Same pipeline names. Same labels. Yet the data tells two structurally distinct stories. Here are the ten divergences our 42 million cards reveal β and what they mean for publishers operating in both markets.
Most publishers and SEOs think of Google Discover as a global product. One algorithm, one logic, language adjustments. Our data shows something else: the French feed and the English feed are two different products sharing the same infrastructure.
The same 20 pipelines exist in both markets. But their proportions, their sources, and sometimes their very existence diverge radically. A publisher who transposes their Discover strategy from one market to the other without adjustment is making a mistake β and our data explains why.
For detailed per-market analyses: FR reference | EN reference
Before diving into the detail, the summary table. Each row is a finding in its own right β and each links to its detailed analysis.
| # | Dimension | FR | EN | Gap |
|---|---|---|---|---|
| 1 | Moonstone (reach) | 19.3% | 9.4% | 2x FR advantage |
| 2 | AIO | ~0% | 3.5β29% depending on pipeline | EN-only |
| 3 | Video in the feed | < 5% of volume | 13% (neoncluster alone) | EN = video |
| 4 | Social source | x.com 75% (creatorcontent) | YouTube 72% | Inverted |
| 5 | neoncluster | 36 hits in 3 months | 4.5% of volume, 13% reach | EN-only |
| 6 | feedads (reach) | 24% | 58.4% | 2.4x EN |
| 7 | Shopping lifespan | 3.7 days | 2.5 days | FR lasts longer |
| 8 | Multi-labeling | 58% in 2+ pipelines | 37% in 2+ | FR more multi-pipeline |
| 9 | EPL exclusion | N/A | Active in 7+ pipelines | EN-specific |
| 10 | Pipeline growth | creatorcontent 33x | neoncluster 18x | Different explosions |
The 10 major divergences between Discover FR and EN. Same infrastructure, two functionally different systems.
This table is not a summary. It's a map. Each divergence is an architectural choice β or a reflection of underlying markets β that changes the optimal strategy for a publisher.
Before diving into each divergence, let's place the two maps side by side.
FR: moonstone dominates in reach (19.3%). neoncluster is invisible (36 hits). feedads is present but moderate (24%). The feed is centered on editorial and engagement.
EN: neoncluster appears at 13% reach (absent from FR). feedads explodes at 58.4%. moonstone is halved. The feed is centered on video, advertising, and AIO.
The visual comparison is striking. Two products. The same labels.
FR 19.3% reach β EN 9.4%. A 2x gap.
Moonstone is the engagement broadcast pipeline β it selects articles that generate clicks and shows them to the maximum number of devices. In France, it's the pipeline with the highest editorial reach. In English, its reach is halved.
The gap is not a measurement artifact. It reflects a structural difference in broadcast competition:
Sources change too. In FR, moonstone is dominated by traditional press β Ouest-France (9%), BFM TV (8.8%), Le Figaro (6.8%). In EN, YouTube (23%) and x.com (20.4%) take the top two spots. English moonstone digests social content β French moonstone stays anchored in the press.
What this changes: for an FR publisher, moonstone is THE visibility lever. For an EN publisher, it's one lever among several β and not the most powerful.
FR: 72 AIO hits in 3 months. EN: 1.1% of volume in a dedicated pipeline, 29% in mustntmiss.
The AI Overview β the AI-generated summary β has arrived in Discover. But only in English.
AIO content rate by EN pipeline. discover_ai_summary at 99.997%, mustntmiss at 29%. In FR: effectively zero.
The AIO source club is exclusive: Reuters, NYT, CNBC, Financial Times, Guardian. It's factual, structured, financial press. AIO doesn't democratize visibility β it concentrates it.
What this changes: for an EN publisher, AIO readiness is a real competitive advantage. For an FR publisher, it's not a priority today β but it's a weak signal to watch. What's in English today is often in French tomorrow.
FR: < 5% video in the feed. EN: three video pipelines totaling 13% of volume.
FR: all pipelines cluster near 0% video. The French feed is massively textual. Even freshvideos is only 53% video.
EN: a full spectrum. neoncluster at 100% video / 13% reach. freshvideos at 94%. creatorcontent at 72%. The video cascade is a structural fact.
The divergence is binary: the FR feed is a textual product with video accents. The EN feed is a hybrid text-video product with a three-stage video cascade.
Volume figures confirm it:
What this changes: an EN publisher without a YouTube video strategy is giving up 13% of feed volume. An FR publisher can focus on text without significant penalty.
FR creatorcontent: x.com 75%. EN creatorcontent: YouTube 72.4%. The perfect inversion.
The same pipeline β creatorcontent β captures social content for Discover. But the source is completely inverted between the two markets:
| Source | creatorcontent FR | creatorcontent EN |
|---|---|---|
| x.com | 75% | 23.2% |
| YouTube | 5.1% | 72.4% |
The inversion propagates through the cascade: FR freshvideos contains 47% articles (TF1, L'Equipe) and 53% video. EN freshvideos is 94% YouTube. The video cascade works because YouTube feeds the pipeline β in FR, x.com doesn't produce video content that feeds the cascade.
creatorcontent FR = 75% x.com. creatorcontent EN = 72.4% YouTube. The same pipeline, two opposite sources.
The implication is cultural as much as technical: in France, social content circulating on X (politics, sport, crime stories) feeds Discover. In English, it's YouTube (news, politics, entertainment). The same system, two cultural ecosystems.
What this changes: an FR content creator who wants to reach creatorcontent must be on X. An EN creator must be on YouTube. The entry platform into Discover differs by market.
FR: 36 hits. EN: 4.5% of volume, 13% reach, 100% YouTube.
This is the sharpest divergence. Neoncluster β the YouTube broadcast pipeline β is the third pipeline by reach in EN (13%). In FR, it doesn't exist. 36 hits over 3 months is statistical noise.
Why? Neoncluster is the third stage of the video cascade. It amplifies the best YouTube content filtered by freshvideos. In FR, the cascade doesn't work: creatorcontent is fed by x.com (not YouTube), freshvideos is half textual. The conditions for YouTube broadcast aren't met.
Neoncluster's growth in EN is staggering: 18x in 3 months. It's the fastest-growing pipeline in Discover, across all markets. Google is investing massively in YouTube/Discover integration β but only in English, for now.
What this changes: for an EN YouTube creator (news, politics, science), neoncluster is the route to 13% broadcast. This route doesn't exist in French.
FR: 24% reach. EN: 58.4%. A 2.4x gap.
The English feed is massively more monetized than the French one. More than half of EN devices see each ad.
| Metric | FR | EN | Ratio |
|---|---|---|---|
| Reach | 24% | 58.4% | 2.4x |
| Volume | 3% | 11.1% | 3.7x |
| YouTube in ads | β | 53.7% | β |
| Growth | 2.7x | 4.1x | EN accelerates faster |
YouTube is the source of 53.7% of EN ads β video advertising dominates. In FR, ads are mostly text links (hotels, fashion, SME e-commerce).
The advertising ecosystem is completely closed in both markets β 99.8% exclusive URLs. It doesn't interfere with the editorial feed. But its weight in the user experience is radically different: an EN user sees 2.4x more ads than an FR user.
What this changes: for an EN publisher, competition for attention in the feed is fiercer β editorial cards are diluted by more aggressive advertising. For an FR publisher, the feed is more "editorial."
FR: 3.7-day lifespan. EN: 2.5 days. French product content recycles more.
The shoppinginspiration pipeline has similar reach in both markets (FR 19.7%, EN 13.1%). The silo is the same β low co-occurrence with other pipelines. But lifespan diverges: French product content stays 48% longer in the feed.
Sources differ too:
The hypothesis: the EN product review market is more competitive β more fresh content pushes articles out faster. In FR, the tech/review content pool is smaller, so each article circulates longer.
What this changes: an FR tech publisher benefits from a longer shopping visibility window. An EN publisher must publish more frequently to maintain presence.
FR: 58% of URLs in 2+ pipelines, max 14. EN: 37% in 2+, max 12.
FR (blue) vs EN (orange). The FR tail is longer β some articles reach 14 pipelines. In EN, the maximum is 12. Multi-pipeline is a more French phenomenon.
Which EN pipelines share the same URLs? The video cascade (creatorcontent-freshvideos-neoncluster) forms a co-occurrence block unique to EN. EN multi-labeling is more structured by cascades; FR is more diffuse across editorial pipelines.
Multi-labeling is structurally more common in French. The reason: the FR feed is dominated by editorial pipelines (content, aura, moonstone, paginationpanoptic, relatedcontentruby, deeptrendsfable, mustntmiss) that share many URLs β a Le Monde article can traverse 6β8 of these pipelines.
In EN, the video cascade (creatorcontent β freshvideos β neoncluster) is a multi-labeling vector by design β but it only concerns YouTube content. EN text URLs traverse fewer pipelines than FR text URLs.
| Distribution | FR | EN |
|---|---|---|
| 1 pipeline | 42% | 63% |
| 2 pipelines | 20% | 19% |
| 3 pipelines | 13% | 9% |
| 4+ pipelines | 25% | 9% |
What this changes: the multi-pipeline strategy β publishing content that traverses the maximum number of pipelines β has more leverage in FR. In EN, the YouTube video strategy gives "free" multi-labeling (3 pipelines via the cascade).
The Premier League is systematically under-represented in 7+ EN pipelines. In FR: no equivalent exclusion.
The terms Premier League, football, Arsenal, Liverpool, Chelsea show a coherent negative signal in aura, deeptrendsfable, deeptrends, geotargetingstories, astria, freshvideos, and other EN pipelines.
Unaffected sports: NFL, NBA, Olympics, rugby, cricket, Formula 1. The exclusion is specific to the EPL.
Systematic under-representation of EPL terms in 7+ EN pipelines. Specific to the Premier League β other sports are unaffected.
In French, no equivalent sports exclusion exists. Football (Ligue 1, Champions League) circulates normally through trend, local, and diversification pipelines. The EPL is a case specific to the English market β likely tied to broadcasting rights.
What this changes: an EN sports publisher covering primarily the EPL has a structural ceiling in diversification pipelines. To maximize visibility, diversify toward other sports or other angles (transfers, business of sport, tactical analysis vs raw results).
FR: creatorcontent 33x. EN: neoncluster 18x. Both markets are evolving fast β but not in the same direction.
FR: creatorcontent explodes at 33x. paginationpanoptic at 7x. feedads at 2.7x. userpersonascontent in decline (-73%).
EN: neoncluster at 18x. creatorcontent and freshvideos at ~8x. feedads at 4.1x. userpersonascontent at 0.4x.
Growth directions tell Google's strategy:
| Direction | FR | EN |
|---|---|---|
| Social | creatorcontent 33x (x.com) | creatorcontent 7.8x (YouTube) |
| Video broadcast | Absent | neoncluster 18x |
| Advertising | feedads 2.7x | feedads 4.1x |
| Scroll infra | pagpan 7x | pagpan 4x |
| Personalization | userpersonascontent -73% | userpersonascontent 0.4x |
| Trends | deeptrends contracting | deeptrends 0.5x |
Both markets converge on one point: the gradual abandonment of userpersonascontent (the persona system) and deeptrends. Google seems to be migrating toward other personalization and persistence mechanisms.
They diverge on video: EN is accelerating massively toward YouTube broadcast (neoncluster 18x). FR is accelerating toward x.com social intake (creatorcontent 33x). Two different trajectories.
What this changes: tomorrow's Discover feed won't be today's. Continuous monitoring is necessary β the pipelines exploding in February 2026 didn't exist in December 2025.
FR: traditional publishers dominate. Ouest-France shows the widest multi-pipeline spread. Social and video columns are nearly empty.
EN: YouTube dominates the social/video column (49.9%). Quality UK press (Guardian, BBC) shows rich editorial spread. YouTube advertising dominates feedads.
The heatmap comparison speaks for itself:
The "multi-pipeline" model doesn't execute the same way:
The question everyone asks: will the two markets converge?
Convergence signals:
Divergence signals:
Our hypothesis: the markets will partially converge β AIO will arrive in FR, video will gain importance β but structural divergences (moonstone gap, social source, mono/multi-pipeline) will persist. Discover will remain two distinct products sharing the same infrastructure.
And that's precisely why continuous monitoring has value. Not to predict the future β to see it coming.
The feed evolves. The cards too. Like any good navigator, it's by keeping an eye on the horizon that you anticipate the currents.
Data: 42 million Discover cards, December 2025 to February 2026. Analysis: 1492.vision. Internal mechanisms are presented as our interpretations based on observed data and available public research.
Posted on 2026-03-28