Discover for beginners: from first signal to traffic spike

You just published your first articles. You know Google Discover exists, you've seen competitors pull 60 to 80 % of their mobile traffic from it, and the entry ticket is free. Yet your Search Console stays painfully empty on the Discover side. That's not a bug — it's a silent validation chain running in several phases, each with its own thresholds. This guide takes that chain apart, from the first signal Google quietly logs to the traffic spike that actually changes a site's trajectory.

Before your first impression: what Google silently observes

Most beginner publishers assume that hitting "publish" immediately triggers a Discover evaluation. It doesn't. Discover doesn't look at articles in isolation — it looks at sites. Before any impression at all, Google runs a purely passive observation phase that typically lasts 14 to 21 days.

During that window, the algorithm collects three families of signals without ever showing you in the feed:

• Publishing cadence. A site that ships 4 articles and then goes silent for three weeks sends an "abandoned" signal. Aim for a minimum of 2 to 3 articles per week across the first 30 days, ideally at consistent times.
• Topical coherence. Discover wants to bind your domain to 1 to 3 main entities (e.g. "mobile tech", "vegetarian cooking", "personal finance news"). If your first 12 articles scatter across 12 different topics, the algorithm has no idea who to recommend you to.
• Technical authority signals. Schema.org Article or NewsArticle in place, the max-image-preview:large meta tag, valid HTTPS, mobile Core Web Vitals in the green. None is enough on its own — but a single one missing can stall the whole pipeline.

This observation phase is non-negotiable. No trick shortens it — not even a site with stellar classic SEO, because Discover runs on a separate evaluation pipeline, as covered in detail in our piece on the 7 fundamental differences between classic SEO and Discover SEO.

Step 0: your Google Web Profile

If you take only one action away from this guide, make it this one: verify that your Google Web Profile exists and is properly populated. It's the internal identity card Google uses to attach your articles to your brand. Without it, your engagement signals fly off into the void.

Concretely, the Google Web Profile holds:

• Your main category (sports outlet, food blog, finance site…)
• Your target country (critical for the feed's language segmentation)
• Your official logo (shown as a thumbnail under every Discover card)
• A description sourced from Wikipedia, the knowledge graph or your Organization markup
• Your verified social links (a secondary trust signal)

To check in 30 seconds whether your profile is indexed, run the Profiler — drop your domain in and it returns the profile.google.com/cp/… URL if it exists, or an eligibility score if it hasn't been generated yet. On brand-new sites, expect to wait 4 to 8 weeks before Google generates that card automatically. Until it exists, you stay in "passive observation" mode and zero Discover impressions will ship.

"Discover treats every domain as an entity first, and a publisher second. If we can't resolve the entity, we don't recommend the content." — paraphrasing public statements from John Mueller (Google Search Central)

The first feed entry: 3 hidden checks

Once observation is done and the Web Profile is validated, Google tests you. That first exposure is deliberately tiny: between 50 and 500 impressions over 24 to 72 hours, distributed to a panel of users selected for topical affinity. Three checks fire in parallel:

1. The relevance check (user ↔ entity matching)

Are your articles being served to the right profiles? If Google thinks your site covers vegetarian cooking but the panel users have shown zero historical interest in the topic, the signal you produce is bad — not because of your content, but because the entity was misclassified. That's exactly what happens to multi-topic sites ("general lifestyle"): Discover refuses to promote them as long as their identity stays fuzzy.

2. The instant engagement check (thumbnail CTR)

Are your hero image and headline clickable enough? The bar sits at 4 to 6 % CTR on average across the Discover feed — substantially higher than classic SERPs. Below 2 %, the article is pulled from the test fast. Above 7 %, the panel automatically widens.

3. The post-click quality check (dwell time + back button)

Once you have the click, Google watches what happens next. A return to Discover in under 10 seconds (pogo-sticking) is the single worst signal — it tells the system your headline over-promised. On the flip side, dwell time over 40 seconds plus meaningful scroll triggers a boost into the next phase.

Spike #1 vs spike #2: why the second matters more

When the first real Discover spike lands — typically 800 to 5,000 visitors over 48 hours on an article that clears all three checks — most beginner publishers assume they've made it. That's a misread. Spike #1 only proves you nailed one article. Discover wants to know whether you can nail a second one.

The window is short: you have roughly 10 to 14 days after spike #1 to ship a follow-up that performs at least at 60 % of the first one. Pull it off and the algorithm reclassifies your site into a higher cohort and the exposure frequency doubles. Miss it and the comedown is brutal — typically 3 to 5 weeks of silence before another test slot opens.

That's exactly the pattern we documented in our +215 % Discover traffic in 30 days case study: the trajectory wasn't changed by the viral piece — it was changed by the ability to ship three solid articles in two weeks. Spike #2 and spike #3 are the real Discover health indicators.

The 3 signals your readers send without knowing it

Discover doesn't ask for your opinion on your own content. It watches what your readers actually do. Three behavioral signals carry disproportionate weight in the ranking:

1. Deep scroll (depth signal)

Hitting 75 % of article height is a strong signal. That implies your article has real length (1,200 words minimum) and a pacing that holds the reader. Short pages rarely "stick" in Discover long-term, even with a strong initial CTR.

2. Active exit (clicks on internal or external links)

A reader who clicks a link after reading sends a far stronger signal than a reader who closes the tab. That's why internal linking — out to other articles, to the Profiler, to a tool page — is a deeply underrated lever.

3. Direct return (non-Discover revisit)

If your reader comes back to your site within 72 hours through a direct channel (typed URL, bookmark, branded search), that's the strongest possible signal. It tells Google you built a brand relationship, not just a moment of caught attention. This signal explains why sites with an active newsletter or a mobile app accelerate faster on Discover.

The 5 mistakes that cap beginners

1. Publishing too little, too irregularly. One article per week works on average; one every 4 days works better; one every 12 days on average is below the consistency threshold.
2. Diluting the topic. Mixing tech, travel, finance and lifestyle on one domain stops Google from classifying you. Better to run one site on one topic than one site on ten.
3. Ignoring the hero image. A thumbnail under 1,200 px wide, or a soft JPEG, or cropped wrong — that's the #1 elimination factor during the test phase. Run through our 7 techniques to blow up your Discover traffic as a checklist.
4. Writing ambiguous headlines. Discover rewards clarity, not mystery. ✓ "The 5 mistakes that kill your Google Discover" / × "You'll never guess what's slowing your site down…".
5. Optimizing for keywords instead of topics. Discover doesn't run on queries. It runs on interests. An article titled "iPhone 17 battery life autonomy test" is a keyword article. An article titled "what the iPhone 17's battery life actually changes for heavy users" is a topic article — and that's what Discover rewards.

Conclusion + quick diagnostic

The beginner's Discover path isn't measured in days, it's measured in validation cycles. Observation phase (14-21 days), Web Profile creation (4-8 weeks), first test (50-500 impressions), spike #1, critical 10-14 day window, spike #2 confirming the trajectory. Each step has its own thresholds, and each one knocks out 60 to 80 % of the sites that quit halfway through.

Three concrete actions for next week:

1. Check your Web Profile with the Profiler. If missing, ship schema.org Organization + Wikipedia (if eligible) and wait it out.
2. Audit topical coherence: across your last 20 articles, how many distinct topics? Above 4, you're diluting the entity.
3. Measure your cadence: open your archive and compute the average interval between publishes. Under 1 article every 4 days, raise the cadence before changing anything else.

For more on the underlying mechanics, read our French version of this article — Discover pour débutants : du 1er signal au pic de trafic — and our complete guide to how Google Discover works in 2026.

Frequently asked questions