YouTube's 18x surge in Google Discover reveals a hidden video pipeline
A new study maps the three-stage YouTube cascade inside Google Discover that grew 18x in a single quarter, now reaching 13% of English devices by February 2026.
Search
A research study published on April 27, 2026, by Sylvain Deaure and Damien Andell of 1492.vision has mapped a hidden three-stage YouTube amplification pipeline operating inside Google Discover. The pipeline - built from three named components, creatorcontent, freshvideos, and neoncluster - grew 18.2x between December 2025 and February 2026, ending the quarter with a reach of 13% of devices in English-language markets. The study draws on 42 million Discover cards collected across hundreds of devices during that three-month window.
The numbers are striking for a distribution channel that operates largely out of sight. Publishers and marketers focused on traditional web traffic have had little visibility into how YouTube videos travel through Discover's internal architecture. This research is among the first systematic attempts to document the cascade mechanics in detail.
Three stages, one funnel
The pipeline begins with creatorcontent, described by the researchers as the entry point for social content into Discover. According to the study, YouTube accounts for 72% of the content in this pipeline in English-language markets, with x.com contributing 23%. The median content age at this stage is 1.9 hours - near-real-time ingestion. Reach sits at 6.7%. Between December 2025 and February 2026, creatorcontent grew 7.8x, expanding from roughly 48,000 hits to 371,000 hits per month.
The second stage, freshvideos, acts as an amplifier with a delay of approximately seven hours after content enters creatorcontent. According to 1492.vision, freshvideos in English markets is 94% video and 94% YouTube. Reach at this stage is 7.1%, and the pipeline grew 7.2x over the same three-month period. The content filtering is visible in the source domains. ESPN, CNN, Formula 1, and the Olympics appear here alongside TikTok, suggesting that sports and news video from premium publishers passes through this stage on the way to broader distribution.
The third stage is neoncluster, the broadcast endpoint. It is 100% YouTube and 100% English. Reach reaches 13% - higher than the general content pipeline, which the researchers place at 8.8%. According to the study, neoncluster arrives approximately 15 hours after freshvideos and 23 hours after creatorcontent, completing a full transit time from social intake to broadcast of roughly 17.3 hours. Each video that reaches neoncluster generates an average of 8.8 hits per URL, meaning a single video surfaces across many devices once it clears the final filter. Growth from December to February was 18.2x, with hits expanding from 24,000 to 431,000 per month.
The content profile of neoncluster is specific. According to 1492.vision, the pipeline carries news-heavy YouTube channels focused on international news, celebrity and entertainment, and politics. Gaming and vlogging content does not characterize this layer. What emerges is YouTube functioning as a structured news distribution network inside Discover, not a general video platform.
The funnel mechanics
Not all content entering creatorcontent survives to broadcast. According to the study, only 16% of creatorcontent URLs appear in freshvideos. Neoncluster is more selective still: the researchers found 74,000 URLs from 130,000 in creatorcontent. The cascade is, by design, a funnel - wide at the intake point, narrow at broadcast. Content quality increases at each stage, measured by the proportion of video: creatorcontent is a mixed feed, freshvideos is 94% video, neoncluster is 100% video.
The researchers describe this structure as a quality filter and amplifier, structurally similar to another internal Discover pipeline they identify as deeptrendsfable leading to deeptrends, but built for video content. This comparison to a parallel pipeline architecture suggests the three-stage cascade is not an isolated design choice but part of a broader architecture Google has developed for different content categories.
An English-language phenomenon
The language dimension of this data is notable. The cascade as described does not function in French. According to 1492.vision, neoncluster in France recorded only 20 hits in February 2026 - effectively non-existent in that market. freshvideos in French markets is only 53% video, with TF1 and L'Equipe accounting for 47% articles. The creatorcontent composition inverts entirely: in France, x.com dominates at 72% while YouTube is under 5%.
The authors describe this inversion as a reflection of platform culture. YouTube holds dominant influence in anglophone markets as both a creator platform and a news distribution channel. x.com carries outsized influence in francophone media and politics. According to the study, Discover mirrors this, the same pipeline tuned to different ecosystems.
For English-language publishers, the implication is that the amplification path from creatorcontent through freshvideos to neoncluster is effectively a video-only track. Text content does not follow the same route. x.com posts do enter Discover through creatorcontent - they account for 23% of that pipeline in English markets - but the downstream stages filter them out. There is no equivalent broadcast amplification path for text at scale.
Scale in context
The 13% reach figure for neoncluster needs framing. Google Discover is primarily accessed through the Google app on Android and iOS devices, and through mobile Google Search. It is the default view when swiping left on many Android home screens and a prominent feature of the Google app on iPhone. A pipeline that reaches 13% of devices in English markets is not a marginal test. For comparison, the general content pipeline sits at 8.8% reach according to the same data, meaning neoncluster now exceeds it.
The full pipeline dataset from the Discover Pipeline Explorer provides a wider context. feedads, representing sponsored feed ads, records the highest reach of any pipeline at 58.4%, with visibility of 11.6%. shoppinginspiration, covering shopping and product cards, reaches 13.1% - nearly identical to neoncluster. discover_ai_summary, covering AI Overview cards, shows reach of 3.5%. The astria pipeline for niche and vertical content reaches 5.7% with 1.8% visibility. Among all pipelines, neoncluster's 18.2x growth makes it the fastest-growing by a substantial margin.
These figures come from the Discover Pipeline Explorer, a free interactive tool published at 1492.vision/interactive/xpl/, mapping the full set of identified Discover pipelines across the December 2025 to February 2026 measurement window.
The growth trajectory
The researchers are explicit that the pipeline is not a mature system operating at steady state. According to the study, these pipelines were launched or dramatically expanded in late 2025. The growth curves are vertical, not gradual. At 18x in three months, neoncluster moved from negligible to the third-largest English Discover pipeline. The study notes that at 18x growth in three months, if these growth curves continue even partially, the social and video pipeline family will rival the core editorial pipelines by mid-2026.
This trajectory fits within a broader documented pattern. In September 2025, Google added content from X, Instagram, and YouTube Shorts to the Discover feed alongside a follow mechanism for publishers and creators. That announcement framed the changes as a response to user interest in mixed-format feeds. The pipeline data from 1492.vision provides a view of what happened inside Discover's architecture in the months that followed.
By December 2025, data from Marfeel showed that YouTube had absorbed a material share of Discover positions, with 51% of feed positions in the United States, Brazil, and Mexico occupied by AI Summaries, and YouTube taking a large share of the remaining space. The 1492.vision pipeline study extends that picture into the internal architecture, naming the specific pipelines and measuring their growth.
Android, Google Search, and YouTube as an integrated system
The pipeline data illustrates a dynamic that antitrust proceedings have examined from a different angle. Google Discover is primarily accessed through the Google app on Android devices and through mobile Google Search - both surfaces controlled by Google. Android is preinstalled on hundreds of millions of devices globally, and court findings in the Google search antitrust case established that Google paid $26.3 billion in 2021 alone to maintain its position as the default search engine across smartphones and browsers. That default placement is the mechanism through which most Discover users encounter the feed in the first place.
YouTube is a Google property acquired in 2006. A federal court antitrust case brought by newspaper publishers alleged that Google used its acquisitions of Android, YouTube, and DeepMind to foreclose competition in lines of commerce related to its general search monopoly. The court dismissed those claims as time-barred under the four-year Clayton Act limitations period. But the structural argument - that Android and Google Search function as distribution infrastructure for YouTube - is made visible by the pipeline data. A video that enters the creatorcontent pipeline and completes the full cascade to neoncluster does so inside a Google-controlled app, running on a Google-defined default surface, delivered to a device where Google Search is the default.
The Rumble antitrust case, dismissed on statute-of-limitations grounds in May 2025, had argued that Google's Mobile Application Distribution Agreements required device manufacturers to bundle YouTube with other Google apps, and that Google rigged search algorithms to favor YouTube over competing video platforms. No court has ruled on the merits of that structural argument. The neoncluster growth curve is consistent with the described dynamic, but the research does not attribute intent - it measures outcomes.
What this means for media buyers and publishers
For performance marketers and media buyers, the pipeline study raises practical questions about how video content intersects with the Discover surface. The feedads pipeline - sponsored feed ads - has a reach of 58.4% in the dataset, confirming that paid placements operate at substantial scale in Discover. The organic video cascade documented by 1492.vision runs alongside this paid layer, and their interactions remain unmapped by this study.
The February 2026 Discover core update prioritized locally relevant content and reduced sensational material, changes that affected text publishers heavily. Video content moving through the neoncluster pipeline - news-heavy, international, high-reach - was not explicitly the target of those changes according to the documented objectives of the update. The December 2025 core update had triggered severe Discover traffic losses for news publishers, with some reporting 70-85% declines in daily visitor counts. The 1492.vision pipeline study was conducted during and after that disruption period. Neoncluster's growth from 24,000 to 431,000 hits per month occurred in the same window where text publishers were losing traffic. The researchers do not claim a direct connection between those two trends, but the simultaneous occurrence is documented in the data.
The cascade rewards consistency. With 8.8 hits per URL at the neoncluster stage, content that clears the filter surfaces across many devices rather than spiking once and disappearing. A single viral moment does not characterize what this pipeline amplifies. Regular production from channels covering news, politics, and international current affairs appears to be what the pipeline selects for, based on the content profile described in the study.
The data behind the study
The 1492.vision research covers 42 million Discover cards collected between December 2025 and February 2026 across hundreds of devices. The methodology involves tracking individual pipeline labels attached to Discover cards - internal identifiers that Google uses to categorize how content was selected. By logging these labels at scale, the researchers measure reach (the proportion of devices that received at least one card from a given pipeline) and visibility (the share of total impressions attributed to that pipeline).
The study is published as part 5 of a seven-part series on Google Discover pipeline architecture. Earlier in the series, part 3, published April 13, examined how a viral article travels through Discover. The interactive pipeline explorer at 1492.vision/interactive/xpl/ allows comparison of pipelines by reach, visibility, content type, and growth over the measurement period.
Timeline
- August 5, 2024: Judge Amit Mehta rules Google illegally maintained a search monopoly. Court findings establish that Google paid $26.3 billion in 2021 to maintain default search engine placement across devices - the same distribution infrastructure that delivers Discover to Android users. (PPC Land)
- May 21, 2025: Rumble's $2 billion antitrust lawsuit against Google is dismissed on statute-of-limitations grounds. The case had alleged Google's Android agreements required device manufacturers to bundle YouTube and that Google favored YouTube in search results.
- September 17, 2025: Google announces updates to Discover, adding content from X, Instagram, and YouTube Shorts alongside a follow mechanism for publishers and creators.
- Late 2025: According to 1492.vision, the creatorcontent, freshvideos, and neoncluster pipelines are launched or dramatically expanded inside Google Discover.
- December 11, 2025: Google's December 2025 core update triggers severe Discover traffic collapses for hundreds of publishers, with some reporting 70-85% declines in daily visitor counts within 48 hours.
- December 2025: Data from Marfeel shows 51% of Discover feed positions in the US, Brazil, and Mexico occupied by AI Summaries, with YouTube absorbing a material share of remaining positions.
- December 2025 - February 2026: 1492.vision collects 42 million Discover cards across hundreds of devices. neoncluster grows from 24,000 to 431,000 monthly hits - 18.2x in three months.
- February 5, 2026: Google releases the February 2026 Discover core update targeting English users in the US, prioritizing locally relevant content and reducing clickbait material.
- April 27, 2026: Sylvain Deaure and Damien Andell publish "YouTube, X, and the social pipeline explosion" on the 1492.vision Substack as part 5 of a seven-part series on Google Discover pipeline architecture.
Summary
Who: Sylvain Deaure and Damien Andell, researchers at 1492.vision, published the study on April 27, 2026. The findings are relevant to YouTube video publishers, news organizations, media buyers, and digital marketers operating in English-language markets.
What: A three-stage pipeline inside Google Discover - composed of creatorcontent, freshvideos, and neoncluster - amplifies YouTube video content from social intake to broadcast-level distribution. The neoncluster stage grew 18.2x between December 2025 and February 2026, reaching 13% of English Discover devices. The pipeline is 100% YouTube and 100% English at its final stage, carrying news-heavy video content covering international news, celebrity, entertainment, and politics.
When: The pipeline data covers December 2025 through February 2026. The study was published on April 27, 2026, as part 5 of a seven-part series on Google Discover pipeline architecture.
Where: The pipeline operates inside Google Discover, primarily accessed through the Google app on Android and iOS devices. The neoncluster pipeline is an English-language-only phenomenon. In France, it recorded only 20 hits in February 2026. The study draws on 42 million Discover cards collected across hundreds of devices.
Why: The findings document a distribution pathway for video content that operates largely invisibly and has grown faster than any other Discover pipeline in the measurement window. For publishers making decisions about video production, the data provides concrete reach figures. For antitrust observers and market analysts, the pipeline illustrates how Android's default distribution infrastructure and Google's control of Discover create conditions that structurally favor YouTube - a Google property - within Google's own content recommendation system.