Why LinkedIn's New Algorithm Makes Warm Outreach Your Best Bet

In late 2024 and through 2025, LinkedIn replaced its entire content ranking infrastructure with a single AI model called 360Brew, a 150-billion-parameter decoder-only foundation model that handles feed ranking, job recommendations, content distribution, and connection suggestions through one unified system. Where the old algorithm counted hashtags, measured follower graphs, and tallied likes, 360Brew reads the text of each post, the metadata of each profile, and the context of what each viewer engages with, then generates a relevance score from semantic understanding.

The practical result: organic reach dropped roughly 50% year over year, engagement fell 25%, and follower growth slowed 59% according to Richard van der Blom's 2025 Algorithm Insights Report analyzing 1.8 million posts across 58,000 profiles. Who posted something matters less than what it's about and whether it earns sustained attention. A post from a stranger with demonstrated expertise on a topic you care about can now outrank a post from someone you've worked with for years.

For sales teams and founders running LinkedIn outreach, this shift changes the playbook. Cold connection requests and templated DMs were already getting deleted on sight. Now the algorithm actively penalizes engagement bait, external links, and AI-generated templates while rewarding content that earns saves, sends, and dwell time. The path that works: comment-led outreach that builds familiarity before the connection request ever lands.

What 360Brew changed about how LinkedIn evaluates content

Every LinkedIn post in 2026 passes through three sequential gates. First, a quality filter classifies content before any human sees it. Posts that resemble templated AI output, engagement-farming prompts, or spam patterns get throttled at this stage and never reach a real audience. The model handles relevance scoring across feed, search, and recommendations through the same unified system, so the quality gate's rules are consistent everywhere your content surfaces.

Posts that clear the gate enter a velocity test. The algorithm pushes each one to a small slice of your followers and watches what happens. Strong early engagement, especially comments and saves, signals to 360Brew that the content is worth distributing further. Roughly 70% of a post's total reach is decided in the first 60 to 90 minutes.

The third phase is the long tail. In mid-2025, LinkedIn tested an algorithm change that aggressively surfaced two to three week old posts in feeds. After user complaints about staleness, the company walked back the most aggressive version in July 2025 but kept what LinkedIn spokesperson Bhairavi Jhaveri called a little bit of a flex on recency to prioritize relevance when appropriate. Evergreen content with strong velocity and saves can still pull impressions for weeks after publishing.

Which engagement signals carry the most weight under the new algorithm

Saves and sends carry more algorithmic weight than likes. A save is roughly 5x the weight of a like and 2x the weight of a comment, based on independent analyses of post performance under the new system. Comments still outweigh likes, commonly cited at 5 to 15x. Dwell time, the seconds someone spends reading your post, is a top ranking factor. Likes are the weakest signal of the four.

Dwell time is the clearest illustration of how the system has shifted. LinkedIn's engineering team has published research on how they use dwell time as a ranking signal. A fast like is a noisier indicator of value than a long read. The percentage of feed posts hitting a 30-second dwell threshold varies sharply by content type. The algorithm pushes posts that earn sustained attention wider than posts that get a like and scroll.

For content designed to support outreach, this means writing for saves rather than surface engagement. The posts that get bookmarked are the ones with frameworks, specific stats, step-by-step breakdowns, and templates. They're screenshot-worthy in a literal sense, the kind of thing readers want to reference later.

How this affects connection acceptance and reply rates

Across 180,155 LinkedIn connection requests given at least 30 days to mature, 27.1% were accepted, 27.6% of accepted connections produced a reply, and 2.0% of accepted connections produced a booked meeting, according to research published by Reachium in August 2026. Replies were 7.5% of all matured requests sent, and booked meetings were 0.54%.

Those numbers are cold outreach baselines. The distance between cold and warm shows up in separate research from Belkins covering 14,077 contact records tracked from invite to connection to reply to booked meeting. Requests with a personalized note saw a 25.3% connection rate and 8.2% reply rate, while requests without a note saw 27.6% connection and 5.3% reply. Blank requests get accepted slightly more often because there's no friction and nothing to reject before clicking accept, but the downstream engagement tells a different story. People who accepted a note-accompanied request are 55% more likely to reply once the conversation starts.

The interpretation is straightforward: a note acts as a filter. Prospects who accept despite knowing what the outreach is about are self-selecting into engagement. Those who accept a blank request may simply be LinkedIn-active people who accept connections broadly and have no particular reason to reply when a sales message arrives afterward.

Why comment-led outreach works better under 360Brew

The old LinkedIn algorithm prioritized content from your connections, which made network size a growth lever. The new algorithm prioritizes content tied to your professional interests, regardless of who posted it. Your profile is now part of the ranking input. 360Brew reads your headline, About section, experience, and posting history as a credibility signal before deciding how to distribute what you publish.

This is the structural reason comment-led outreach converts better than cold requests. When you comment daily on a prospect's posts before sending a connection request, you're doing two things the algorithm now rewards. First, you're creating repeated exposure in their feed and notifications, building familiarity through what psychologists call the mere-exposure effect. Second, you're signaling to 360Brew that you share a professional interest with that person, which increases the likelihood your profile and future posts will be surfaced to them organically.

By the time the connection request lands, it's no longer cold. The prospect has seen your name multiple times, read your comments, and likely visited your profile. The request itself carries context. In our experience, a warmed connection request converts 3 to 5x better than a cold one, and the reply rate after acceptance is higher because the conversation already started in the comments.

What the algorithm penalizes in outreach activity

Five behaviors trigger distribution penalties under 360Brew: AI-templated content, engagement bait, automated pod engagement, external links in body, and hashtag stuffing. AI-templated content gets throttled at the quality gate before reaching a real audience. The fix is specific point of view plus concrete detail, the stuff a generic LLM output won't produce by default.

Engagement bait, the comment YES and tag someone prompts, is pattern-matched as low quality and suppressed. Comment pods or automated engagement tools result in detected automated comments capped in visibility, often below the post's organic baseline. Posts with external links in the body see roughly 60% reach reduction versus identical link-free posts, and the first-comment workaround has been penalized since early 2026.

Hashtag stuffing reads as spam. The algorithm now uses semantic intent rather than hashtag matching, so stacking five or ten hashtags at the end of a post does nothing. If you use them, cap at three relevant hashtags maximum and place them inline rather than in a block at the end.

How to work with the algorithm when running outbound

Engage with 5 to 10 posts in your feed for 10 to 15 minutes before you publish. Post. Then stay active for 90 minutes: reply to every comment within 15 minutes, engage with others' content, and watch your engagement velocity. Pre and post engagement around publishing can lift reach by up to 20%, which compounds against the algorithm's velocity test.

The LinkedIn algorithm uses your recent activity as one input to decide who's likely to see your post in the test phase. If you've been engaging in the right communities just before publishing, the initial test audience is warmer. If you've been silent for hours, you're starting cold. Replying to comments within 15 minutes pulls each commenter back into the thread and extends total dwell time on the post, which the algorithm tracks.

The operational layer that turns this into pipeline: track who engages with your posts and run personalized follow-up within 24 to 48 hours. Someone who commented on your framework post, saved your benchmark data, or spent 45 seconds reading your case study breakdown is warmer than a random profile from a target account list. The comment-led approach works in both directions. You comment on their posts to build familiarity before connecting, and you act on their engagement with your posts to open conversations while the context is fresh.

The formats that perform best for thought leadership and outbound in 2026

Document posts, the multi-page PDFs that replaced native carousel uploads in December 2023, hit 6.6% average engagement, the highest of any LinkedIn format. They outperform video by 278%, text-only posts by 596%, and single images by 303%. Native video, 30 to 90 seconds, comes in second at 5.6% engagement. Standard text posts struggle to break 2% unless they deliver genuine value with strong structure.

Document posts work because they earn higher dwell time with each swipe, they're mobile-optimized for maximum screen real estate, they're save-worthy when they contain frameworks or data people want to reference later, and LinkedIn measures completion tracking to see how many slides people view. Best practices: keep it 5 to 10 slides, use 1080x1080px square or 1080x1350px portrait, export as PDF for cross-device consistency, start with a strong hook slide stating clear value, one main point per slide, and end with a clear call to action.

For B2B outreach, the highest-impact content combines thought leadership on personal profiles with comment-led engagement on prospects' posts. The algorithm rewards specialization over broad-spectrum posting. If you post about ten different topics, the model can't decide which audience to show your content to, and your reach suffers. Posting across ten themes signals to 360Brew that you're an authority on none of them. Topic discipline is now an algorithm strategy in its own right.

Why personal profiles dominate consumption over company pages

Personal profiles dominate consumption on LinkedIn. For B2B teams, the highest-impact investment sits in executive and employee LinkedIn presence rather than the brand page. The audience-first approach, building a following through organic content on personal profiles before amplifying with paid, is now compounded by the algorithm's structural preference for individual creators over brand pages.

Company pages should post 3 to 5 times per week. More frequent posting causes your content to compete with itself in followers' feeds. Personal profiles have more flexibility and can post daily if quality stays high. The critical mistake is posting just to stay active. The 2026 algorithm aggressively penalizes low-quality content. One valuable post per week outperforms five forgettable ones.

Investing in your founders and salespeople as creators is the strongest distribution move available on the platform in 2026. When they comment on prospects' posts, the engagement carries individual credibility rather than corporate noise. When prospects comment on their posts, the reply comes from a person with a visible point of view rather than a brand voice. The connection request that follows lands warm because the relationship already started.