes (click-through rate patterns, query reformulation rates) that will be difficult to isolate from the core update signal in early volatility data. Second, sites recovering from December 2025's significant traffic declines are still in a partially-stabilized state — their March recovery curves have not fully settled, which means the May signal will layer on top of unresolved March noise for many affected domains.

Drawing on the March 2026 pattern, the historical cadence of back-to-back updates, and Google's unchanged official guidance, the following projection framework distinguishes the most likely continuations from the most plausible changes.

Likely to repeat: The first-party correction pattern. Every confirmed signal from the March cycle — Solis's aggregator-shift framing, Ray's first-party-correction framing, Sistrix's "authority over interchangeability" framing — is consistent with the direction of Google's stated goals around E-E-A-T (experience, expertise, authoritativeness, trustworthiness) and its emphasis on "helpful content" as the core evaluation framework. None of those goals changed between March and May. The March pattern is the best available prior.

What may intensify: The AI Overviews displacement effect on reference and informational content. Dictionary sites, knowledge databases, and information-lookup pages continued to lose visibility in March even though they are high-quality by traditional standards. As AI Overviews continues to expand query coverage — particularly for informational and definitional queries — the incentive for Google to rank those reference pages highly in the blue-link results diminishes. May may continue or accelerate this displacement, particularly if AI Mode's query volume growth since March has expanded the set of informational queries AI Overviews now answers directly.

What is most uncertain: How the update interacts with sites that already lost significantly in March. Historical precedent from back-to-back update pairs (e.g., the August → November 2024 sequence) suggests that sites hit hard in the first update sometimes stabilize or partially recover in the second, if the first update's signal was over-applied. But there is no systematic rule — Google does not guarantee or target recovery in a follow-up update. Sites that lost in March should not assume May will restore their rankings; sites that gained in March should not assume May will extend those gains.

The Sistrix vs Amsive divergence on Reddit and Instagram is a clean pedagogical example of a problem that affects every piece of core-update analysis: the numbers are tool-specific, geography-specific, and methodology-specific, and treating any single tool's output as ground truth produces misleading conclusions.

Before drawing conclusions from a winners-and-losers list, four questions are always worth asking. First: what is the geographic scope? UK-only datasets (Sistrix UK Radar) will show different results than global datasets (Amsive's 2,076-domain global analysis) because ranking systems vary by market. A site that gained in UK results may have lost globally, or vice versa. Second: what is the measurement unit? Percentage visibility change and absolute visibility-point change answer different questions. A small site that doubles its visibility shows a large percentage gain but a small absolute point gain; a large platform that loses 10% of its visibility shows a large absolute point loss but a moderate percentage change. Third: what keyword set is included? Different tools index different keyword universes, and the keyword selection heavily influences which domains appear on the winners and losers lists. Fourth: what time window? Winners and losers lists published during a rolling update — particularly in the first week — reflect noise more than signal. Wait for rollout completion.

The ideal methodology is what both Sistrix and Amsive practice individually and what this post practices comparatively: attribute every data point to its tool and methodology, note geography explicitly, distinguish percentage change from absolute change, and flag when multiple tools give contradictory signals rather than picking the most convenient one. The content-quality signals 2026 core updates reward uses the same cross-tool methodology for E-E-A-T signal analysis.

One additional methodology note specific to the March–May comparison: the March 2026 update ran concurrently with a March 2026 spam update. When a core update and a spam update run simultaneously, volatility tool readings reflect both signals combined. Sites that experienced ranking changes in March cannot definitively attribute those changes to the core update versus the spam update without additional signal — specifically, whether they received a manual action notice in Google Search Console, and whether their link profile or content practices are consistent with spam-update targets. The May 2026 update appears to be running without a concurrent spam update (no spam update has been announced as of this writing), which should make attribution cleaner once rollout data is available.