Google March 2026 Core Update: What Actually Changed

The Google March 2026 core update ran from March 27 to April 8, 2026. Google called it a regular update and published nothing else. That silence got filled fast, with named AI filters, new speed thresholds and dramatic volatility scores, and most of it has no source. This article sticks to what Google confirmed and what Sistrix measured. It also covers the May core update, which came seven weeks later and pushed the same pattern further.

When the March 2026 core update ran

For dates, the only record that counts is Google's Search Status Dashboard. It's where Google logs every ranking update it confirms, and for 2026 it lists six so far:

Two details matter if you're reading your own traffic data. The spam update was over before the core update began, so a drop between March 24 and 26 belongs to the spam update, not the core update. And 12 days is an ordinary rollout. Most core updates take about two weeks.

What Google said about the March 2026 core update

One sentence. Google described it as a regular update meant to "better surface relevant, satisfying content" for searchers. There was no blog post, no list of changes and no new guidance. The May update came with the same wording.

That matters more than it sounds. When Google doesn't name a system, every specific claim about what a core update "introduced" is someone's reading of ranking data. Some of those readings are good, but they aren't announcements.

Who won and who lost in Germany

The most careful public analysis for the German market comes from Sistrix. Johannes Beus started with 1,371 domains that moved during the rollout, checked each one against a year of visibility history and threw out everything that couldn't be told apart from normal fluctuation. 166 domains were left: 134 losers and 32 winners. Four losers for every winner. The table shows a selection, with each domain’s Sistrix Visibility Index before and after the update.

The losers cluster by type. 39 of them were online shops. Language tools lost almost everywhere, including verbformen.de, bab.la (−22%) and linguee.de (−10%). Recipe portals lost too, with chefkoch.de the one that held steady. The winners include the airports of Stuttgart, Cologne-Bonn, Hamburg and Munich, plus official sites like hessen.de and arbeitsagentur.de.

Sistrix's international analysis found the same pattern in the UK, France, Spain and Italy. It also called the update small next to earlier core updates. Keep that in mind before reading too much into any single domain.

Why interchangeable sites lost

Sistrix sums the update up as "authority beats interchangeability", and I think that's the right frame. Look at what lost. A conjugation table on verbformen.de is the same table on ten other sites. A recipe portal's lasagne competes with thousands of other lasagnes. An answer on gutefrage.net competes with every other forum answer, and more and more with the AI Overview sitting above it.

Then look at what won. Stuttgart Airport is the source for Stuttgart Airport. Ratiopharm is the source for ratiopharm's own drugs, and nobody can be more authoritative about those searches than they are. Size didn't decide this update. Being the original source for the thing someone searched for did.

Here's the uncomfortable part for publishers. Dictionaries, reference sites and quick-answer pages lost visibility at the same time as AI Overviews answered more of exactly those queries. Sistrix pointed out the overlap without claiming cause and effect, and I won't claim it either. But if your page answers in one sentence what Google can also answer in one sentence, your competitor is Google, not the other sites.

The May 2026 core update followed the same line

Google released the next core update on May 21, and it finished on June 2, again with the "regular update" wording and nothing more. The Sistrix analysis for Germany found a sharper version of the March pattern: generic comparison and service content lost across many domains, and brands with clear product ranges gained.

The most useful detail sits inside a single domain. On nivea.de, the advice section under /beratung/ lost 29% while the product section lost 7%. Coolblue publishes advice too, and it gained. The difference Sistrix points to: Coolblue's advice leads to a purchase on the same site, and Nivea's advice stands on its own. Content tied to what a site actually sells held up. Content that exists mainly to catch traffic didn't.

Reddit is worth a note as well. It lost almost a fifth of its German visibility in May, so treat any "Google loves Reddit now" advice as a snapshot of one update.

Claims about this update that don't hold up

The first version of this article repeated two of these, and they still show up all over search results.

• "The update introduced the Gemini 4.0 Semantic Filter." No Google source uses that name. Google didn't name any system for this update, and the phrase only appears on SEO blogs that cite each other.
• "A good INP is now 150 ms." Google's INP documentation still rates 200 ms or less as good and anything above 500 ms as poor. Core Web Vitals thresholds don't change inside a core update. The switch from FID to INP was announced on web.dev months ahead.
• "The rollout ran from March 6 to March 27." It started on March 27 and ended on April 8, according to the Search Status Dashboard.
• "Volatility hit 9.3 out of 10." I couldn't trace that number to any tracker. Sistrix, which publishes its method, described the update as small.
• "Information gain became a ranking factor in March." Google holds a patent on information gain scoring, but a patent shows a system could exist, not that it runs in Search. Google hasn't confirmed it.

All five follow one pattern. A specific-sounding detail travels further than "Google didn't say". So when a claim about a core update names a system, a threshold or a percentage, look for the Google page it came from. Usually there isn't one.

The February Discover update

This one was new: the first update Google ever announced for Discover alone. According to Google's announcement, it shows more locally relevant content from sites based in the reader's country, less sensational content and clickbait, and more in-depth, original reporting from sites with expertise in a topic.

It launched for English-language users in the US, with other countries and languages to follow. For German publishers that means one practical thing: before you blame a Discover drop on this update, check whether Google has confirmed the rollout for German. Once it arrives, the local part could help DACH publishers, because it favors sites based in the reader's own country.

How to recover from a Google core update

Start with Google's core update guidance, because it's calmer than most recovery checklists. Most sites don't need to do anything. A small drop doesn't mean something is wrong, and Google advises against changing content that's already performing well. Improvements can take several months to show, sometimes not until the next core update.

If you did lose real traffic, this is the order I'd work in:

1. Compare the right windows. In Search Console, put two or three weeks before March 27 against two or three weeks after April 8, and leave the rollout itself out. A drop that started before March 27 has a different cause.
2. Group losses by page type, not by URL. The March pattern hit templates: conjugation pages, recipe pages, generic category pages. If one type lost and everything else held, you've found it.
3. Ask whether each losing page is interchangeable. Would a searcher notice if your page were swapped for the next result? If not, rewording won't help. Adding something only you have can, like your own data, your own tests or first-hand photos.
4. Tie advice to what you offer. The Nivea and Coolblue split from May is the clearest example in the data. A guide that leads into your own product or service is sturdier than one written only to rank.
5. Merge pages that compete with each other. Two thin pages on one topic are two interchangeable results. See the keyword cannibalization guide for how to find them.

The GSC Dashboard runs the window comparison in your browser, and the E-E-A-T guide goes deeper on showing first-hand experience.

FAQ