Google core update: three checks to make before touching your site

Every time Google announces a core update, the same scene plays out in agencies in Kirchberg and in SMEs in the Nordstad: someone opens Search Console the next morning, sees a curve down 12%, and proposes rewriting half the site before noon.

It is almost always a mistake. Not because the drop is not real, but because a core update rolls out over one to two weeks, and the figures on day three do not look like those on day fifteen. We have seen sites lose a quarter of their clicks in the first week and end up higher by the end of the rollout. The reverse happens too.

Before deciding anything, three checks take less than an hour and prevent most decisions made under pressure.

1. Confirm that it really is the core update

The reflex is to blame every fluctuation on the announced update. Yet the ordinary causes have not gone away: a tag migration pushed the same day, a caching plugin that crashed, a robots.txt file changed by a developer who was testing something.

The Google Search Status Dashboard shows the start and end date of each rollout. If your drop began three days earlier, it is not the core update. If it affects a single section of the site while the rest is stable, it is rarely the core update either. These updates apply to an entire domain and shift things in a fairly uniform way.

Another point specific to Luxembourg: filter by country in Search Console. A site that publishes in French, German and English receives traffic from France, Belgium and Germany that reacts differently. A drop limited to the French market, for example, may come down to French competitors ranking better on those queries and say nothing about your Luxembourg audience.

2. Separate the pages that matter from the ones that fill space

Export the pages that lost the most clicks over the period, then sort them into two groups. On one side, those that generate leads, sales or engaged reading. On the other, those that exist because something had to be published that month.

Nine times out of ten, the second group is the one that fell furthest. This is no coincidence: the internal Google documents that circulated in 2024 mention a measure of a site's topical focus and of the distance between each page and the core of the domain. An accounting firm's site that published forty articles on workplace wellbeing to “faire du contenu” (“produce content”) dilutes the signal of what it actually does well.

The answer is not to delete everything during the rollout week. It is to wait until the end, look at what remains, then calmly decide what to merge, what to update in depth and what to remove.

3. Reread three pages as a reader, not as an SEO

Take the three most affected pages in the group that matters and read them on a phone, in the position of a reader arriving from Google who does not know you. The questions Google itself suggests in its documentation on helpful content are uncomfortable enough to be effective:

• Does the text provide information that cannot already be found in the other ten results?
• Does the person who wrote it show first-hand knowledge of the subject?
• If the author's name and the company's name were removed, would there still be a reason to trust this page?
• Does the title promise exactly what the page contains?

For a Luxembourg publisher, the last question often comes down to localisation. An article titled “Les aides à la digitalisation pour les PME” (“Digitalisation grants for SMEs”) that actually describes French schemes will not hold up for long against a page covering Fit 4 Digital or Luxinnovation's programmes, with the country's amounts and conditions.

What not to do during the rollout

Change the titles of fifty pages, alter the URL structure, switch to a new theme. Any technical change launched during a core update makes diagnosis impossible: at the end, you will no longer know whether the rise or fall came from Google or from you.

Updating the publication date without touching the content does not help either. Google has long distinguished between the displayed date and the last significant change to the text. An article dated yesterday but identical to the 2023 version is treated as a 2023 article.

And afterwards

If, once the rollout is over, the drop is confirmed and concentrated on the pages that matter, the work begins: one page at a time, adding what only you can bring. Sites that recover after a core update generally do so at the next core update, not in the weeks that follow. It is slow, it is frustrating, and it is the only method that has worked repeatedly.

We will follow the current rollout and publish an update when it ends, with feedback from publishers in the country who agree to share their figures.