Most Discover advice is Search advice with the word "Discover" pasted on top. It tells you to target keywords, build topical authority, and win intent. But Discover has no query. Nobody typed anything. There is no intent to match, no SERP to compete on, and no keyword whose ranking you can check.
That single absence changes the whole optimization problem. In Search, the query is the constraint and your page is the candidate. In Discover, the user is the constraint — their inferred interests — and your page is the candidate. You are not competing for a phrase. You are competing for a slot in a feed assembled around someone's topical profile, and the only thing they see before deciding is a title and an image.
Why "no query" changes the ranking problem
Google documents Discover as showing content based on a user's interests, drawn from their Web & App Activity and from topics they explicitly follow in the Google app. Location and language matter too. There is no equivalent of query-document matching because there is no query document.
The practical consequence: keyword targeting has no lever to pull. You cannot rank for "best hiking boots" in Discover. You can publish something about hiking gear that the system associates with hiking-interested users at a moment when that association is active. Those are different jobs.
| Dimension | Search | Discover |
|---|---|---|
| Trigger | A user's query | A user's inferred interests |
| Matching unit | Query ↔ page/passage | Topic/entity profile ↔ page |
| Competitive set | Other results for that query | Everything eligible for that user's feed |
| Freshness weight | Varies by query intent | Heavily skewed toward recent content |
| Preview surface | Title, URL, snippet | Title and a large image |
| Available data | Queries, pages, positions | Pages only — no query dimension |
The last row is the one that quietly breaks most reporting workflows. Search Console's Discover report gives you pages, impressions, clicks, and CTR. It gives you no queries, because there are none to give.
What Google actually documents about eligibility
This is worth separating from inference, because a lot of Discover folklore is inference presented as fact.
Documented by Google:
- Eligibility is automatic for pages indexed by Google that meet Discover's content policies. There is no submission step, no special markup, no AMP requirement.
- Large images in the feed require either
max-image-preview:largein the robots meta tag or AMP. Google recommends images at least 1,200 pixels wide. - Discover has its own content policies layered on top of Search's spam policies, and violating them can remove content from Discover without affecting Search.
- Google's own guidance warns against clickbait, curiosity-gap titles that withhold information, and using a site logo as the page image.
Reasonable inference, not documented mechanism: that Discover leans on entity and topic representations of your site as a whole, that sites with a coherent topical territory get surfaced more consistently, and that user-level engagement feedback shapes which publishers keep appearing. These are consistent with how the feed behaves, but Google has not published the model.
Here is the only piece of markup that is genuinely non-negotiable:
<meta name="robots" content="max-image-preview:large">Without it, you can still appear, but with a thumbnail instead of the full-bleed image. On a surface where the image is most of the decision, that is a large handicap for one line of HTML.
What interest matching implies for what you publish
If the system is matching topics to profiles rather than pages to queries, three things follow.
Coherence beats coverage. In Search you can rank a one-off article on an unrelated topic because the query matched it. In Discover, a page that sits outside your site's usual subject matter has no interest profile to attach to. Sites that publish across an identifiable territory tend to surface more predictably than sites that publish scattershot. Treat this as a strong pattern, not a documented rule.
Recency is a gate, not a tiebreaker. Discover traffic concentrates overwhelmingly on recently published content. Evergreen pages do appear, usually when something in the world makes the topic newly interesting. If your content strategy is entirely evergreen, Discover is not a channel you can optimize into — it is a channel you occasionally get lucky in.
There is no long tail to farm. Search rewards publishing a thousand narrow pages that each catch a trickle. Discover does not accumulate that way. A page either enters feeds or it does not, and the distribution is brutally uneven: a handful of pages will account for most of your Discover clicks.
Why the image is doing the work your title tag does in Search
In a SERP, a user has already declared intent. Your snippet only has to prove relevance. In Discover, the card has to create interest from nothing. The image carries most of that load because it is the largest element and the fastest to parse.
What that means in practice:
- Use a distinct, high-resolution image per article, at least 1,200px wide, not a stock photo you have used forty times. Repeated imagery reads as generic in a feed.
- Crop for a wide card, not for your article layout. Feed cards crop aggressively; a subject centered for a 4:3 in-body image often loses its head.
- Avoid text overlays and logos. Text baked into the image is unreadable at card scale and competes with the title. Google explicitly advises against using your logo as the preview image.
- Make sure the image you want is the one Google picks. Set it as the
og:imageand as theimageproperty in your Article structured data, and keep it high in the DOM.
How to write titles when the title is the entire result
A Discover title has no query context to lean on, so it must be self-sufficient: a reader with no prior context should understand what the story is. That is the opposite of the curiosity-gap headline, and Google's Discover guidance specifically calls out withholding key information as a policy problem, not just a taste problem.
So: state the thing. "Why the new EU battery rules will raise e-bike prices" works. "You won't believe what the EU just did to e-bikes" is a policy risk with worse retention. Note that Google may use your <h1>, <title>, or other on-page text for the card, so keep them consistent rather than optimizing one and neglecting the others.
How to read the Discover report without fooling yourself
The data has three properties that break naive analysis.
- No query dimension. Do not try to reconstruct one. Analyze at the page and topic level.
- Extreme skew. A single article can produce more impressions in two days than your site gets in a month. Means are useless here; look at medians, distributions, and per-article outcomes.
- CTR is not comparable to Search. Discover impressions are cheap and passive. A Discover CTR of a few percent is a different animal from a Search CTR of a few percent, and mixing them in one dashboard row is a reporting bug.
Also remember the 16-month retention window applies here too. If you want longer Discover history, export it on a schedule.
What to do this week
- Confirm
max-image-preview:largeis present sitewide, including on templates that override the robots meta tag. - Audit the last 20 published articles for a unique, wide-cropped, 1,200px-plus lead image with no baked-in text, and check that
og:imageand Articleimageagree. - Pull the Discover report, sort by clicks, and identify what the top pages have in common — topic, format, publish timing. That is your interest signal, since you have no queries.
- Rewrite any curiosity-gap headlines on high-impression pages into complete-thought titles, then compare the next cohort's CTR.
- Decide explicitly whether Discover is a channel you invest in. If you do not publish timely content in a coherent subject area, the honest answer is that your effort belongs in Search instead — and do not forecast revenue against Discover volume, which can disappear without a policy violation or an algorithm announcement.