Marfeel automatically applies a new method for identifying Google Discover traffic after a signal change that, according to the provider, became more significant on Oct. 6. Its product notice describes an attribution repair: better distinguishing the origin of visits the publisher already receives. The economic consequence follows when that label determines which editorial work receives funding.
The company announced its response on Oct. 7 and says it developed and validated the method within 24 hours of the main change. In its validation, less than 0.6% of Google Search traffic could move into the Discover category. The denominator is Search, rather than all Google traffic or visits labelled Discover. The notice does not disclose the sample, validation period or complete classification matrix.
Search remains a common entry point; that does not measure a publisher’s visits
Spanish context helps explain why an origin distinction can affect consequential decisions. The GWI series published through Statista records monthly search-engine use among 87.4% of internet users aged 16 to 64 in the second quarter of 2025. Monthly ChatGPT use was 30.5%. These are potentially overlapping activities by the same people, rather than a distribution of the visits publications receive.
Spain: two tools for accessing information
Monthly use among internet users aged 16 to 64. Second quarter of 2025.
Tap or hover for details. All values are also available in the data table.
Collapse notes
GWI online panel, reported by Statista and DataReportal. Sample size is not published in the statistic’s details. Overlapping activities; this does not measure news, clicks or publisher referral shares.
Explore the data table
| Category | Monthly use |
|---|---|
| Search engines | 87.4% |
| ChatGPT | 30.5% |
For an editor, that presence is not an instruction to move budget from one channel to another. A tool answering a question can resolve it without sending the user to a website. An actual arrival presents a subsequent decision: read, return or subscribe. Attribution should preserve that sequence rather than convert a tool’s popularity into expected revenue.
Moving visits between columns can consume resources without changing readership
The useful comparison starts with total arrivals from Google and then their distribution between Search and Discover. If the total barely moves while categories shift around the update, classification deserves investigation before commissioning more content or withdrawing an apparently weaker story. If the total also falls, the diagnosis requires additional evidence. Neither observation identifies the entire cause by itself.
Marfeel’s published percentage cannot establish Discover precision above 99.4%. That calculation would need correctly and incorrectly classified visits within the Discover category. A small fraction of Search can also represent a larger portion of Discover when the first source is much bigger. Without the publisher’s own volumes, there is no universal conversion between those denominators.
Social distribution raises a different question
Another Statista series, based on GWI for the third quarter of 2025, provides context for the social platforms where content circulates. Monthly Instagram use reached 73.5% and Facebook use 68.5% among Spanish internet users aged 16 to 64. That scale describes potential distribution environments, rather than people exposed to the article or readers who would leave Google to follow the publication there.
Spain: monthly presence on social platforms
Four selected platforms among internet users aged 16 to 64. Third quarter of 2025.
Tap or hover for details. All values are also available in the data table.
Collapse notes
GWI / DataReportal, via Statista. Online panel; sample size is not published in the statistic’s details. Platform use overlaps and must not be added. This does not measure news consumption or a particular publisher’s available audience.
Explore the data table
| Category | Monthly use |
|---|---|
| 73.5% | |
| 68.5% | |
| TikTok | 50.5% |
| 28.1% |
The comparison supports a second budget discipline: diversifying distribution also consumes editing, adaptation and attention. Maintaining content across networks requires the story’s promise to remain recognisable when somebody arrives. Choosing a channel solely by a national percentage confuses a broad population with the journey the publisher is prepared to serve.
Google and the website observe different moments in the same journey
The Search Console Discover report counts exposure and selection on Google’s surface. The publisher’s analytics observes arrivals and subsequent activity. The figures need not match: a click can precede a failed page load, and records can use different reporting days or URL groupings. Period, pages and event definitions need aligning before comparison.
The method-change date should remain beside the series when the report reaches management, editors or sales. Nor should the entire historical record be assumed to have been recomputed: Marfeel’s notice describes automatic application without specifying every customer’s historical reconstruction. An explained break is more informative than an apparently continuous line whose equivalence was never established.
Google describes Discover as less predictable than keyword-driven search . Repairing recognition helps investigate a publication’s dependence, but does not control distribution. The next editorial investment should follow evidence about readers and commercial continuation. The repair earns its value by preventing a creative response to a change that may have occurred in the report’s column.
In this article
Marian Vasile
Marian Vasile is a MarTech and growth professional. His work spans product marketing, SEO, paid advertising and deep online tracking, including behavioral analytics and attribution.