The May 2020 Core Update: Why You Must Never Diagnose a Core Update on Traffic Alone

A big update, at the worst possible moment to read it

The May 2020 core update is the cleanest case study in the entire core-update history for one reason: it proves you must never diagnose a core update on traffic alone.

On May 4, 2020, Google began rolling out a broad core algorithm update. By every tracker that measured it, this was big - the second-largest update of 2020, behind only December that year. Analysts measured ranking swings of 10 to 20 percent for top results, against the 1 to 3 percent that an ordinary update moves, and entire domains lost more than half their visibility in a single sweep.

That alone would make it memorable. What makes it instructive is the calendar. The update shipped into the middle of the global COVID-19 lockdown, a moment when query demand, click behavior, intent, and seasonality were all already broken. People were not searching for flights, concert tickets, or gyms. They were searching for symptoms, sourdough, and home workouts. So when a music site or a travel aggregator watched its traffic fall off a cliff in early May 2020, there were two completely different possible causes overlaid on the same graph: Google had re-scored the site down, or the real-world demand behind those queries had simply evaporated.

The timeline and the announcement

There was no dedicated blog post for this one. The May 2020 update predates Google's Search Status Dashboard, whose core-update incident log only began in 2021, so the first-party record is a pair of tweets from the Google SearchLiaison account plus Google's standing core-updates guidance.

Later today, we are releasing a broad core algorithm update, as we do several times per year. It is called the May 2020 Core Update. Our guidance about such updates remains as we've covered before. Please see this blog post for more about that.Google - @searchliaison, May 4, 2020

The rollout began at roughly 3:50pm ET on May 4 and was confirmed complete at 12:37pm ET on May 18, when SearchLiaison posted that "The May 2020 Core Update rollout is complete." That is a two-week window, which is normal for a broad core update.

Note what Google did not say. There was no update-specific guidance and no new advice. Google pointed to its generic 2019 core-update post, the same standing message it repeats every time: there is nothing wrong with pages that drop, the work is to keep making genuinely good content. The rollout shape was the one quirk worth flagging - several analysts noted a quiet day one followed by an enormous day-two spike in volatility, so anyone who checked their rankings on May 5 and breathed a sigh of relief got a shock on May 6.

A clear vertical signature, and a geographic split

The winner and loser data for May 2020 is unusually well documented, better than most 2020 updates. A strikingly clean vertical pattern emerged in the UK: music, streaming and entertainment sites were hammered, while reference, dictionary, news-medical and large aggregator or directory sites surged. The figures below come from SISTRIX's UK Visibility Index, with effects first visible around May 7 and data through May 12.

The pattern was geographic as well as topical. In the UK, music was the common thread among losers. In Germany, the loser list was dominated instead by health-related sites - the same broad update produced a different vertical signature by country. That split is itself a clue: a single algorithm change expressing differently across markets tells you the update was a genuine broad re-assessment of quality and relevance, not a single targeted lever.

The case-study work backs this up. Lily Ray, then at Path Interactive, catalogued roughly 550 winners and losers across nutrition, fitness, news, medical science, banking, music and natural medicine. Glenn Gabe of GSQi published four detailed case studies: a health and medical site that recovered about 53 percent after a 14-month fix cycle, a news publisher up about 35 percent, an affiliate site down 40 percent on aggressive disruptive ads and thin content, and a news publisher that compounded a January-update loss to be down more than 90 percent. Even Search Engine Roundtable, the industry's own watering hole, took a hit - a useful reminder that broad means broad.

Separate the signal from the noise before you touch anything

Here is the part that still matters in 2026. Because COVID-19 was simultaneously reshaping query demand, the traffic deltas around this update conflated two completely different things: Google re-scoring quality and relevance (the signal) and the real-world demand behind queries collapsing (the noise). Several careful analysts, Gabe and Marie Haynes among them, warned at the time that raw winner and loser traffic figures had to be treated with care for exactly this reason.

The owners who got hurt worst in the months after May 2020 were not the ones the algorithm demoted. They were the ones who panicked at a traffic graph, assumed a phantom penalty, and gutted perfectly good content chasing a problem that did not exist. A music venue listings site did not have a quality problem in May 2020. It had a no-live-music problem. Rewriting its pages would have fixed nothing.

The diagnostic discipline that separates the two is mechanical, and it is the single most transferable lesson from this update.

1. Segment by query category before you read any total. Split traffic into the topics and intents you actually serve. A sitewide line graph hides everything; a per-category view shows you which segments moved and which held.
2. Compare share of the visible SERP, not absolute clicks. If your clicks fell but your average position and impression share held, demand dropped and you did not. If your position fell while demand was flat, that is the algorithm talking.
3. Check whether real-world demand for the topic moved. Cross-reference demand trends for your category. Travel, events, gyms and venues all cratered in spring 2020 for reasons that had nothing to do with Google.
4. Only then look at content and quality. If, after isolating the noise, a specific set of pages genuinely lost ranking share against stable demand, that is your signal. Now you have a real target worth the work.

The confound never went away - it multiplied

You might think a 2020 lockdown lesson expired the moment the world reopened. It did the opposite. The confound May 2020 exposed - traffic graphs that mix algorithmic re-ranking with demand and behavior shifts - is now permanent, and in 2026 there are more layers of noise sitting on top of any core update than there were six years ago.

• AI Overviews: answer boxes now absorb clicks that used to reach your page, so a ranking that held can still show falling clicks. That looks identical to a demotion on a traffic graph.
• Zero-click behavior: more queries resolve in the SERP without a visit, adding a second demand distortion that has nothing to do with whether Google re-scored your quality.
• Seasonality and demand cycles: these never stopped. Every category has its own demand curve, and core updates routinely land in the middle of one.

So the May 2020 method is not nostalgia. It is the baseline procedure for reading every update since. The sibling updates from the same era reward the same discipline - see the January 2020 core update and the larger December 2020 core update that closed out the year - and the modern AI-search shift is covered in our piece on Google AI Mode and the post-blue-link era.

How we read it in the accounts we audit

When a core update lands and a client account moves, the first thing we do in the accounts we audit is refuse to answer the question "did we get hit?" until we have segmented the data. A sitewide traffic number is the least useful figure in the building. The May 2020 update is the story we tell to explain why.

• No content changes on day one: rollouts take weeks and rankings churn the whole time. May 2020's quiet-day-one, big-day-two shape is the warning. Reacting to a partial rollout is reacting to noise.
• Demand audit first: before any content work, we confirm whether the topic's real-world demand moved. If demand fell and rankings held, there is nothing to fix.
• Category segmentation always: we read movement by query category and by SERP share, never by total clicks, so we can see the signal under the noise.
• The real target is durable quality: once the noise is stripped out, if specific pages genuinely lost share against stable demand, that is where the work goes - and it is the same work every core update has rewarded since.

For the full procedure, our step-by-step core-update diagnosis framework turns this discipline into a repeatable checklist. May 2020 is six years old and no site is still waiting to recover from it specifically. But the lesson it taught is the one we apply to every update that has come since.