How articles actually enter the Discover feed

Most SEO teams treat Discover as "Search with pictures". It isn't. There is no query. There is no position. An article doesn't rank in Discover — it gets admitted, evaluated, amplified or dropped. Understanding that lifecycle is the difference between publishers who get occasional lucky spikes and publishers who ship articles into the feed on demand.

What follows is the entry mechanism as we observe it across the properties we monitor — 10M+ URLs in our index, growing by 10,000 a day.

Step 1 — Source eligibility comes first

Before any single article can enter the feed, the site must be eligible: sufficient E-E-A-T signals, a clean quality-tier standing, valid technical foundations (large images enabled, correct metadata, fast rendering). A perfect article on an ineligible source gets zero impressions. This is why our audits start at the domain level, never at the article level.

Step 2 — The evaluation window

When a fresh URL from an eligible source is picked up, the feed gives it a small test audience — typically within minutes to a few hours of publication. This window is short and unforgiving. The signals that matter:

• Click-through on the card — driven almost entirely by headline & image pairing;
• Engagement response — do the first readers stay, scroll, come back?
• Velocity — an audience that responds within the window beats a bigger audience that responds late.

Step 3 — Amplification tiers

Pass the evaluation and the article climbs through audience tiers: interest cluster → related clusters → broad distribution. Each tier is a re-evaluation. Most articles die between tier one and two — usually because the headline promised more than the content delivered, and the engagement curve sagged.

Step 4 — Decay and the follow-up window

Discover articles decay in 48–96 hours. The mistake is to mourn; the play is to follow up. A URL that performed opens a temporary trust corridor for the same cluster: follow-up angles, internal links from the hot article, and a second push often turn one entry into a week-long traffic event.

What you can't see in GSC

Search Console shows you Discover data with up to 48 hours of delay, aggregated, after the game is over. The evaluation window — the only moment you can act — is invisible in GSC. That blind spot is exactly what real-time Discover tracking exists to close: seeing feed entries as they happen, so the newsroom can respond while the article is still being evaluated.

Once you see the lifecycle, the strategy writes itself: earn eligibility, engineer the spark, feed the amplification, exploit the corridor. Everything we do — consulting, operations, tooling — maps to one of those four moves.