Google Discover in 2026: The Traffic Source Most Sites Ignore

Google Discover can send more traffic to a post in a weekend than that post’s organic search rankings will send it in a year — and then, without warning, send it almost nothing for months. I’ve watched this whiplash happen to sites I work with more than once, and the frustrating part is that Google’s own guidance is deliberately vague about why. That vagueness is worth understanding on its own terms, because it changes what’s actually worth doing about it.

What Discover eligibility actually requires

Google’s official documentation is short on requirements by design. To be eligible for Discover, a page needs to be indexed by Google and follow Discover’s content policies — there’s no special markup, no dedicated schema type, and no separate submission process. The main practical levers Google names directly are image quality (images at least 1200px wide, ideally 16:9, are eligible for the large-image Discover format) and avoiding what Google explicitly calls out as disqualifying: clickbait and sensationalism.

That clickbait guidance got a real revision in February 2025, rolled out alongside a core update — Google split its title guidance into more specific items and sharpened the language around what counts as exaggerated or misleading framing. If your headlines lean toward “you won’t believe” territory, that’s now explicitly the kind of thing Discover’s guidelines are written to exclude, not just discourage.

Why Google won’t tell you more than that

Google’s own documentation states plainly that traffic fluctuations from Discover may have nothing to do with a decline in content quality or publishing frequency — Discover is described as an evolving product, and shifts in what it surfaces can be about product changes on Google’s end rather than anything happening on your site. That’s not a dodge exactly, but it does mean the standard SEO instinct — “traffic dropped, something must be wrong with the page” — doesn’t map cleanly onto Discover the way it does onto search rankings.

Discover isn’t a ranking system you can audit the way you audit search visibility. It’s closer to a recommendation feed, and recommendation feeds change their taste without notice.

That volatility has been very real for publishers. Reporting around the December 2025 core update described some publishers seeing Discover traffic declines in the 70-85% range, with a few reporting drops as steep as 98% within the first weekend of the rollout. I’d treat those specific percentages carefully — they’re publisher self-reports circulated through industry channels, not numbers Google has confirmed or published itself. But the direction they point in — that Discover traffic can move dramatically and quickly, independent of anything measurable in a site’s search rankings — matches what a lot of site owners have experienced directly.

What’s actually different about optimizing for Discover

Search rankings reward matching a specific query. Discover doesn’t have a query at all — it’s surfaced based on a reader’s inferred interests and behavior, which means the content that performs well there tends to look different from content built purely around keyword targeting:

Timely and evergreen both work, but stale doesn’t. Discover favors content that feels current to the reader encountering it, which isn’t the same as “recently published” — a well-maintained evergreen piece can perform if it still reads as relevant.
Images carry more weight than in search. The large-image Discover card format means your featured image is doing real work, not just decorating the page. Low-resolution or generic stock photography visibly underperforms.
Titles need to be accurate, not just compelling. Given Google’s explicit February 2025 crackdown on sensationalism, a title that overpromises relative to the content is a genuine risk to Discover eligibility, not just a trust problem with readers.

What I’d actually do about the volatility

Given that Google itself says Discover fluctuations may be unrelated to your content, chasing every dip with changes to your publishing pattern is often wasted effort. What’s more useful:

Don’t build revenue plans around Discover traffic. Treat it as a bonus channel on top of a search strategy that stands on its own, not as a dependable baseline.
Keep featured images genuinely high quality — properly sized, relevant, and not generic stock. This is one of the few concrete levers Google names directly.
Write titles that would survive being read completely out of context, since that’s exactly how they’ll appear in a Discover feed next to unrelated content from other sites.
Watch the trend over months, not days. Given how much Discover volatility appears to be product-driven rather than content-driven, a single bad week tells you very little.

A pattern worth watching for in your own data

If you have historical Discover traffic in Search Console, it’s worth looking at whether your drops and spikes cluster around dates that match publicly reported core updates and Discover-specific guideline changes, rather than around your own publishing changes. When a decline lines up with a broad, industry-wide event like the December 2025 core update, that’s evidence pointing toward “this is a platform-wide shift,” not “something specific to my site broke.” When a decline is isolated to just your site with no corresponding industry-wide event, that’s a better signal to actually audit your own recent content and image choices rather than waiting it out.

Discover is genuinely one of the more frustrating channels to optimize for precisely because Google has been unusually candid that a lot of its behavior is out of publishers’ hands. The sites that handle it best, in my experience, are the ones that treat it as an unpredictable multiplier on already-solid content rather than a system they can reverse-engineer.