In February 2026 Google adjusted Discover independently of search ranking for the first time, and the pool of domains appearing in US Discover feeds narrowed from 172 to 158. That matters more to publishers than any core update of the same period, because Discover can be a majority of a news site’s traffic and it now has to be diagnosed as its own channel. A publisher can hold every search ranking it had and lose half its traffic, and nothing in the web Search report will explain why.
Discover is a separate system now
Until February 2026, Discover visibility moved as a side effect of broad core updates. The February 2026 Discover core update, which ran from 5 to 27 February, was the first time Google changed what gets recommended in the feed without touching search rankings.
The practical consequence is a diagnostic one. If you report Discover and Search together, a Discover collapse looks like a general traffic decline and you will spend months on search quality work that cannot fix it. Switch Search type to Discover in Search Console before drawing any conclusion. That single filter change resolves a large share of publisher misdiagnoses.
| What you see | Measurement (where to check) | What it means |
|---|---|---|
| Discover fell, web search held | Search type: Discover versus Web, same date range | A Discover-specific change, not a ranking loss |
| Both fell together | Compare the two reports across the same window | Check the core update dates as well |
| Discover has always been spiky | Look for a sustained step, not day-to-day swings | Normal feed behaviour rather than an update |
| No Discover report at all | The report only appears once you have Discover traffic | You were never featured, which is a different problem |
What the February 2026 update targeted
Google published three goals for this update, which is more detail than core updates usually get.
- Surface more locally relevant content. Why: Discover is a personalised feed, and geographic relevance was weighted up as a recommendation signal. Regional publishers benefited relative to national aggregators.
- Reduce clickbait. Why: Headlines that overpromise relative to the article were the clearest losers in post-update analysis. This is an editorial standard rather than an SEO tactic.
- Prioritise original and in-depth reporting over rehashed material. Why: Shallow coverage of trending topics, the most common Discover growth tactic of the last five years, lost ground to original reporting.
The measurable outcome was that the pool narrowed. Unique domains appearing in US Discover feeds fell from 172 to 158, which suggests Google tightened which publishers it considers recommendable at all rather than simply reshuffling the order. Several large general publishers lost visibility while specialist and Google-owned properties gained.
At launch the update applied to English-language content in the United States only, with Google confirming plans to expand to other languages and regions later in 2026. Publishers outside the US should treat this as a preview rather than as somebody else’s problem.
What the measurement showed
Third-party tracking during the rollout gives the clearest picture of what changed, and the geographic pattern is the part most publishers missed.
| Measurement | Before | After | What it means |
|---|---|---|---|
| Unique domains in US Top 1000 Discover placements | 172 | 158 | An 8.1 percent decline. Consolidation toward established publishers |
| Share held by US-based publishers | 88.86% | 89.94% | Domestic publishers gained |
| Share held by international publishers | 8.52% | 7.04% | Foreign outlets lost ground in US feeds |
| Reuters visibility | Down by up to 20 percent | Even major international outlets were affected | |
The second and third rows matter more than the headline number. Google now favours content from publishers based in the user’s own country, which means a strong international outlet can lose US Discover visibility without its content changing at all. If you publish from outside the market you are trying to reach, that is a structural disadvantage rather than a quality problem, and no amount of editorial improvement fully offsets it.
Expertise is also now assessed topic by topic rather than across a whole site. A publisher with genuine authority in one area no longer carries that credit into unrelated coverage, which is the mechanism behind broad general outlets losing to focused specialists.
Figures from NewzDash DiscoverPulse tracking during the February 2026 rollout.
Headlines: the fastest thing to fix
Clickbait reduction was a stated goal, and headlines are the one lever a publisher can pull this week rather than this quarter.
The working test: would a reader who clicked feel the headline described the article honestly? Not whether it was exciting, and not whether it earned the click. Whether it kept its promise. Curiosity gaps, withheld outcomes and implied revelations that the piece never delivers are exactly the pattern that lost ground.
This is uncomfortable for publishers whose Discover strategy was built on headline optimisation, because the tactic that drove the growth is the one being penalised. There is no way to keep the traffic model and drop the practice, since they were the same thing.
Aggregation versus original reporting
Worth being precise, because aggregation is not inherently the problem. Aggregation with original selection, context and clear attribution is a legitimate publishing model that has worked for decades.
What lost ground was reproduction with nothing added: covering a trending story because it is trending, with no reporting, no local angle and no expertise, published fast enough to catch the search volume. That is the model the update narrowed, and it overlaps with the scaled content abuse policy introduced in March 2024. A publisher on both sides of that line has a Discover problem and a spam policy problem at once.
- Add reporting to stories you would have aggregated. Why: One phone call, one local source, one figure nobody else has. It does not need to be an investigation to count as original.
- Cover your region properly rather than the national story badly. Why: Local relevance was weighted up. Competing with national outlets on their story is a losing position that geography can reverse.
- Publish fewer pieces with more in them. Why: The narrowed domain pool means being recommendable at all is now the bar, and volume does not help you clear it.
- Make bylines real and verifiable. Why: Named journalists with a history readers can check. Anonymous news content fails the trust standard before any other factor is considered.
Recovering Discover traffic
Discover recovery behaves differently from a core update recovery, and the difference works in your favour.
- There is no scheduled reassessment to wait for. Why: The feed evaluates content as it is published, so a genuine editorial change starts being tested within days rather than waiting on a core update cycle.
- New content carries the recovery, not the archive. Why: Discover is a recommendation feed for current material. Rewriting old articles does almost nothing here, which is the opposite of a search quality recovery.
- Consistency matters more than any single piece. Why: Being reconsidered as a recommendable domain depends on a sustained pattern. A month of good work followed by a return to aggregation resets it.
- Expect volatility even after recovery. Why: Discover traffic is inherently spiky. Judge it on monthly trends, never on daily numbers.
If your web search traffic also fell, that is a separate diagnosis running in parallel. Use the drop analyzer on the Search report specifically, and the page scope analyzer to see whether the loss is site-wide or confined to one section.
Frequently asked questions
Why did my Discover traffic drop but search traffic stay the same?
Because Discover now runs under its own evaluation. The February 2026 Discover core update changed feed recommendations without touching search rankings, so the two can move independently. Switch Search type to Discover in Search Console to confirm.
What did the February 2026 Discover update target?
Three stated goals: more locally relevant content, less clickbait, and original in-depth reporting over rehashed material. Unique domains in US Discover feeds fell from 172 to 158, indicating a narrower pool rather than a reshuffle.
How long does Discover recovery take?
Faster than a core update recovery, because there is no scheduled reassessment. The feed evaluates new content continuously, so an editorial change can be tested within days. Being reconsidered as a recommendable domain takes a sustained pattern, usually months.
Should I rewrite old articles to recover Discover traffic?
Generally no. Discover recommends current content, so recovery is carried by what you publish next rather than by the archive. That is the opposite of a search quality recovery, where the archive is often the whole problem.
Does this update apply outside the United States?
At launch it applied to English-language content in the US only, with Google confirming plans to expand to other languages and regions later in 2026. Publishers elsewhere should treat it as a preview of what is coming.
Discover, search, or both?
Reported together they look like one problem. They are two systems with different fixes, and publishers routinely spend months on the wrong one. Send your Search Console access and within 48 hours I will tell you which channel actually moved. Free, no obligation.
Get free recovery audit