August 31, 2026

What Google Discover is, and why it decides the day in a newsroom

There is a scene that repeats in any digital newsroom. It is eleven in the morning, someone looks at the realtime panel and sees 4,000 people inside. Nobody has done anything different. A piece has entered Discover.

And in the afternoon it repeats the other way: the same panel, 300 people, and the same look of bewilderment.

That is Discover. For many publishers with a mobile-heavy audience it is between 20% and 40% of total traffic, and it is the channel they control the least. This guide explains what it is, how a site gets in, what changed in February 2026, and — the most useful part — what you can do and what you cannot.

What Google Discover is

It is Google’s personalised content feed: it appears in the Google app, on the home screen of many Android phones, in Chrome on mobile, and, since 2025, on desktop as well, in a rollout that has been arriving country by country.

Google’s own definition is that it shows content related to a person’s interests, based on their Web and App Activity. The important part is what it does not say: there is no query. Nobody searched for anything. Google decides, on its own, to show your article to someone who was looking at their phone.

If you come from the Google News guide, that is the difference: News and Top Stories answer a search; Discover gets ahead of it. They are different systems, with different reports in Search Console, and the same publisher can do well in one and badly in the other in the same week.

What follows from there being no query

Three things, and all three hurt:

You do not rank. There is no keyword to work and no SERP to analyse. You can improve your odds; you cannot control the channel. Anyone selling you “Discover ranking” is selling smoke.

It is volatile by design. Google says so in its own documentation: because it is served proactively, Discover traffic varies more than Search. A 500% spike does not mean you got something right, and a 60% drop does not mean you broke something.

It arrives with less intent. Whoever comes from Discover was consuming, not searching. It is scroll traffic. That conditions everything that happens after the click, and it is why a record day in pageviews can be a bad day of real reading.

How a publisher gets into Discover

Same as News: you do not apply. Any content that is in Google’s index and meets Discover’s content policies can appear. No special tags or structured data are needed to be eligible.

Two nuances that save a nasty surprise:

Meeting the requirements does not guarantee you will appear. Google uses those words.
Old content can appear if it is useful and relevant to someone. Discover is not only last hour: a feature from eight months ago can blow up your traffic on a Tuesday.

And something that is constantly mixed up: the news sitemap does nothing here. It is a Google News requirement. Touching it to “fix Discover” is time wasted.

What you do control

Little, but it is real, and almost nobody does it all the way.

The image. It is the factor with the most direct impact. Google asks for at least 1,200 px wide, more than 300,000 pixels in total, a 16:9 ratio, and the image enabled with max-image-preview:large or via AMP. Never the publisher logo. And check the crop: a perfect photo can lose its head on the card.

The headline. It has to reflect the content. Google explicitly advises against sensationalism and tactics that exploit morbid curiosity, titillation, or outrage. Since February that is no longer a polite recommendation.

Authorship and demonstrable experience. A byline, an author page, who we are, contact. That is what lets Google’s systems attribute experience on a topic to you.

Speed. Pure mobile traffic. An article opening that takes four seconds loses half the readers before it starts.

The 2026 change: the Discover core update

On 5 February 2026 Google launched the first core update dedicated exclusively to Discover, without touching Search results. The rollout finished on 27 February. It started with English-speaking users in the United States and is spreading gradually to other countries and languages.

Two ideas sum it up:

Expertise topic by topic. The systems identify experience by area, not as one global score for the domain. A local paper with a gardening section kept up over time can show expertise in gardening; a film-review site that publishes one stray article on gardening cannot. For a small publisher this is good news: depth in what is yours is worth more than breadth.

Geographic proximity. Discover favours content from sites in the reader’s own country. Google was explicit: sites outside the United States that publish for a US audience can lose traffic, and over time that content should rise in its local regions.

The reading for Spain is that this is one of the few Google decisions in years that plays in favour of the local publisher. The practical consequence: make your location unmistakable — a physical address, contact, a who-we-are that says where you are — and keep your local coverage original, not agency leftovers.

The uncomfortable change: Discover is no longer only news

In parallel the feed has been turning into a hybrid, with video, social posts, and creators competing for the same space publishers used to occupy. It is obvious in the United States and it is increasingly visible in Spain.

And in August 2026 Google announced that the feed can be personalised in natural language: the reader writes what they want to see — less celebrity, more technology analysis, less of one subject — and Discover adjusts. It comes in the same package as the Preferred Sources button for publishers.

Add it up: less serendipity, more weight on the brand. The reader who chooses you explicitly is worth much more than the reader who crosses you by chance. That favours whoever has an identity and punishes whoever lives on interchangeable headlines.

The risk almost nobody talks about

Discover is the only traffic source that can disappear on a Tuesday with no warning and without you having done anything. Chartbeat data in the Reuters Institute report put the drop in Discover traffic to US publishers at about 29% year on year through 2025.

A publisher that takes 40% of its audience from Discover does not have a channel: it has a dependency. The strategic conversation is not “how do I get more from Discover”, it is “what share of my audience arrives without going through Google, and how do I raise it”.

How it is measured

In Search Console, the performance report, the Discover tab. It only appears if your site has generated relevant traffic from the channel in the last 16 months; if you do not see it, you have not yet entered with enough volume.

Three things worth knowing when you read it:

There are no queries. You can filter by page, country, and appearance type, but not by keyword, because there is none.
It includes Discover traffic in Chrome, not only in the app.
Today’s figures are preliminary and they change.

And a hygiene rule: when a core update is rolling out, Google recommends not making emergency changes. The volatility of the first 72 hours says nothing. You wait until the rollout ends, and then you look.

Mistakes that keep repeating

Using the logo, a generic montage, or an image under 1,200 px.
Not checking the crop on the card.
A clickbait headline — exactly the pattern the February update punishes.
Reacting in the heat of a Discover drop by rewriting half the site.
Measuring Discover with Search’s stick and not understanding why the reader’s behaviour is worse.
Mixing it up with Google News and touching the news sitemap.
Celebrating the spike and not asking what happened afterwards.

A 30-minute checklist

Open the Discover report in Search Console. If it is not there, you already know where you start.
Take your five most-viewed pieces from last month and look at their image: width, ratio, and the real crop.
Check that max-image-preview:large is in the template, not only on some pages.
Check that your who-we-are and your contact make it unmistakable where the newsroom is.
Pick two or three topics you want Google to recognise you for, and see whether you publish on them steadily or in bursts.
Work out what share of your audience depends on Discover. That number is your risk level.

The spike is not the result

Discover will tell you how many people came in. It will not tell you if anyone stayed. And on a scroll channel, that is the whole difference: 4,000 visits that leave without opening a second page are worth less than 400 that read the piece through and keep going.

That is where Desvelio works: a HUD on your own site that, on the piece you are looking at, tells you how many people are reading now, how long they really stay, how many leave without opening anything else, and which door each one came through. When the Discover spike arrives, that is the information that decides if it is worth repeating tomorrow. If you want to see that half on your own homepage, the 30-day trial is already open.

Frequently asked questions

Do you have to do something to appear in Google Discover?

There is no sign-up and no application. Any indexed content that meets Discover’s content policies can appear, with no special tags or structured data. Meeting all of that still does not guarantee you will appear.

Is it the same as Google News?

No. News and Top Stories answer a search; Discover does not. They are separate systems with separate reports in Search Console.

What size should the image be?

At least 1,200 px wide, more than 300,000 pixels in total, a 16:9 ratio, and enabled with max-image-preview:large or AMP. Never the logo.

Why do I not see the Discover report in Search Console?

Because it only appears when the site has had relevant traffic from Discover in the last 16 months.

Can old content appear?

Yes. Discover can bring back pieces from months ago if they are useful and relevant to a particular reader.