LinkedIn Algorithm & 360Brew

Last updated: Wednesday, June 03, 2026  |  Competitive Intelligence Scan

Weekly Scan Fresh
Freshness note: No platform announcements found in the 7-day window (May 27–June 3, 2026). Most recent sourced intel dates to late May 2026. All findings marked accordingly. No stale content padded in.

Key Metrics

Organic Reach Decline YoY
~50%
Intentional — 360Brew prioritizes relevance over broadcast [3]
Save vs. Like Signal Weight
A save drives 5× more reach than a like; 2× more than a comment [4]
Company Page Reach Drop
60–66%
Since 2024; personal profiles vastly outperform [5]
360Brew Model Size
150B
Parameters; built on LLaMA 3, trained on LinkedIn proprietary data [6]

360Brew Watch

No new 360Brew model updates announced in the last 7 days. The deployment reference period is March 12, 2026 (official LinkedIn Engineering Blog post by Hristo Danchev). Ongoing rollout continues through A/B testing waves.
Confirmed Recent — Apr 21, 2026 Official Deployment Date: March 12, 2026
The public announcement of 360Brew's deployment in the feed was made on March 12, 2026, via the article "Engineering the next generation of LinkedIn's Feed" authored by Hristo Danchev on the LinkedIn Engineering Blog. At the core of the change: 360Brew, a 150-billion-parameter AI model that replaces the fragmented recommendation machinery built over the years.
Confirmed Recent — Apr 21, 2026 Architecture: LLaMA 3-based Decoder-Only Transformer
360Brew is a decoder-only foundation model, developed by LinkedIn's FAIT (Foundation AI Technologies) team, with 150 billion parameters trained exclusively on LinkedIn's proprietary data (profiles, posts, professional interactions, job descriptions), with architecture derived from Meta's LLaMA 3 family.
Confirmed Recent — Apr 13, 2026 Momentum Model Replaces the "Golden Hour"
The 2026 algorithm replaced the "Golden Hour" with the "Momentum Model." LinkedIn now evaluates content over a much longer, more forgiving 3–8 hour evaluation phase. The platform tests your post with a small sample of your audience. If it maintains a high Depth Score and consistent dwell time over several hours, the AI will push it to second and third-degree connections — meaning a post might start slowly in hour one but achieve viral distribution by hour eight.
360Brew Technical Architecture Details
Confirmed Background — Jan 23, 2026 30+ Predictive Tasks in One Model
LinkedIn's engineering team published research describing 360Brew as capable of handling over 30 different predictive tasks that previously required separate systems maintained by dedicated teams. The platform essentially rebuilt its entire recommendation infrastructure around a single AI model that comprehends professional context.
Confirmed Background — Apr 21, 2026 Generative Recommender Analyzes 1,000+ Past Interactions
The 2,000 shortlisted posts pass through the Generative Recommender (GR), a transformer model that analyzes more than 1,000 of your past interactions as a chronological sequence — not as a bag of independent events. The old system scored each post in isolation. The new one understands your trajectory: if you read three articles about B2B marketing this week, then looked up growth hacking, then saved a post about lead generation, it infers where your interest is heading.

Algorithm Signals

Boosted
Penalized
Depth Score — Full Signal Breakdown
Confirmed Recent — multiple sources Three-Component Depth Score
Depth Score is calculated from: (1) Reading Time (Dwell Time) — how long users spend reading or viewing your content, the primary component; (2) Comment Depth — not just how many comments, but how substantive the discussions are, with multi-reply threads carrying significantly more weight; and (3) Saves for Later — a strong quality signal indicating the content has lasting reference value.

Content Format Performance

Sources: Buffer (45M+ posts, Mar 2026), meet-lea.com (Feb 2026), GrowLeads analysis (~2 weeks ago), PostEverywhere (~2 weeks ago), SuccessKnocks (June 2, 2026)

Format Avg Engagement Rate Reach Profile Dwell Time Best For
Carousel / Document PDF 21.77% median (Buffer) / 24.42% avg [10] High 15–20 sec Frameworks, how-to, saves
Multi-image post 6.60% [11] Medium-High Moderate Visual storytelling
Video (native, <60s) 6.47% avg; +53% for short clips [11] Medium Varies Trust, personality, cold reach
Text only ~4% [11] High (burst) 8–10 sec Opinions, stories, comment volume
Poll 4.40% (+2× since 2023) [11] Medium Low Market research only
Single image 4.85% (underperforms text) [8] Low Low Brand awareness (limited)
Format Detail Notes
Recent — ~2 wks agoCarousel Optimal Slide Count: 8–12
The sweet spot is 8–12 slides. Fewer than 5 slides often underperform because they don't earn enough swipe completion to justify algorithmic distribution. More than 15 slides can lose viewers before completion, hurting the dwell-time-completion-rate signal.
Recent — ~2 wks agoRecommended Weekly Mix
The strongest LinkedIn strategies split roughly 60% text posts (volume) and 40% carousels (depth). Accounts rotating between formats achieve 37% more follower growth and 28% more consistent visibility compared to accounts using single formats repeatedly.
Recent — Jun 2, 2026Document Posts Lead on Engagement Rate
Native documents lead with average engagement rates hitting 7.00% — a 14% year-over-year jump. Multi-image carousels follow closely at around 6.6–6.8%. Short native videos deliver strong results when kept under 60–90 seconds.
Recent — Mar 2026Poll Fatigue: Reach Up, Engagement Down
Polls are in a mixed place: reach is actually up (polls stay visible longer) but engagement is significantly down. It's likely due to overuse — poll fatigue is real. Run them for 7 days; 1-day polls incur a 70% engagement penalty.
Recent — Mar 2026Video: Subtitles Add +29% Engagement
Videos with subtitles achieve 29% higher engagement and 32% longer retention than those without. Always add captions — the majority of LinkedIn users watch video with sound off, especially during work hours. A video without captions is invisible to most of your audience.

Platform & Creator Program Changes

Confirmed Background — Oct 2025 Top Voice Badges Retired; Collaborative Articles Winding Down
In October 2025, Top Voice badges were retired and the Collaborative Articles feature began winding down. No replacement creator program announced as of June 3, 2026.
Confirmed Recent — Feb 2026 Lempod Banned from Chrome Web Store
Lempod was banned from the Chrome Web Store in February 2026, and flagged accounts face a 60-to-90-day shadow ban.
Confirmed Recent — Mar 12, 2026 (Authenticity Update) Engagement Bait & Automation Pods Officially Killed
The March 2026 "Authenticity Update" officially killed engagement bait, legacy automation pods, and external link spam. The platform has fundamentally pivoted away from rewarding superficial, high-velocity reactions and is ruthlessly enforcing a new metric designed to measure true human attention.
Confirmed Recent — Apr 22, 2026 360Brew Deployment Continues in Testing Waves
360Brew deployment continues through testing waves in 2026. Semantic reasoning replaces traditional signal-based ranking for an expanding subset of users. External links with context see modest reach gains.

Opportunity Radar

01
Small Accounts Can Now Punch Above Weight
Smaller accounts with genuine expertise can now reach relevant audiences without first building massive followings — the AI identifies who would find content valuable by understanding topic, interests, and professional context. Build a 90-day topic cluster immediately.
02
Saves as a KPI — Underexploited Signal
The old algorithm used likes as a signal of social proof, whereas 360Brew interprets bookmarking as a stronger quality indicator. AuthoredUp's analysis of over 3 million posts found that a save drives 5× more reach than a like and 2× more reach than a comment. Shift content briefs toward "reference-worthy" frameworks and checklists.
03
Post Half-Life Now 2–3 Weeks
The half-life of all content is longer now than a year ago — regular engagement on old posts is something that rarely happened before 2025. Posts can now surface for 2–3 weeks, compared to days on other platforms. Repurpose evergreen carousels and re-engage with old comments to re-trigger distribution.

Source Log

  1. Yep Ads — "LinkedIn Algorithm Changes 2026: Why LinkedIn Reach Dropped" — May 2, 2026 Recent
  2. Try Ordinal — "How LinkedIn's Algorithm Works in 2026: Complete Guide" — [undated, ~2026]
  3. Falia — "360Brew: LinkedIn's New Algorithm Explained (2026)" — Apr 21, 2026 Recent
  4. Botdog — "5 Biggest LinkedIn Algorithm Changes in 2026" — Mar 16, 2026 Recent
  5. GrowLeads — "LinkedIn Algorithm 2026: Text vs Video Strategy Exposed" — ~May 20, 2026 (2 weeks ago) Recent
  6. DEV.to — "Reverse-Engineering LinkedIn's 360Brew From Their Engineering Blog" — ~May 2026 (1 month ago) Recent
  7. designACE — "LinkedIn Algorithm 2026: How Reach, Engagement, and Visibility Work Now" — Apr 8, 2026 Recent
  8. LinkBoost — "LinkedIn Algorithm Changes 2026: Beat the Depth Score" — Apr 13, 2026 Recent
  9. PostEverywhere — "LinkedIn Carousel vs Text Post: 2026 Format Guide" — ~May 20, 2026 Recent
  10. Meet-Lea — "LinkedIn Content Formats: Performance and Statistics 2026" — Feb 21, 2026 Recent
  11. Buffer — "Best Content Format on Social Platforms in 2026: 45M+ Posts Analyzed" — Mar 19, 2026 Recent
  12. SuccessKnocks — "LinkedIn Content Formats That Drive Engagement in 2026" — Jun 2, 2026 New
  13. First AI Movers — "LinkedIn's AI Shift Reveals How Platforms Now Reward Substance Over Scale" — Jan 23, 2026 Background
  14. Pettauer.net — "LinkedIn 360Brew and the New Physics of Visibility" — Jan 26, 2026 Background
  15. PostEverywhere — "How the LinkedIn Algorithm Works in 2026" — Apr 22, 2026 Recent

Lilypath Competitive Intelligence — For internal use only. Generated June 03, 2026.