Google Discover in 2026 - How to Earn Feed Traffic When Search Sends Fewer Clicks
RETURN_TO_BLOG
AI & SEO 12 min

Google Discover in 2026 - How to Earn Feed Traffic When Search Sends Fewer Clicks

Paweł Wiszniewski
Paweł Wiszniewski
SEO & GEO Specialist · AI Engineer

As clicks from search fall because of AI Overviews, many companies look for traffic that doesn't depend on a typed query. Google Discover, the recommendation feed in the Google app and on the Google homepage on phones, is a natural candidate: it shows content to people before they start searching for anything. For publishers it has long been one of the biggest sources of Google traffic. It isn't immune to change, though - according to data compiled by Press Gazette, in 2025 Discover referrals to US publishers fell 29%, and classic search referrals 38%. Discover won't replace search, but run well it can partly offset search declines.

In 2025 Discover referrals to US publishers fell 29%, less than search (38%), and in February 2026 Google ran its first Discover-only update. How the feed works, what changed in 2026, technical requirements, content that gets shown, titles without clickbait and measurement in Search Console.

In 2026 the rules of the game changed noticeably. In February Google ran the first-ever algorithm update aimed solely at Discover, and the feed gained AI-generated summaries, social media posts and a "Follow" button. In this post I explain how Discover works, what has changed in recent months, which technical requirements you need to meet, which content stands a chance, what to avoid and how to measure results in Search Console.

/// WHAT CHANGED IN DISCOVER

Sep 2025
FOLLOWING SOURCES
A "Follow" button for publishers and creators, posts from X, Instagram and YouTube Shorts
Dec 2025
AI SUMMARIES IN THE FEED
51% of positions in the US, Brazil and Mexico (Marfeel); 21.6% in slots 1-5
Feb 2026
FIRST DISCOVER-ONLY UPDATE
More local and expert content, less clickbait; starting in the US
Apr 2026
PREFERRED SOURCES IN ALL LANGUAGES
Favorite sites appear more in Top Stories and Discover
Jun 2026
AI FEATURES OPT-OUT
A Search Console setting after the CMA decision; tested in the UK
Sep 2026
DESKTOP VERSION ENDS
Reportedly pulled from google.com on desktop

How Google Discover works

Discover doesn't answer queries; it predicts interests. The system combines a user's activity across Google services (if they've allowed it) with topics recognized in the Knowledge Graph, and picks content they may find interesting. That means you're not optimizing for a keyword, but for a topic someone follows and for the moment that topic is alive.

A few principles follow directly from Google's documentation:

  • There's nothing to submit and no special markup to add. Any indexed page that meets Google's content policies can appear in Discover.
  • Freshness matters, but it isn't a requirement. Discover mostly shows new content, but from time to time also evergreen material if it matches interests.
  • Traffic is less predictable than from search. Google explicitly calls it supplemental, because it depends on audiences' shifting interests.

Discover is available in the Google app, on the Google homepage in mobile browsers and on the home screen of many Android phones. Google also tested it on desktop, on the google.com homepage, but according to reports from September 2026 it pulled it from every market where it had been available. When planning content for Discover, still think phone first.

What changed in 2025 and 2026

The past year and a half has brought more change than the several years before it combined.

  • Following sources (September 2025). Signed-in users can follow publishers and creators directly in Discover, and the feed also shows their posts from X, Instagram and YouTube Shorts. Every follow is a lasting signal that can stabilize traffic.
  • Preferred Sources. Launched in 2025, it lets people pick favorite sites that appear more often in Top Stories and in the Discover feed. In April 2026 Google made it available in all languages, and in August it gave publishers a button to embed on their own sites.
  • The February Discover update (February 2026). The first algorithm update aimed only at Discover: more local content from sites based in the user's country, less sensationalism and clickbait, more in-depth, original material from sites with expertise in the topic. It first covered English-language users in the US; Google announced it would expand to all countries and languages in the following months.
  • AI summaries in the feed. In the US, Brazil and Mexico, part of the feed is taken up by AI-generated summaries. According to Marfeel data from December 2025, they made up 51% of positions, though only 21.6% of the top five. That's one analytics vendor's data, but the direction is clear.
  • An opt-out from AI features. Since June 2026, following a decision by the UK regulator CMA, Search Console is getting a setting to exclude a site's content from AI Overviews, AI Mode and Discover's AI features. For now it's being tested with a subset of UK publishers, and Google says it will roll out globally.

Technical requirements

The technical side of Discover is simple, but neglecting one element can shut a page out of large feed cards entirely.

/// DISCOVER TECHNICAL CHECKLIST

✓
Image at least 1,200 px wide
At least 300K pixels, ideally 16:9
✓
max-image-preview:large
Without it, only a small thumbnail in the feed
✓
Image in structured data and og:image
Not a logo - an image showing the topic
✓
Indexed and fast on a phone
Core Web Vitals in good shape
✓
A title without clickbait
Says what the piece is about, no exaggeration
✓
Compliance with content policies
Stricter than in search
  • A large image. Google recommends images at least 1,200 pixels wide and at least 300,000 pixels in total, ideally in a 16:9 ratio. Large image cards generate noticeably more visits than small thumbnails.
  • Permission for a large preview. A robots meta tag with the max-image-preview:large directive allows Google to show a large image. Without it, a page may only reach the feed with a small thumbnail.
  • Specifying the image. Point to the main image in structured data (e.g. Article with the image property) and in the og:image tag. Don't use your logo as the main image.
  • Indexing and speed. The page must be indexed, and on a phone it has to load quickly and comfortably - how to measure that I cover in the post on Core Web Vitals.
  • Compliance with content policies. Discover has stricter rules than search: no misleading titles, no shocking content and no dangerous material.
discover-head.html
<meta name="robots" content="max-image-preview:large"><meta property="og:image" content="https://yourdomain.com/image-1200x675.jpg">

Both tags go in the page's head section. How to describe and prepare the image itself so it also works in visual search I show in the post on image SEO.

Which content stands a chance in Discover

The February update shifted the weight toward quality and expertise. Google put it plainly: more in-depth, original and timely content from sites that know their subject. A site that writes regularly about one field has better odds than a site that writes about everything.

In practice, these do well in Discover:

  • New information from your industry - changes in regulation, launches, research results - covered quickly and accurately.
  • Expert analysis and commentary that adds something the press release doesn't.
  • Guides that match current interests, e.g. seasonal ones or ones tied to events.
  • Material with your own data, photos and examples, not rewritten content from elsewhere.

Discover strongly rewards trust in the author and the site. A clearly credited author, information about the editorial team and a consistent specialty are signals I covered in more depth in the post on E-E-A-T. It isn't a channel for everyone: a site without regular content production - say, a company site with a few service pages - usually won't build a steady presence in Discover.

Titles and images without clickbait

Discover is a feed, so the title and image decide the click. That tempts people to overdo it. Since February 2026 Google has placed its clickbait warning at the very top of the Discover guidelines and names two practices to avoid: exaggerated or misleading details in the title, snippet or image, and withholding information needed to understand what the piece is about.

Instead ofBetter
"You won't believe what's changing in taxes""New VAT rules - what changes for businesses and when"
"This mistake costs companies millions""Invoice errors in e-invoicing: the most common mistakes and how to avoid them"
An image unrelated to the content, chosen for emotionA photo or graphic that actually shows the article's topic

Clickbait can bring a short spike in traffic, but Discover learns from audience behavior - if people quickly return to the feed or hide your content, later pieces may get less reach.

Why Discover traffic swings so much

Anyone with Discover traffic knows the pattern: a day with thousands of visits, then weeks of quiet. That's normal - Discover shows content for as long as the topic is alive in audiences' interests, then lets it fade. Don't judge Discover by one day or one article, but by the longer-term trend and the number of pieces that appear at all.

Two things help stabilize traffic. The first is consistency in one topic - the more pieces in a given field, the more chances the system has to match them to someone's interests. The second is building a direct relationship with your audience: encouraging people to follow the site in Discover and add it to their preferred sources. It's the same direction I described in the post on a zero-click era strategy - less dependence on one algorithm, more on a loyal audience.

How to measure Discover traffic

Search Console has a separate performance report for Discover. It only appears once a site reaches a minimum number of impressions in the feed. It shows clicks, impressions and CTR, and can be grouped by page, country, appearance type and day.

/// WHAT TO MEASURE IN DISCOVER

Pages with impressions
how broadly Discover surfaces the site, not just single hits
Search Console, monthly
CTR per page
low CTR with many impressions usually means a weak image or title
Search Console
Topics with traffic
which fields Discover associates with your site
Search Console, grouped by page
Engagement
engagement time of Discover-driven pages vs the rest
GA4
  • Number of pages with impressions per month - shows how broadly Discover surfaces your site, not just whether you had one hit.
  • CTR per page - a low CTR with many impressions usually means a weak image or a title that doesn't draw people in.
  • Topics of the top-traffic pages - show which fields Discover associates with your site.
  • Engagement in GA4 - Discover traffic usually lands in GA4 as Google / organic, and some visits from the Google app show up as direct traffic; it helps to tag the pages that got Discover traffic in Search Console and compare their engagement time with the rest.

Remember that Search Console data is attributed to the canonical URL. I cover combining Search Console and GA4 data in the AI era in more depth in the post on measuring AI traffic.

Discover and AI-powered search are two different channels that run side by side. In search you compete to be cited in AI Overviews and AI Mode - how to do that I described in the post on AI Overviews and AI Mode. In Discover you compete for a place in a feed that itself increasingly contains AI summaries. The common denominator of both channels is the same set of signals: original, timely content from a site with clear expertise. That's why you don't build a separate strategy for Discover; you add three elements to your content strategy: publishing pace, good images and titles without exaggeration.

A step-by-step implementation plan

  1. 1.Check whether you have a Discover report in Search Console and which pages have already appeared there.
  2. 2.Add max-image-preview:large to your page template if it's missing.
  3. 3.Take care of images at least 1,200 pixels wide, in 16:9, specified in structured data and og:image.
  4. 4.Pick 1-3 topics you want to be known for and publish in them regularly.
  5. 5.React to new events in your industry quickly but accurately - with your own commentary or data.
  6. 6.Rewrite titles so they say what the piece is about, without exaggeration or withholding information.
  7. 7.Strengthen trust signals: a credited author, an about-the-team page, a consistent specialty.
  8. 8.Encourage people to follow the site in Discover and add it to their preferred sources.
  9. 9.Check monthly the number of pages with impressions, CTR and the top-traffic topics.

---

I help companies and publishers build Google traffic beyond classic search results - from technical setup for Discover, through topic and title strategy, to measurement in Search Console and GA4. I do this as part of SEO content marketing and AI optimization (GEO). Get in touch - I'll start with a review of your Discover report and a list of the pages with the most potential.

Worth reading next:

/// RELATED_RECORDS

AI & SEO

Brand Reputation in AI - What to Do When ChatGPT Says Something Untrue About Your Company

In a July 2026 Searchable study, 93% of 1,704 brands had at least one fact wrong or missing in AI answers, and small businesses got fabrications far more often. Where models get untrue facts about a brand, how to audit them, five correction paths and what the law says when an error costs a contract.

13 min
AI & SEO

Image SEO in the AI Era - Google Lens, Visual Search and Graphics Models Cite

Google Lens now processes close to 20 billion visual searches a month, and Circle to Search runs on over 580 million Android devices - images stopped being an afterthought to text and became a search surface of their own, with its own rules. Here's what alt text actually does (and doesn't do), how to wire images into ImageObject structured data, why an infographic built as HTML beats an infographic-as-image for citability, which file format to pick in 2026, and what the EU AI Act says about labeling AI-generated graphics.

14 min
AI & SEO

Bing, Microsoft Copilot and IndexNow - the Forgotten Ecosystem Feeding AI Answers

On July 1, 2026, Fabrice Canel, Bing's longtime head of crawling and indexing and the chief architect of the IndexNow protocol, retired from Microsoft. It's a good moment to notice how much of today's AI-generated answers actually run on the infrastructure he built. The whole SEO world is debating Google AI Overviews, while Microsoft Copilot entirely, and ChatGPT substantially, pull sources from the Bing index - and since February 2026, Bing itself shows you when you're being cited. Here's how the ecosystem works, how to configure it, and how much traffic it actually delivers.

13 min
/// AUTHOR
Paweł Wiszniewski – AI & Web Engineer

Paweł Wiszniewski

SEO & GEO Specialist & AI Engineer

SEO/GEO specialist (10 years) and AI engineer (3 years). I build search visibility, AI systems and automations that reduce costs and improve operational efficiency.

Signal received?

Terminate
Silence

Initiate protocol. Establish connection. Let's build something loud.

> WAITING_FOR_INPUT...