Google's February 2026 Discover Core Update: 22 Days That Reset Publisher Recommendation Rules

What changed: a standalone algorithm for the recommendation feed

The February 2026 Discover core update was the first core algorithm change Google scoped exclusively to its Discover feed - the AI-curated recommendation surface that now drives more publisher traffic than traditional Google Search for many news sites.

This is a broad update to our systems that surface articles in Discover. The update will improve the experience by showing users more locally relevant content from websites based in their country, reducing sensational content and clickbait in Discover, and showing more in-depth, original, and timely content from websites with expertise in a given area, based on our systems' understanding of a site's content. Google via Search Engine Roundtable - seroundtable.com

Previous core updates affected Search and Discover at the same time. This one operated independently and focused solely on the Discover feed algorithm, signaling Google's intention to treat Discover as a distinct product with its own quality standards. The separation matters because the two surfaces have different mechanics: Search responds to explicit demand from a typed query, while Discover proactively pushes content into a feed without being asked. Google has historically applied a higher quality bar to what it recommends than to what it returns for explicit queries.

The context adds urgency. Search Engine Roundtable's December 2025 webmaster recap noted that "Google said Discover is minimally aligned to search rankings, which may be why there is so much spam in Google Discover." February's update is best read as Google's response to that spam problem - a dedicated Discover ranking system that no longer borrows quality signals from Search. The three pillars Google named - geographic relevance, clickbait suppression, and topic-level expertise - are the levers it intends to pull from now on.

Effective date: the 22-day rollout and what surrounds it

Google originally estimated the rollout would complete within two weeks. It took 22 days - about eight days longer than expected. Search Engine Roundtable confirmed completion at 5:02 am ET on Friday, February 27, 2026, with Google posting that "the rollout was complete as of February 27, 2026." Effects are now live in production.

• Announced - February 5, 2026 (Google Search Status Dashboard)
• Rollout begins - February 5, 2026, global to Discover surfaces
• Original estimate - up to two weeks
• Actual completion - February 27, 2026 (22 days, 8 days over estimate)
• Follow-up updates - March 2026 spam update and March 2026 core update both rolled out within weeks of the Discover update completing

The surrounding update calendar matters for diagnosis. Search Engine Land reported the March 2026 core update began rolling out on March 27, 2026, and completed on April 8, 2026 after 12 days and 4 hours. If you are seeing traffic volatility, the first diagnostic question is which update window the change falls inside - February 5 to February 27 isolates this to Discover; later windows may overlap with the March spam update or the March core update.

Who's affected: severity by publisher profile

Impact concentrates in news publishing and content sites that have built Discover-heavy traffic mixes. The further your site is from concentrated topical depth, the higher the severity. The further from a US (or home-country) editorial base when serving that market, the higher the geographic exposure.

One caveat. A drop in Discover traffic during this window does not indicate a Search ranking problem. Google Search Console separates Discover performance from Search performance in its reports, and the first diagnostic step is to confirm which channel is responsible for any traffic change before drawing conclusions about overall SEO health.

What to do this week: diagnose before you change anything

Priority order: confirm the channel, isolate the cause, capture the highest-ROI technical fix, then wait at least two weeks before making structural content changes. Aggressive mid-rollout changes are historically associated with worse recovery, not better.

1. Separate Discover from Search in Search Console. Open Google Search Console and compare the Discover performance report against the Search performance report for the February 5 to February 27 window vs the prior 22 days. If only Discover changed, you are dealing with this update specifically. If both changed, something else is also in play.
2. Sort the Discover tab by impressions, not clicks. This is the most useful diagnostic in Search Console. Impressions reveal what Discover's algorithm considers a good audience fit for your site, even when click-through rates were modest. These are the content patterns to produce more of - the algorithm has already validated them as recommendation-worthy.
3. Run a geographic distribution analysis. If you have state-level or country-level data for your Discover traffic, compare pre- and post-update distributions. A loss concentrated in feeds outside your home market signals geographic recalibration, not a content quality problem. A loss distributed across all markets signals a content evaluation issue.
4. Audit the max-image-preview:large meta tag. Discover is a visually driven feed. Pages without the max-image-preview:large robots meta tag get smaller image previews and lower engagement. Verify the tag is present on all article pages and use original images at minimum 1200 pixels wide. This is the highest-leverage technical fix available right now.
5. Wait 14 days after the rollout completes before making major changes. The rollout finished February 27. Historically, sites that rewrite large sections of content, delete pages, or make aggressive structural changes mid-rollout struggle to recover more than sites that wait for stability. Use the window to diagnose; act on the diagnosis when signals settle.

What to do this quarter: build a Discover-resilient operation

The strategic shift in one line: Discover now deserves its own content operation rather than being treated as an accidental bonus on top of classic SEO. The February update reinforces a pattern that has been building for two years - Google is becoming more selective about topic-level expertise, more sensitive to original value, and less tolerant of sensational packaging or scaled sameness.

Concentrate topical depth over topical breadth

Sites that cover dozens of unrelated topics dilute their authority signals and are less likely to be recommended for any individual topic. Five excellent articles in your niche are now worth more than twenty thin posts across unrelated areas. Build internal linking between related articles to strengthen topic clusters - a site with ten interconnected pieces on cloud security sends a stronger topic-authority signal than a site with one cloud security piece among hundreds of unrelated posts.

Establish a consistent publishing cadence

The target is not volume - it is consistency. Sites that publish at least two to three high-quality articles per week typically achieve more stable Discover visibility than those with sporadic schedules. Publishing 10 articles one week and then nothing for a month sends mixed signals. Consistent cadence builds the publisher trust signal that keeps your site in the recommendation pipeline.

Invest in visual quality at the article level

Discover is a feed users scan visually before reading. Original, high-resolution images at minimum 1200 pixels wide are not optional. Pages without compelling imagery are effectively invisible in the feed. Combine the visual investment with the technical max-image-preview:large meta tag and you have done the two highest-leverage Discover technical optimizations.

Strengthen E-E-A-T at the author and entity level

Detailed author bios with credentials, links to author profiles on professional platforms, and demonstrated first-hand experience in the content all contribute to the entity signals Discover now weights more heavily. Beyond author pages, consider whether your organization has a clear entity presence - a Wikidata record, schema declaring relationships between your company and your products, and consistent author attribution across platforms.

From our accounts: what the tracker data shows

Note: the patterns below are aggregated from third-party Discover tracking tools (NewzDash, DiscoverSnoop) and from audit work on news and content sites during the February 5 to February 27 window. Tracking tools agreed on several patterns and diverged on others - which is itself instructive when interpreting any single data source.

Counterpattern worth flagging. The trackers disagreed on some individual sites, depending on which measurement window each vendor used. A tool measuring mid-rollout (say February 15) shows different results than one measuring post-completion (after February 27). Before drawing conclusions from any single third-party report, confirm the measurement window matches the rollout phase you are interested in. The cleanest comparison is post-completion vs the equivalent pre-launch window of the same length.

What we're still watching

Four open questions are shaping how we sequence Discover diagnostic work for the next two quarters.

• Discover-only cadence going forward: Whether Google continues to ship Discover-only updates independently of Search core updates, or whether February 2026 was a one-time recalibration. The next Discover-only update will tell us whether this is a new product line or a single intervention.
• Geographic precision over time: Whether the country-level domesticity signal evolves into state- or region-level precision in subsequent updates. NewzDash's tracking already shows meaningful "local layers" of content differing between California, New York, and US national feeds. Granular geographic ranking would be a significant follow-on.
• Interaction with the March 2026 core update: Whether sites hit by the Discover update saw further movement during the March 27 to April 8 core update window. Compounding effects across surfaces could either accelerate recovery or deepen losses, and the data on this is still settling.
• Recovery patterns for topic-focused sites: Whether sites that respond by concentrating topical depth (rather than chasing breadth) recover within one to two update cycles, or whether the new topic-expertise signal takes longer to accrue. The historical recovery pattern is "usually with future updates, not immediate fixes" - the open question is how many cycles.