Google Discover surfaces stories from any site that is indexed by Google Search and complies with Discover's content policies — no special markup, tags, or separate optimization track required, according to Google's own documentation. What actually gets shown to a given reader is personalized to that reader's activity, which is why the same headline can drive a traffic spike on one site and nothing at all on another.
What makes a story eligible for Discover in the first place?
Eligibility has two parts, per Google's Search Central documentation: the page must be indexed by Google Search, and it must comply with Discover's content policies. Google is explicit that no special tags or structured data are required to appear — Discover pulls from the same index as Search rather than running a separate submission process.
The policy layer does the real filtering. Discover excludes dangerous content, deceptive practices, harassing or violent extremist material, manipulated media, medical content, sexually explicit or violent content, and vulgar language, per Google's content-policy page. Publishers are also required to show clear dates and bylines and to disclose sponsored content so it isn't mistaken for independent journalism. Google added a site reputation abuse policy effective May 5, 2024, aimed at manipulative publishing practices; violations can get individual stories pulled from the feed or, for repeated breaches, make a whole site ineligible for Discover and for Chrome's "Articles for you" panel.
What signals decide whether an eligible story actually gets shown?
Once a story clears the eligibility bar, Discover ranks it using "many of the same signals and systems used by Search to determine what is helpful, people-first content," per Google's documentation — not a distinct Discover algorithm. The factors it names are relevance to a reader's interests, drawn from that reader's Web & App Activity; page experience quality; image quality and specifications; and timeliness or storytelling value. Recency isn't mandatory: Google says older content can still surface if it's relevant to what a reader is currently interested in.
Why does the same site's Discover traffic swing so hard week to week?
Discover traffic is "somewhat less predictable than Search traffic," Google's Search Console help documentation states, because Google is proactively serving content rather than answering a specific query. Search Console's Discover performance report only becomes available once a property has crossed a minimum impression threshold, and it counts three metrics: impressions (a story scrolled into view, capped at one per result per session), clicks (shares don't count), and average click-through rate. The report defaults to a trailing three months, and the most recent days show as preliminary, dotted-line data that keeps updating for roughly 24 hours. None of that volatility is necessarily a signal that a site's editorial quality changed — it's closer to a function of how many readers currently have that topic in their Web & App Activity at a given moment.
That distinction matters for how an operator reads a Discover dip. A drop in Search impressions usually traces to something diagnosable — a ranking change, a technical error, a lost backlink. A drop in Discover impressions can simply mean the pool of readers whose current interests match a given topic shrank that week, with nothing on the publisher's end having changed at all. Search Console's own framing — proactive serving rather than query response — is why Discover reporting reads noisier than Search reporting even for a stable, policy-compliant site.
What did Google change in September 2025, and what does it mean for publishers?
In September 2025, Google revamped the Discover page to let readers follow specific publishers and creators by name and tap through to see everything from that source, TechCrunch reported. The rollout was set to expand to more content types beyond articles, including YouTube Shorts and posts from X and Instagram, in the weeks after the announcement. TechCrunch tied the change to a broader pattern: an August 2025 update that let readers pick preferred sources for Top Stories, and Google's earlier rollout of "Offerwall," a tool meant to help publishers diversify revenue as AI Overviews cut into their search traffic. TechCrunch also flagged an open question worth carrying into any planning around this feature: it reported that "it is not clear what Google's criteria are for creator content to show up on the Discover page," meaning the follow feature adds a new distribution lane without a published rulebook for who gets surfaced through it.
What can a publisher actually do about it, and what's outside their control?
Strip the mechanism down and the controllable levers are narrow:
- Get indexed and stay compliant with Discover's content policies — bylines, dates, and sponsored-content disclosure are non-negotiable inputs, not nice-to-haves.
- Invest in page experience and image quality, which Google names directly as ranking factors, rather than chasing a Discover-specific trick that doesn't exist.
- Treat the follow feature as a new but uncertain lane: it rewards being memorable enough that a reader taps a name, even though Google hasn't published how creator content is chosen for it.
- Don't read a single bad week of impressions as an editorial verdict — Discover's own documentation describes the channel as inherently less predictable than Search.
What's outside a publisher's control is the personalization layer itself: Discover decides, per reader, whether a story matches that reader's current interests, and no amount of on-page optimization substitutes for that judgment.
None of this makes Discover a lottery. It makes it a distribution channel with a published rulebook for entry — indexing, policy compliance, page and image quality — and an unpublished, personalized rulebook for who actually sees a given story once it clears that bar. Publishers who treat the eligibility criteria as a checklist and the ranking layer as something they influence indirectly, through consistent quality rather than direct targeting, are working with the mechanism as Google has described it. The channel rewards being eligible and being read, not being gamed, and Google's own documentation gives no indication that changes anytime soon.
For a related media news perspective, read Ben Affleck and Matt Damon Make a Splash on Gossip Stone TV.