Updated · March 2026

LinkedIn 360Brew Algorithm — what changed in 2026

Q: How can I tell if my profile is coherent?

Read your headline, your about, and your most recent experience in sequence. If the three pieces tell the same professional story, the profile has basic coherence. For a structured analysis of the 5 coherence dimensions, Karvi identifies where the signal is fragmented.

360Brew is LinkedIn's ranking model, deployed in March 2026. 150 billion parameters, it replaced the previous specialized models. The practical difference: the algorithm now evaluates semantic coherence across profile sections, not keyword counts. Scattered profiles lose reach; profiles with a consistent signal win.

Find out where your profile is inconsistent — free diagnosis

The model

What 360Brew is

360Brew is a 150-billion-parameter language model built by LinkedIn to unify ranking across the platform. It was described in a paper published on arXiv in January 2025 by researchers on LinkedIn's Foundation AI Technologies team, led by Hamed Firooz.

Before 360Brew, each ranking task (feed, job suggestions, connection recommendations, recruiter search) was handled by a different specialized model. Each of those models was developed and maintained by separate teams over several years. 360Brew replaced more than 30 distinct predictive tasks with a single foundation model.

In March 2026, LinkedIn confirmed the model's rollout to feed through the post "Engineering the next generation of LinkedIn's Feed" on its engineering blog. The practical shift: instead of counting keywords per section, the model reads the entire profile as text and checks whether the parts form a consistent professional identity.

Primary source — paper: 360Brew: A Decoder-only Foundation Model for Personalized Ranking and Recommendation — arXiv 2501.16450

Primary source — feed announcement: Engineering the next generation of LinkedIn's Feed — LinkedIn Engineering

Comparison

What changed in ranking

Swapping specialized models for 360Brew changed what the algorithm considers in each profile dimension. The table below summarizes the practical differences.

Aspect	Before (specialized models)	After (360Brew)
Profile evaluation	Keyword count per section	Semantic coherence across sections
Hashtags	Direct weight in ranking	Reduced weight — indirect topic signal
User activity	Evaluated in isolation by volume	Must reinforce the profile's identity
Recruiter search	Match by exact terms	Match by meaning and context

For a practical guide to optimizing the 6 profile dimensions (headline, about, experiences, skills, completeness, activity), see the full guide: How to optimize your LinkedIn profile in 2026.

Framework

How to evaluate your profile's coherence

360Brew evaluates coherence across 5 relationships between profile sections. Understanding each already shows where the signal is fragmented.

Headline ↔ About: aligned positioning?

The headline declares a positioning. The about should develop it with context and proof, not repeat the same text or introduce a different identity. If the headline says "Data Engineer · ML Infrastructure" and the about talks about leading sales teams, there's signal dissonance.

About ↔ Experiences: does the trajectory match?

The about describes a professional direction. The experiences need to show how you got there. An about focused on digital product but experiences describing only operations and logistics creates an incoherence that 360Brew detects as a fragmented signal.

Experiences ↔ Skills: do the skills match what you've done?

Declared skills that don't appear in any experience get reduced weight. The algorithm cross-checks what you claim to know with evidence of use in your experiences. Skills that appear both in the dedicated section and in role summaries carry more weight in semantic ranking.

Specialization: is there a clear thematic cluster?

Profiles with multiple unrelated areas (e.g., digital marketing + accounting + mobile development) create a scattered signal. 360Brew tends to rank profiles with a defined thematic cluster (a central area with related extensions) higher than generalists without a through-line.

Activity: do posts and comments reinforce the positioning?

Posts and comments on random topics dilute the signal. A developer who posts exclusively about software engineering builds a consistent "semantic fingerprint." Posting about cooking, motivation and tech in the same week fragments that signal.

Data from the State of LinkedIn in Brazil 2026 report show that the Skills section averages 3.1/10 across 1,998 analyzed profiles. It's the dimension with the highest rate of incoherence detected by Karvi.

Concrete actions

Practical implications of 360Brew

Five edits with the largest impact on semantic coherence. None require rebuilding the profile from scratch.

Rewrite the headline to reflect a single positioning

A headline with two or more unrelated positionings (e.g., "UX Designer & Financial Analyst") creates an ambiguous signal. Pick the primary positioning and build the headline around it.

Cut skills that don't appear in your experiences

Review your skills list and remove the ones that don't appear in any role summary or listed achievement. Skills without experience backing carry reduced weight in 360Brew.

Narrow the topic range in posts

Focusing on 2-3 topics related to your field creates a sharper semantic fingerprint for the algorithm.

Make sure the about and experiences tell the same story

Re-read your about and your most recent role. If someone read both in sequence, would they reach the same conclusion about what you do? If not, adjust the about to reflect the real direction shown in the experiences.

Update old role summaries to align with the current direction

Old experiences with descriptions that contradict your current positioning create semantic noise. You don't need to delete them: just rewrite the summary to highlight the parts most coherent with your current profile.

What to avoid

What no longer works with 360Brew

Practices that worked with the previous models are now penalized by 360Brew.

Anti-pattern	Why 360Brew penalizes it
Keyword stuffing in the about	Repeating "project management" or "leadership" ten times in the about used to work with count-based models. 360Brew penalizes artificial term density: the semantic signal collapses when the keyword concentration sounds mechanical.
A list of 50+ generic skills	Skills like "Communication," "Teamwork" and "Leadership" without backing in experiences are filtered by 360Brew as low-value signal. Quantity doesn't compensate for lack of specificity and coherence.
Hashtags in volume without thematic coherence	Adding 10+ hashtags to posts to maximize reach used to work with older models. With 360Brew, hashtags on topics unrelated to your profile's positioning dilute the semantic fingerprint instead of amplifying it.
Scattered posts on unrelated topics	Posting about tech one week, mental health the next, and recipes after creates a semantically undefined profile. The algorithm uses your activity history as an identity signal, and topic inconsistency generates noise.

Frequently asked questions

Questions about the 360Brew algorithm

When was 360Brew launched?

The research paper describing 360Brew was published on arXiv in January 2025. LinkedIn confirmed the model's rollout to feed in March 2026 through the post "Engineering the next generation of LinkedIn's Feed" on its engineering blog.

How does LinkedIn evaluate semantic coherence?

360Brew is a decoder-only model with 150 billion parameters trained on LinkedIn data. Instead of analyzing each section separately, it reads the entire profile as text and checks whether the parts form a consistent professional identity.

Do hashtags still work?

Hashtags weren't eliminated, but they lost direct weight. With 360Brew, what matters is the alignment between the hashtag and the profile's positioning, not volume. Using 2-3 hashtags directly related to your field has more impact than 10 varied ones.

Do I need to redo my entire profile?

No. The highest-impact edits are: rewriting the headline for a single positioning, aligning the about with the experiences, and removing skills without experience backing. Each of these can be done in under 30 minutes.

How much weight does the Skills section carry in the new algorithm?

Skills still carry significant weight, but with an additional criterion: they need backing in your experiences. Skills that appear both in the dedicated section and in role summaries have more semantic weight. Karvi's data on 1,998 profiles shows an average of 3.1/10 in Skills, the dimension with the worst average performance in Brazil.

How can I tell if my profile is coherent?

The simplest test: read your headline, your about, and your most recent experience in sequence. If the three tell the same professional story, the profile has basic coherence. For the 5 coherence dimensions broken down by section, Karvi identifies where the signal is fragmented.

Does 360Brew affect recruiter search the same way it affects feed?

360Brew consolidates more than 30 predictive tasks, including feed, candidate search, and job recommendations. On recruiter search, the model moved from exact-term matching to meaning-and-context matching. A recruiter searching for "growth specialist" can find profiles that use "user acquisition" and "conversion funnel," as long as the whole profile signals that positioning.

Next step

Find out where your profile is inconsistent

This guide explains how 360Brew evaluates coherence. Karvi identifies exactly where your profile loses signal: a diagnosis across the 5 coherence dimensions with concrete suggestions for each section.

Find out where your profile is inconsistent — Karvi analysisFree diagnosis — no credit card