Google Discover breaks the model most SEO is built around: there is no search query. Google decides on its own that a particular person might want your article, and drops it into the feed they see in the Google app, on the Chrome new-tab page, and increasingly on the desktop homepage. For a publisher that lands, Discover can rival or even exceed the traffic from ordinary search. For one that doesn’t, it pays nothing.

That gap is why Discover gets misunderstood. There is no keyword to target, no ranking position to chase, and no switch that turns it on. According to Google’s own documentation, you can only make a page eligible; whether it actually appears is a separate event that depends on one person’s interests at one moment. This is a practitioner’s read of what genuinely matters, prompted by Olga Zarr’s in-depth Discover guide on SEOSLY. The honest version for a small or EU-based publisher comes first: nobody can promise you Discover traffic. What follows is how to earn eligibility and stop getting in your own way.

How does Google decide what shows up in Discover?

Discover is personalization, not retrieval. Google’s systems match content to what an individual user seems to like, drawn from their Search history, their activity across Google products, their location, and the topics and sites they have chosen to follow. Two people opening the feed at the same moment can see almost nothing in common.

That has a practical consequence worth sitting with. Being eligible for Discover (indexed, compliant, technically sound) is necessary but not sufficient. Appearing is a second, separate event that depends on a specific person’s interests at a specific time. It is also why Discover traffic spikes and collapses in ways search traffic rarely does, and why an evergreen piece can resurface months later when it suddenly fits someone’s moment. Don’t read a quiet week as a penalty, or a viral day as a strategy you can repeat on demand.

One shift to plan for: Discover is expanding to desktop. Google confirmed the rollout at Search Central Live in Madrid in 2025, and it is appearing on the google.com homepage and Chrome’s new-tab page for logged-in users — initially in a handful of countries (the US, Canada, the UK, Australia, Germany, per that announcement). Treat it as a direction of travel rather than a switch that has already flipped everywhere.

E-E-A-T: who Google is willing to recommend

Discover leans hard on trust, because Google is staking its own credibility on what it pushes to people who didn’t ask. Experience, Expertise, Authoritativeness and Trust are not a score you set; they are signals Google assembles about who is behind the content.

The boring, fixable parts matter most. Real author bylines, on real author pages, with a genuine bio and a stable author.url. A publisher identity that stays consistent across the site and its structured data. Clear sourcing on anything medical, financial or scientific. None of it is glamorous, and all of it is the difference between a site Google trusts a person to read unprompted and a domain it would rather not vouch for. There is a useful overlap here: the same trust and quality signals that earn Discover eligibility are the ones that increasingly earn citations in AI Overviews and AI search, which is the heart of our argument in “good SEO is good GEO.” Discover work is rarely a single-channel bet.

Content that earns a place in the feed

Discover rewards content people want to read, not content that answers a query they typed. In practice that favours a few recognisable shapes: data-led analysis with a number nobody else has, a contrarian take grounded in real experience, and locale-specific angles a global outlet would never bother with. A generic “Best budget laptops 2026” roundup is exactly what a language model produces on request, so it has little reason to travel. “The budget laptop we returned after a week, and what we bought instead” carries a point of view that can’t be faked, and that is what the feed tends to surface.

Headlines carry more weight in Discover than almost anywhere else, because there is no query to set expectations: the headline is the pitch. A little curiosity is fine; a headline that oversells or hides the actual payoff is the fastest way to teach Google that your content disappoints the people it recommends. Make a promise the article keeps. Freshness helps too — Discover skews toward recently published material — but don’t game it: bumping a date without a real update is the kind of signal Google’s guidance explicitly warns against.

Publisher Center, the Follow button and topic signals

You do not need Google News membership to appear in Discover, but a complete profile in Google Publisher Center reinforces the publisher identity Google reads, and it is worth setting up properly. The Follow button matters more than most realise: when a reader follows your site or a topic you cover, your content becomes eligible to surface in the “Following” view of their feed — a direct line back to an audience that has already raised its hand. Building genuine depth on a topic, rather than scattering one-off posts across unrelated subjects, is what makes those follow and topic signals add up over time.

What image size does Google Discover require?

Google requires Discover images to be at least 1,200 pixels wide, at least 300 KB, in a 16:9 aspect ratio, with more than 300,000 total pixels (a 1280×720 image clears it comfortably). That is the spec, and it is non-negotiable for the large-image treatment that makes Discover work.

The piece that is easy to miss is the meta tag. Large image previews require max-image-preview:large in your robots meta; if it is absent, or set to standard or none, you are not eligible for the large-image treatment no matter how good the picture is. Pair it with a relevant og:image (or schema markup) so Google picks the right one, and avoid the obvious mistakes: your site logo is not a Discover image, and overlaid text, heavy filters or misleading visuals work against you. In the Discover audits we run, the failure is far more often a missing or wrong robots meta tag than the image size itself — the spec can be perfect and still yield no large-image treatment.

The technical foundation

None of the above matters if the page can’t be crawled, indexed and loaded cleanly. The baseline:

  • Indexable and in Google’s index. Discover draws from the same index as Search; an unindexed page can’t appear.
  • Core Web Vitals in the green. Largest Contentful Paint ≤ 2.5s, Interaction to Next Paint ≤ 200ms, Cumulative Layout Shift ≤ 0.1. A feed user who taps through has zero patience for a slow, jumping page.
  • Mobile-first and HTTPS. Discover began as a phone experience; a clean mobile render and a secure connection are table stakes.
  • Structured data. Article, NewsArticle or BlogPosting markup — with datePublished, dateModified, an image array and author.url — helps Google understand and trust the page.
  • A working feed. An RSS or Atom feed, linked with rel="alternate", gives Google a clean signal when you publish.

The policy gate you can lose in one move

Discover eligibility is conditional on Google’s content policies, and this is where sites self-destruct. Mislabelled sponsored content, manipulated media, or dangerous and hateful material can all pull you out of the feed. Two newer risks deserve naming: mass-produced, low-effort AI content (“AI slop”) and site reputation abuse, where a trusted domain rents space to third-party content for the rankings. Google has been visibly tightening on both. The practical posture is conservative — label sponsorships clearly, keep headlines honest, verify sensitive claims, and don’t publish anything you would be uncomfortable defending. For news publishers, where the policy bar and the upside are both highest, that gate is worth a dedicated news site audit before you lean on Discover.

Does Discover traffic show up in Search Console?

Yes — Discover performance lives in its own report in Google Search Console, with up to 16 months of clicks, impressions and CTR. The catch is that the report only appears once you have had enough Discover traffic to populate it. Use it to learn which formats and headline styles actually travel, and watch CTR trends rather than obsessing over any single day. Keep an eye on the Security & Manual Actions section too: if Discover traffic vanishes overnight, a policy issue is the first thing to rule out.

The trap is reading volatility as signal. Discover is spiky by design. Judge it over weeks, against the business outcomes that matter — subscriptions, leads, return visits — not against the shape of a single graph.

What this means for a smaller or EU publisher

Strip away the checklist and Discover comes down to a few honest moves. Build genuine author and publisher trust into the site. Write the pieces only you can write, and headline them honestly. Get your images to Google’s actual spec, since that one technical fix separates sites eligible for the large-image treatment from those that aren’t. Keep the foundation fast, mobile-clean and policy-safe. Then measure over weeks, and resist the urge to read every quiet day as failure.

European and multilingual publishers have two extra levers most US-centric guides skip. Correct hreflang tells Google which language and region version of a page to surface, so a Ukrainian or Polish edition lands in the right national feed instead of competing with itself. And because GDPR limits how much personal data Google can use to personalise for EU users, publisher-side signals — topic authority, follow relationships, clean structured data — tend to carry relatively more weight here than in markets with looser consent norms.

None of this guarantees a place in the feed. It just makes you the kind of site Google can afford to recommend. The eligibility work is also easy to get subtly wrong — a meta tag set to standard, an image cropped a hair off 16:9, a dateModified that quietly contradicts the visible date — which is exactly the kind of thing a structured Google Discover audit is built to catch.

Sources