Google's First Discover-Specific Core Update Took 22 Days. Here's the Winner-Loser Data.

person

AI SEO Intelligence

calendar_today May 24, 2026

schedule 10 min read

Google's First Discover-Specific Core Update Took 22 Days. Here's the Winner-Loser Data.

Forbes lost 67% of its Google Discover audience. Parade.com gained 1,300%. Yahoo dropped from 3rd place among US Discover publishers to 9th. These are not typos. They are the numbers from the first core update Google has ever publicly labelled as Discover-specific, separate from the broader core update sequence — and three months on, the post-rollout analyses converge on the same story.

If you run a publisher site, a content business, or a media operation that gets meaningful traffic from the Google app on mobile, this update probably moved you in a measurable direction. If you do not run one, the update still matters: it is the cleanest data point we have on how Google is rebalancing the entire "content via Google" surface area in 2026. Read alongside the May 7 FAQ rich result deprecation and the March 2026 Core Update, a single pattern emerges.

Key Takeaways

The February 2026 Discover Core Update rolled out Feb 5 – Feb 27 (~22 days, eight days longer than Google's two-week estimate). First-ever Discover-specific core update; English-language US users only; global expansion announced but undated.
Top US Discover publishers contracted from 172 to 158 unique domains — Discover got more concentrated, not more diverse.
Concrete winners: Parade.com +1,300% audience, Fortune +800%, WSJ +86%, Axios +33%, YouTube +15%. Concrete losers: Fox Business −90%, Forbes −67%, Yahoo −62%, The Sun −67%, The Independent −57%.
Discover now drives ~68% of Google-sourced traffic to news publishers, up from 37% in 2023. Traditional Search dropped from 51% to ~27% over the same period.
All three 2026 algorithm moves — February Discover update, March Core Update, May 7 FAQ deprecation — push the same direction: original destinations up, aggregators / templated / shallow content down.

The Rollout

Google announced the Discover Core Update on February 5, 2026 and confirmed completion on February 27. Search Engine Roundtable's tracking and Search Engine Journal's coverage both note this was the first time Google publicly framed a core update as Discover-specific, decoupling it from the broader algorithmic refreshes that hit Search.

That decoupling is the part most analysts undersold at launch. Discover and Search had been treated as different surfaces with shared signals, occasionally divergent rankings, but no separate update cadence. Pulling the Discover update out of the broader rollout is the algorithmic equivalent of declaring Discover a first-class product line, not a downstream consumer of Search ranking. Whatever signals Google is now weighting differently in Discover, they are different enough to warrant separate testing, separate rollout windows, and separate official communication.

Scope was deliberately narrow: English-language US users only. Google has said global expansion will happen but has not committed to a date. For non-US publishers, the implication is straightforward — your home market is intact for now, but your traffic from US-based readers is on the move.

The Numbers

ALM Corp's analysis of US top-1,000 Discover domains is the most detailed publicly available dataset. The headline movements (placements measured as share of Discover slot impressions; audience measured as reached unique users):

Losers: - Fox Business: −52% placements, −90% audience - Forbes: −21% placements, −67% audience (ranking dropped from #56 to #193) - The Sun: −67% placements - The Independent: −57% placements - Yahoo: −47% placements, −62% audience (ranking dropped from #3 to #9) - Economic Times (India, targeting US): −43% placements

Winners: - Parade.com: +208% placements, +1,300% audience (ranking jumped from #220 to #10) - Fortune: +121% placements, +800% audience - WSJ: +86% placements - Axios: +33% placements - YouTube: +15% placements

The shape of the redistribution is more interesting than any single number. The top of the leaderboard got reshuffled — Yahoo lost six positions, Parade gained 210, the publisher list contracted from 172 to 158 unique domains. Google's Discover algorithm is now both more selective (fewer unique publishers in the top tier) and more dynamic (the publishers who do get in can rise or fall hundreds of positions in a single update). For a publisher running a Discover-heavy strategy, that's a different risk profile than the Search side, where movements of even 50 positions for a major domain would make headlines.

A separate finding: local content appeared roughly 5× more often in matching regional feeds post-update. Syracuse.com and CBS6 Albany lost out-of-state traffic but held — and in some cases grew — their home-region audience. Geography became a stronger weighting input in Discover, not weaker.

Why Discover Now Matters More Than Search (for News)

The single most consequential number in the post-update writeups is not in the winner/loser table. It is this: Discover now accounts for roughly 68% of Google-sourced traffic to news publishers, up from 37% in 2023. Traditional Search over the same period dropped from 51% to about 27%.

For a category of site that thinks of itself as ranking on Google, "Google" no longer means "Search." It means "Discover," with Search as a meaningful secondary stream. The February 2026 update is the moment Google stopped treating Discover as an experimental side surface and started treating it as the primary content distribution layer for the Google app on mobile.

The implication for publisher strategy is large. SEO tools, dashboards, and KPI structures built around Search rankings are measuring the smaller half of the relationship. Discover-specific signals — image dimensions, predicted engagement, headline-content alignment, freshness windows — are not just "nice to have"; they are now the dominant ranking inputs for the larger half of news traffic. The publishers who jumped 200 positions in February did not crack a new Search ranking factor. They got better at the signals Discover specifically weights.

How Discover Actually Ranks

Discover does not use queries. There is no keyword to optimize against, no search intent to match. The ranking is, in Google's framing, "predicted engagement" — the algorithm guesses which articles a specific user, given their interests and recent behavior, is most likely to tap on.

That changes what optimization means. ClickMediaLab's 2026 Discover ranking factors guide consolidates the signals that have produced measurable Discover gains in 2025–2026:

Image dimensions matter literally: posts with images at least 1200 pixels wide and meta name="robots" content="max-image-preview:large" see roughly 45% higher click-through in Discover. Smaller images get a smaller card; smaller card means lower CTR; lower CTR means future deprioritization. The deeper dive — including Google's own Kirbie's Cravings case study with a 79% Discover lift after enabling the directive — is in Discover Optimization Is Not SEO Optimization.
Freshness windows are short: most Discover traffic to a given article arrives in the first 48 to 72 hours. After that the algorithm rotates it out. Articles that were never going to be evergreen — news, opinion, real-time coverage — fit Discover natively. Evergreen reference content does not.
Predicted engagement, not relevance: the algorithm is trying to model "will this user tap this card" rather than "does this article match this query." That favors content with strong opening hooks, visible authority signals, and headlines that promise something specific rather than vague.

None of this looks like classic SEO. It looks like editorial choices the algorithm is now scoring at scale.

The Pattern Across All Three 2026 Algorithm Moves

The February Discover Core Update did not arrive alone. The March 2026 Core Update finished rolling out April 8, with SISTRIX visibility losses on a scale not previously seen for major platforms: YouTube lost 567 visibility points, Reddit 64.2, Instagram 48.1, X (Twitter) 45.9. "Arts & Entertainment" dominated the loser lists. Destination sources gained, aggregator and templated content lost. Then on May 7, 2026, Google deprecated FAQ rich results entirely, removing a SERP feature that had been gamed for years.

Read separately, these are three different updates with three different scopes. Read together, they push the same direction:

Original destinations up. Parade jumped 200 positions in Discover. WSJ, Fortune, and Axios — strong-byline original-reporting publishers — gained. Forbes and Yahoo — strong aggregators of contributor and wire content — lost.
Intermediaries down. UGC platforms (Reddit, Instagram, X) lost meaningful Search visibility in the March update. Aggregator and templated content categories led the loser lists.
Low-effort SERP signals removed. FAQ schema lost its rich-result reward on May 7. HowTo went the same way in 2023–2024.

The through-line is not subtle. Google is consistently demoting content patterns that earn ranking via structure rather than substance, and consistently rewarding content patterns that demonstrate original work. (We dug into the corresponding April 2026 amplification of Experience signals in our companion post on the FAQ deprecation; the Experience signal weight increase fits the same pattern.)

That pattern is not Google policy — Google has not published a unified statement tying these three updates together. It is an inference from observed traffic movements across multiple independent analyses. Treat it as a strong working hypothesis rather than confirmed mechanics. The hypothesis is, however, actionable enough to drive priorities.

Recovery Playbook for Discover-Hit Publishers

If you lost Discover traffic in the February rollout, five tactics show up consistently across the post-update analyses (most concretely in ALM Corp's local-publishers analysis):

1. Run a state-level geographic audit. Discover losses for many publishers concentrate in out-of-state regions where the algorithm now favors local sources. Syracuse.com lost national traffic but held Syracuse. Look at Search Console region-by-region rather than just at the national total — the recovery plan for "lost most readers in 38 states" is different from "lost readers everywhere."

2. Close the headline-content gap. Discover's predicted-engagement model penalizes mismatches: a headline promising one thing while the article delivers a thinner version of it. The publishers that gained tend to write headlines that under-promise relative to the content underneath, not the other way around.

3. Concentrate on 2–3 core topics, not broad coverage. Topical depth beat topical breadth in this update. Sites that ranked for everything lost ground to sites that owned narrower verticals.

4. Invest in image quality, not just image presence. Native 1200×630 (or wider) images with max-image-preview:large aren't a nice-to-have for Discover, they are the input that determines card size and therefore CTR. Stock photos paste poorly; original photography pastes well.

5. If your strategy was aggregation, change it. This is the hardest one because it changes content economics, not just content production. Aggregator publishers got hit hardest. Sites doing original reporting — even small original reporting layered on top of wire content — fared better. Forbes' contributor-network model is now under direct algorithmic pressure; WSJ's staff-byline model is benefiting from the inverse.

What Non-US Publishers Should Do Before the Global Rollout

Google's announcement explicitly limited the February update to US, English-language users with global rollout to follow. Non-US publishers have a finite window — exact length unknown — before the same signals start weighting their home markets. The actionable preparation:

Baseline Discover impressions now, in Search Console, separated by country. Once the global rollout lands, you want a clean before/after.
For publishers heavily targeting US audiences from outside the US: expect a hit. The Economic Times (India) and similar US-targeting international publishers were already affected. Reduce dependence on US Discover traffic in revenue forecasting until you understand the rebalanced baseline.
For publishers serving their home markets only: prioritize regional / local signal density — author bios with regional credentials, local references in content, regionally specific imagery. The signal that lifted Syracuse.com domestically is the signal you want strong before the global rollout.
For publishers in non-English markets: the update is currently English-only. Treat this as research time, not safety. The same algorithmic preferences will almost certainly apply to your language once Google extends.

What This Doesn't Mean

A short list of things the data does not say, in case anyone is over-extrapolating:

It does not mean Discover is replacing Search. Search is still 27% of news traffic and ~100% of long-tail / commercial query traffic. The Discover share is news-specific.
It does not mean every Forbes article will lose Discover traffic — Forbes' contributor-network model is what took the hit, not Forbes-as-a-publisher. Strong staff bylines on Forbes still appear in Discover.
It does not mean Google has officially tied this update to the March Core Update or May 7 FAQ deprecation. The through-line is an analyst-level inference, not Google's framing.
It does not mean recovery is a 30-day project. Most publishers reporting recovery did so 60–90 days post-rollout. The February update is fresh enough that some of the "winners" may yet slip and some of the "losers" may yet recover partially.

The honest read is that this is the cleanest signal yet that the Discover surface needs to be optimized differently from Search — and that the publishers who understand the difference will gain visible ground in 2026. We're not Discover-specific in our audit (yet), but the Experience and original-reporting signals that win in Discover overlap heavily with the signals our content checks already evaluate.

Want to see how your pages score on the signals that overlap between Search, Discover, and AI search? Run a free audit at hybridranking.com — the 99 signals we evaluate include several that map directly to the patterns Discover now rewards.