How to Get Google Discover Traffic to a New Site (2026)

Google Discover is the personalized content feed on the Google app, on Chrome's new-tab page on Android, and in the Google home-screen widget. It is not a search result — users aren't typing anything. Google decides, based on each person's interests, what stories they'll find compelling and surfaces them passively. For US readers, this is the feed they scroll while waiting in line, on the couch, or on the toilet. Traffic from it behaves differently from search traffic: visits are short, click-through rates are high, and the bounce rate is normal for browse intent. When I started writing on the side outside my old company, the first Discover spike made me assume something was broken — 30,000 sessions in a day from a feed I didn't even know existed. That was the moment I realized Discover wasn't optional; for content sites in 2026, it's a major channel. The thing that makes it different is that you don't optimize for keywords. You optimize for being clickable, fresh, and useful in the context of someone scrolling for curiosity or entertainment. Google's own publisher guidance for Discover is the one authoritative source worth bookmarking: https://developers.google.com/search/docs/appearance/google-discover . For how new sites get any traffic at all, see how to get traffic to a new website.

Google has been deliberately vague, but cross-referencing its public guidelines with what actually surfaces in 2026, four signals matter most. They aren't equally easy to fix, and a new site usually has to earn them in roughly this order:

• Page experience and Core Web Vitals. Sites that load fast, don't shift layout, and feel native on mobile get a large Discover advantage over slow or janky competitors. This is the most controllable signal and the one to fix first.
• High-quality images, ideally over 1,200px wide with proper Open Graph tags. Discover is a visual feed; small or missing images kill your odds before the headline even gets read.
• Clear, non-clickbait headlines that promise something specific. Discover punishes misleading headlines hard by reducing future surfacing, so the headline has to be both compelling and deliverable.
• Topical authority and freshness. The algorithm is trying to predict whether a specific user will find this story interesting right now, and freshness is a strong proxy. Stories more than a few weeks old rarely get Discover traffic unless they're evergreen reference content with sustained user signals.

The implication for a new site is concrete: get your technical performance right, invest in real images rather than stock, write headlines like a magazine editor, and publish consistently in a tight topical area. Core Web Vitals explained goes deep on the performance side.

Three technical mistakes effectively block new sites from Discover, and all three are fixable in an afternoon. The first is missing or incorrect Open Graph image tags: Discover needs a clear, large hero image to surface a story, and if your og:image is missing, broken, or under 1,200px wide, you're invisible to the feed. Test every URL with a sharing debugger or Open Graph preview tool before assuming this is right. The second is lazy-loaded above-the-fold images: some templates lazy-load every image including the hero, which means Google's crawler sometimes doesn't see them or sees them late. Set the hero to load eagerly with explicit width and height attributes. The third is weak structured data: Discover doesn't strictly require Article schema, but pages with proper Article structured data including datePublished, author, and image consistently outperform pages without. The bar for Discover is higher than the bar for regular search results — treat your technical setup like you're publishing to a curated magazine, not dumping HTML on the internet. For the schema patterns, how to write SEO content with AI covers the full structured-data stack.

Across the content sites I've watched succeed in Discover, the winning angles fall into recognizable patterns, and they all share one trait: voice. The first is human-interest framing on practical topics — a guide titled "How to Save $200 a Month on Groceries" gets less traction than "I Tried 3 Grocery-Saving Hacks for a Month — Here's What Actually Worked," because the first sounds like generic SEO content and the second sounds like a magazine story. The second is timely interpretation rather than pure news — Discover surfaces analysis more reliably than breaking news, so "What the new Apple Tax change means for indie devs in 2026" beats "Apple announces new policy." The third is lists with strong personality, like "5 things every US freelancer should do before April 15," which works because it's specific, time-bound, and signals authority. The fourth is counterintuitive framing — "Why I stopped using ChatGPT for client work" outperforms "ChatGPT pros and cons" because it has a clear point of view.

Headlines carry the whole load in Discover because users see only the headline and the image. The patterns that work without crossing into clickbait are leading with a number when the number is genuinely surprising, framing as a question a real person actually asks, and using first person where appropriate. What fails is vague curiosity hooks like "You won't believe what happened next," hyperbolic adjectives, and listicle headlines with no real angle. Google now penalizes manipulative headlines by cutting future impressions, so clickbait isn't just ineffective — it's actively damaging. A useful gut check: would a real magazine editor approve this headline, or redline it as too thirsty? The companion rule is to write the headline as if you must deliver on it within the first 100 words, because if the headline implies a story you don't actually tell, the bounce rate signals Google to stop surfacing you. Trending keywords strategy covers how to find these topics before everyone else does.

Images do more work in Discover than text — the hero image gets the click and the headline closes it. Stock photos rarely surface well, because Google has gotten good at recognizing common stock images and de-prioritizing them. The images that work are original photography you took yourself, screenshots of actual interfaces or processes you're describing, custom illustrations or designed graphics, and edited charts or diagrams. The minimum size that consistently surfaces is 1,200px wide, and the sweet spot is 1,920 by 1,080 or larger at a 16:9 ratio that crops cleanly on mobile. If you're a single-operator site without a photographer, invest in two things: a paid stock service that doesn't appear in everyone else's content (Unsplash+ or 500px rather than the free Unsplash everyone uses), and an AI image tool for custom illustrations where stock won't fit. Be careful with AI images, though — Google has been cracking down on obviously AI-generated hero images on news-style content, so keep AI for diagrams and abstract illustrations rather than photorealistic stand-ins. Best AI image tools 2026 covers what's working without triggering quality penalties.

Discover rewards consistency more than volume, which is good news if you're building income around a day job. A site publishing three strong pieces a week consistently outperforms a site publishing ten mediocre ones, and both beat a site publishing once every two weeks. The minimum cadence I've seen work for real Discover traffic is roughly two to three publishes per week; below that, your topical-authority signal is too weak for Google to reliably predict you have fresh content worth surfacing, and above five per week, quality usually slips. The other thing Discover rewards is tight topical clustering. A site publishing three pieces a week on a coherent theme — say, indie iOS development — beats a site publishing three pieces a week on unrelated topics, because topical authority compounds in Discover when Google's prediction confidence rises around a clear site identity. The practical comparison looks like this:

| Cadence | Topical focus | Discover outcome | | --- | --- | --- | | 1 piece every 2 weeks | Any | Too sparse; weak freshness signal | | 2-3 pieces/week | Tight cluster | Best balance; authority compounds | | 3 pieces/week | Scattered topics | Surfaces inconsistently | | 5+ pieces/week | Any | Quality slips; diminishing returns |

For a new site, this means picking a niche narrow enough to publish in consistently for twelve months without running out of angles. See how to pick a niche for your website for the selection framework.

Google Search Console has a dedicated Discover report under the Performance tab, separate from Search — open it weekly. The metrics that matter are total impressions (how often your content surfaced), CTR from feed to site, and average position (less meaningful in Discover than in Search, but still worth watching). When a piece gets unusual traction, study it: what was the headline pattern, the image style, the publish timing, the angle? Document the patterns and apply them to future pieces. The single most useful metric for refining strategy is per-piece CTR. A piece that surfaces a lot but isn't clicked has a weak headline or a weak image — fix that pattern next time. A piece with high CTR but few impressions means your headline and image are working but Google lacks confidence in your topical authority, which is a signal to keep publishing in the same theme. Treat Discover analytics as a feedback loop rather than a scoreboard: every spike and every flop tells you something specific about which headline shapes, image styles, and topics your audience responds to. Internal linking strategy 2026 covers how internal linking compounds your topical authority over time.