Google Discover: What It Is, How Content Gets In, and What Search Console Measures

On September 14, 2026, we typed “google discover” into Google in Brazil, where we are based. Right below the first link, People Also Ask lined up four questions. How to turn it on, what it is, how to turn it off, and where to find it. ¹

These are questions from people who found the feed on their phone without asking for it. Site owners arrive with different ones. Why did my page show up there yesterday and vanish today? Where do I see that?

This piece covers Discover as a product, through Google’s documentation as read on September 14, 2026. What it is, where it appears, how content gets in, what the rules ban, how to measure it, and what changed. Nothing here promises that your page will appear in it.

What Google Discover is

Think of the guy at the corner newsstand who knows you. When you walk by, he has already set aside the sports section and the new fishing magazine. You asked for nothing. He remembered what you usually take.

On your phone, Google plays that role. It lines up articles and videos on your interests before you search for anything. This is called Discover.

The documentation for site owners defines it this way. Discover is a part of Google Search. It shows people content related to their interests, based on their Web & App Activity. ²

The name has a date. On September 24, 2018, Google gave the name Discover to the feed it had launched the year before. ³ The title of the announcement summed up the idea: new information, with no query required. ³ The launch started in English and Spanish, in the US. ³

The same post brought an idea that still holds. The feed also shows older articles, as long as they are new to the reader. ³ Today’s documentation repeats the rule. Older content may appear if it is helpful and relevant to the person’s interests. ²

Where Discover appears

Google Search Help lists three places. The Google app. Google.com in the browser on an Android phone. And, on some devices, a swipe right from the home screen. ⁴

There is a fourth, in Chrome. Discover also appears on the browser’s new tab page, on Android and on iPhone. ⁵ From February 2021 on, that traffic started to flow into the Discover report. ⁵ Discover’s content policies give this Chrome surface a name, “Articles for you.” ⁶

Desktop joined the list little by little. Today’s help page also shows how to turn off Discover on google.com on a computer. And it warns that the feature is not available in all countries or languages. ⁷

The card itself changed, too. In September 2025, Google announced posts from X and Instagram and YouTube Shorts in the feed, next to articles. ⁸ In October 2025, it rolled out AI-powered previews of trending topics. The post lists the US, South Korea, and India. ⁹

And there are ads. The settings help page is blunt about it: you can’t use Discover without ads. You can change which ads you see. ¹⁰

How content gets in, with no one searching

In search, a person types a query and Google looks for answers. In Discover, the order flips. Google starts from what it knows about the person and picks what to show.

The Search Console help describes this difference. Google serves Discover proactively, without depending on anyone’s query. ¹¹

Google Search Help explains where that knowledge comes from. Google uses information from the device, from other Google products, and from the Google Account. ⁴ That includes Web & App Activity and personal results. ⁴ It also includes information from apps and contacts on the device, and location. ⁴ Without a home location, Google estimates it from other sources, such as the IP address. ⁴

The person also teaches the feed. They can follow a topic, say they are not interested in it, or ask not to see a source. ⁴ They can like a card. Google warns that these choices may be shared with publishers anonymously. ⁴

On August 20, 2026, Google announced one more way to teach it. In the app, people will be able to write, in their own words, what they want to see more or less of. ¹² The announcement says “in the coming days” and names no country. ¹²

On the site side, none of this goes through keywords. There is no query to target. There is each person’s interest, which changes over time. ²

What the “Get on Discover” page asks for

Google’s guide for site owners is called “Get on Discover.” ² It was last updated on March 9, 2026. ²

Eligibility. A page becomes eligible for Discover automatically when it is indexed and meets the feature’s policies. ² It needs no special tag and no structured data. ² And the documentation warns that being eligible is no guarantee of appearing. ²

The same signals as search. Discover uses many of the same systems and signals that Search uses to find helpful, people-first content. ² Anyone looking for success there, Google writes, should read its advice on helpful content. ² What Google wrote about quality is in the piece on E-E-A-T.

The title. The guide asks for page titles and headlines that capture the essence of the content. ² And it closes the shortcut. No misleading or exaggerated details in the title, snippet, or image to inflate clicks. ² No holding back the information that explains what the content is about. ² No sensationalism that plays to morbid curiosity, titillation, or outrage. ²

What the guide asks for instead fits in one sentence. Content that is timely for current interests, tells a story well, or brings unique insights. ²

The big image. Here is the most technical part of the guide. According to it, large images are more likely to generate visits from Discover. ² The recommendation comes with numbers. At least 1,200 pixels wide, more than 300,000 pixels in total, and a 16:9 aspect ratio. ²

Think of a storefront window with room for two photo sizes. The store can use the big photo only if the product’s maker allows it. Without that permission, the thumbnail goes up.

On the page, that permission is a robots rule. It takes three values: none, standard, and large. ¹³ With large, the image preview can fill the full width of the screen. ¹³ Without the rule, Google may show a preview at the default size. ¹³ This is called max-image-preview:large. ^{2 13}

In practice, it is one line in the head of the page, <meta name="robots" content="max-image-preview:large">. The guide also accepts AMP as a route. ²

Notice the design. To get into Discover, no tag is required. ² For the big photo, the page has to grant the permission, or use AMP. ^{2 13}

Which image. Google tries to crop the image for Discover on its own. ² To point to the preferred one, the guide accepts schema.org markup or the og:image meta tag. ² Since March 2, 2026, the image documentation spells out the routes. ^{14 15} They are the primaryImageOfPage property, the image property tied to the page’s main entity, and og:image. ¹⁴ How schema.org markup works is in the piece on structured data.

Two bans come with it. No generic image, such as the site’s logo. And, for best results, no text-heavy image. ²

What Discover filters out. The feed wants content that suits interest-based feeds, such as articles and videos. ² The guide names what it might not recommend. Job applications, petitions, forms, code repositories, and satire without context. ² Beyond SafeSearch, Discover filters out what might seem shocking or unexpected. ²

What Discover’s policies ban

The policies have their own page in Google Search Help. ⁶ Read them in English, even if you use Google in another language. The Portuguese version, for one, carries an AI translation notice. And, as read on September 14, 2026, it lacks the section on followed sites. ⁶

The general rules. Content cannot violate Search’s overall policies or its spam policies. ⁶ Ten policies for Search features also apply. Dangerous content, deceptive practices, harassment, hate, manipulated media, medical content, sexually explicit content, violent extremism, violence and gore, and vulgar language. ⁶

Discover’s own rules. Three belong to Discover alone. ⁶

The first covers ads and sponsorship. Advertising on the page should not exceed the content. ⁶ Sponsorship must be clear to the reader, and paid content cannot pose as editorial. ⁶

The second covers the preview. Google does not allow a preview that promises details the content does not deliver. ⁶

The third covers transparency, and it is the one that speaks to any business. News sources should show clear dates and bylines. ⁶ Also information about the authors, the publication, and the publisher, about the company behind the content, and contact information. ⁶

The penalty. Google may remove the content from Discover. ⁶ In repeated or egregious cases, the site may lose eligibility for Discover, including the Chrome surface. ⁶ Manual actions can show up in Search Console, under Security & Manual Actions. ^{2 6}

Following sites in Discover: what exists today

Here the documentation tells two stories, and they are easier to read apart.

The old button. On September 1, 2021, Google documented a feature for following sites, still in beta. ¹⁵ It relied on the site’s RSS feed. ¹⁵ On November 19, 2025, Google deleted that guidance. The reason, in the changelog’s own words: “The Follow feature is no longer shown in Google Discover.” ¹⁵

Following now. On September 17, 2025, Google announced a different kind of follow. People follow publishers and creators right on Discover, signed in to their Google Account. ⁸ Before following, they can preview what that source publishes. ⁸

The policies page has a section on this. People who follow your site see your content in the “Following” feed. ⁶ The exception is content that violates the overall policies or a list of policies for Search features. ⁶

On June 4, 2026, Search profiles arrived. ¹⁶ Publishers and creators with a sizable following on a social or video platform can claim their own profile. ¹⁶ People who follow a source from its profile are more likely to see its content in Discover. ¹⁶ Google wrote that the profiles would launch first in the US. ¹⁶ If your readers are in the US, they may meet this feature before readers elsewhere.

In the site owner documentation I read on September 14, 2026, no markup or file turns on the follow button. Today, the one who presses it is the reader.

How to measure Discover in Search Console

The Discover performance report. Search Console got this report on April 10, 2019. ¹⁷ It appears only when the property passes a minimum number of Discover impressions. ¹¹

The help page defines the metrics. An impression is a card scrolled into view, counted once per result per session. ¹¹ A click is a tap on the card, and a share does not count as a click. ¹¹ Click-through rate divides one by the other. ¹¹

You can group by page, country, appearance type, and date. ¹¹ The report covers the last 16 months. ² The numbers are credited to each page’s canonical URL. ¹¹

What it does not measure. The Discover report has no queries and no positions. ¹⁸ The Search Console team wrote that in 2021, when it opened this data in the API. ¹⁸ The help page’s list of groupings has no device either. ¹¹ And the table stops at 1,000 rows. ¹¹

The Discover AI report. Since August 31, 2026, a separate report for Discover’s generative AI features has been open to all sites. ¹⁹ It appears only for sites with enough impressions in those features. ¹⁹ It counts impressions, by page, country, and date. ¹⁹ The June 2026 announcement keeps the device breakdown for Search only. ²⁰ And it says this data also feeds the overall report. ²⁰

The Search Console control that takes a site out of generative AI features also covers the ones in Discover. ²¹ What that control costs is in the piece on AI Overviews.

Social posts. Since July 29, 2026, platform properties have been open to everyone. ²² With them, Search Console shows how posts on Instagram, TikTok, X, and YouTube perform in Search, Discover, and Google News. ²²

Outside Search Console. In 2021, Google announced that Discover in Chrome would arrive with the referrer https://www.google.com/. ⁵ It is the same one Discover uses in the Google app. ⁵ To a tool that reads where a visit came from, Discover shows up with Google’s address. That is why I read Discover volume in Search Console. How Analytics splits its channels is in the piece on organic traffic.

Why Discover traffic rises and falls

Google itself writes that Discover traffic is less predictable than keyword-driven search traffic. ² And it recommends treating it as a supplement to that traffic. ²

The guide lists three reasons. ² People’s interests change, and the feed follows. The types of content in the feed change. And Search updates move Discover, which is an extension of Search. ²

There is one sentence every site owner should read before changing everything. Sometimes there is nothing to do after an update. ² Google’s ongoing work on Discover can shift traffic for reasons unrelated to quality or publishing frequency. ²

The February 2026 update. On February 5, 2026, John Mueller announced a core update built for Discover. ²³ It had three goals. More local content, from sites based in the reader’s country. Less sensationalism and clickbait. And more in-depth, original, and timely content, from sites with expertise in the topic. ²³

The system identifies expertise topic by topic. ²³ Google’s example helps. A local news site with a gardening section may have gardening expertise. A movie review site with a single gardening article probably does not. ²³

The update started with English-language users in the US. ^{23 24} The status dashboard records the end on February 27, 2026. ²⁴ If your readers use Google in English in the US, this is the update that reached them. Google promised to bring it to all countries and languages in the months that followed. ²³ On the dashboard as read on September 14, 2026, there is no record of that expansion. ²⁴

On the same day, February 5, the “Get on Discover” guide got the new recommendations. ¹⁵ I know of no measurement of this update’s effect in Brazil, where we are based, with a published method.

What Discover is not

Discover is not search. In Discover, nobody types a query. ¹¹ That is why there is no Discover keyword to target, and why the report shows no queries. ¹⁸

There is no sign-up either. No form and no entry tag. ² An indexed page that meets the policies is already eligible. ² And the traffic from there comes with no guarantee. The guide itself says to treat it as a supplement. ²

One search rule applies here, too. The robots specification lists Discover among the results affected by nosnippet and max-snippet. ¹³ What decides the text snippet in regular search is in the piece on meta description.

What to do starting today

From now on, when your Discover traffic drops, you have three questions before the scare. Did the audience’s interest change? Was there an update on the status dashboard? Did the page’s card follow the guide?

The first step takes ten minutes. Open Search Console, go to Performance, and look for Discover. If the report is not there, the site has not passed the impression minimum yet. ¹¹

If it is there, write down the date, the clicks for the last three months, and the five pages with the most impressions. Then open the source code of one of them. Look for max-image-preview:large and check which image sits in og:image. If it is the logo, swap it for the image that represents the text. ²

Last, look at the page as a reader would. Are the date, the author, the contact, and the company visible? ⁶ The site’s other technical groundwork is in the technical SEO section.