Why LinkedIn's 360Brew Algorithm Is Killing Generic Outreach

Send a connection request today that reads like a template, and 360Brew will quietly throttle it before your prospect ever sees the notification. That's the shift most outbound teams haven't absorbed yet: the model LinkedIn uses to rank your feed is the same model that now decides whether your outreach surfaces, gets demoted, or dies silently in the queue.

The acceptance rates we've watched collapse across agency books since March aren't a copy problem in the way SDR leaders usually mean. The copy is being evaluated against the sender's profile, recent activity, and network graph position — all at once, in natural language, by a 150-billion-parameter model. Generic no longer means low-performing. Generic means flagged.

This post is a diagnostic and a rebuild. We'll walk through what 360Brew actually scores on a connection request, why your engagement-to-outreach ratio is now the single most important sender metric, and the specific template rewrites that stop pattern-matching as spam. If you want the architecture-level view of how this model works across surfaces, we covered that in the 360Brew cloud-vs-extension teardown — this one goes deeper on the sender-content cross-check.

What the linkedin 360brew algorithm actually scores before your message is delivered

360Brew is a decoder-only foundation model developed by LinkedIn's FAIT (Foundation AI Technologies) team, detailed in a research paper published on arXiv in January 2025, and unveiled as the backbone of the feed overhaul on March 12, 2026 — a 150-billion-parameter model that replaced the fragmented recommendation machinery built over the years.

The critical architectural detail for outbound: 360Brew is designed to support many recommendation surfaces, not only the feed — content recommendations, "people you may know," job matching, and more — replacing a fragmented system of many specialized models with a unified approach. People You May Know is the ranking surface that governs connection request visibility. The same model scoring your posts is scoring your outreach.

And because decoder models with a textual interface can generalize across tasks by verbalizing member behaviors and their social connections in natural language, 360Brew doesn't look at your message in isolation. It reads your profile, your last 30 days of activity, your mutual connections, and the message body as one structured prompt — then predicts how likely the recipient is to engage. Low predicted engagement equals demoted delivery.

That's the mechanic. Everything downstream — template fatigue, acceptance decay, the "my copy stopped working" complaint — traces back to this one scoring pass.

The Interest Graph shift: why your network size no longer buys reach

For years, big networks meant big distribution. That lever is now largely broken. The numbers from Richard van der Blom's 2026 Algorithm Insights Report — a 200-page annual study now in its seventh edition, built on 1.3 million posts and more than 1,400 hours of research, cited across the industry when claims about reach are made — are stark.

Connection-based distribution fell from 70% in 2024 to 54% in 2025 to 10% in 2026, while the Interest Graph climbed from 5% to 21% to 50%. Reach on LinkedIn now follows interest, not your connections.

For outbound, this reads two ways:

1. A 10,000-connection profile targeting prospects outside its topical lane will see worse acceptance than a 1,500-connection profile that posts and engages in the same niche as its targets.
2. The "warm up by mass-connecting" tactic — long a staple of agency onboarding — now actively degrades sender trust, because it builds a network that doesn't match the sender's content fingerprint.

The aggregate reach damage from this shift is real. Van der Blom and his team analyzed over 1.3 million posts from 50,000 creators to produce the 2026 report, and reach is down 60% for active creators over the last two years. Senders who built outreach volume on top of broad, incoherent networks are now the biggest losers.

The message patterns 360Brew flags: a before/after teardown

Templates aren't dead. Pattern-matched templates are. Here are three structures our team has seen trigger silent demotion in agency books over the last 90 days, with the rewrites that restored acceptance.

Pattern 1 — The compliment-then-pitch opener

Before: "Hi {firstName}, loved your profile and your work at {company}. We help {industry} teams scale pipeline through…"

This matches a template fingerprint 360Brew has seen tens of millions of times. The sender's profile almost never contains evidence they actually read the recipient's work (no comment, no profile view, no shared post).

After: Reference a specific post the prospect published in the last 14 days, by paraphrase not copy-paste. Do not pitch in the connection request. Pitch in message 2, after a reply.

Pattern 2 — The fake mutual-interest hook

Before: "Saw we both work in B2B SaaS — would love to connect."

360Brew cross-checks the claim against both profiles. If the sender's recent activity shows no SaaS-adjacent engagement — no comments on SaaS posts, no SaaS-topic content — the claim reads as a template token. Demoted.

After: Build 7-10 days of genuine topic activity before sending to a given ICP. The model needs the sender's content fingerprint to match the hook.

Pattern 3 — The role-pitch to a C-suite target

Before: Direct outreach to the CRO or CEO with a demo ask.

Two problems. First, 360Brew weighs recipient response probability heavily, and C-suite reply rates are structurally low. Second, our data shows the fix: directors and VPs reply 2-3x more than CROs, and the model learns that pattern faster than most SDR teams.

After: Multi-thread. Open with the director, name the CRO as a secondary stakeholder in message 3.

The engagement-to-outreach ratio: the sender metric that matters now

This is the number almost no SDR dashboard tracks. 360Brew — and the broader enforcement layer sitting on top of it — looks at the ratio of your genuine engagement (comments authored, meaningful reactions on others' posts, replies to inbound messages, profile views you initiated) against your outbound volume (connection requests sent, cold messages sent).

Accounts with high outbound and near-zero engagement look exactly like what they are: outreach rigs. Accounts with a healthy ratio read as humans doing their job.

We've been tracking what a safe ratio looks like across agency books. Here's the current working table — these are operational targets, not published LinkedIn thresholds:

| Daily sends (connection + msg) | Minimum weekly engagement actions | Profile posts per week | |---|---|---| | 10-15 | 15-20 (comments on ICP posts) | 1 | | 15-20 | 25-30 | 2 | | 20-25 (ceiling) | 40+ | 2-3 |

A few teams push past 25 daily sends with aged, well-engaged profiles — but it's rare and the margin for error is thin. We dug deeper into the engagement side of this scoring in the detection post on engagement-to-outreach ratios.

The mechanical fix: block 25-30 minutes a day for commenting on target-ICP posts. Even Richard van der Blom notes that 80% of the comments he receives in the first five minutes of any post are written by AI — which means thoughtful human comments stand out further than they did two years ago, both to recipients and, apparently, to 360Brew.

Profile optimization that 360Brew actually rewards

Because the model reads your profile as part of the message-scoring prompt, profile and sequence are now one surface. Three changes move the needle:

Topic coherence. 360Brew builds a semantic fingerprint of what you post, comment, and work on. If your headline says "RevOps consultant," your last 10 posts are about pricing strategy, and your connection requests target VPs of Finance about pricing — the model has clean signal. If your posts are a mix of hiring takes, SaaS hot takes, and the occasional pet photo, the signal collapses.

Recent activity density. Dormant profiles sending outreach are now a bright red flag. The model wants to see the sender being a LinkedIn member, not just operating one. Two comments and one post per week is a reasonable floor for any account doing meaningful outbound.

Headline specificity. Generic titles ("Founder" or "Sales Leader") give the model nothing to anchor to. Specific titles with the ICP named ("Helping Series B RevOps teams fix pricing attribution") give 360Brew a semantic match it can score against your message body.

For solo operators and consultants, this is where the 14-day warm-up protocol earns its keep — the first two weeks are entirely about building the sender fingerprint before any outreach fires.

The 7-day sender warm-up for a flagged or new account

If your acceptance rate has dropped below 20%, or you're spinning up a fresh profile, this is the sequence we run before touching send volume:

• Day 1-2: Rewrite headline and About to one specific ICP + outcome. Post one short observation about your niche.
• Day 3-4: 10-15 thoughtful comments on ICP posts (not emojis, not "Great post!"). Accept inbound only.
• Day 5: Second post. Reply to every comment within the hour.
• Day 6: 5 connection requests, all to people whose content you've engaged with that week. Zero pitch in the note.
• Day 7: Review acceptance rate. If above 40%, scale to 10 the following day. If below, extend the warm-up another 7 days.

For agencies running this across client books, the protocol is the same but parallelized — and the operational lift is why we published the whitelabel 24-hour onboarding stack separately.

Sending cadence that respects the new scoring layer

The cadence math has changed. Pre-360Brew, you could spike to 40-50 daily sends on an aged account and absorb the throttling. Post-360Brew, the throttling isn't a soft cap — it's a scoring input that compounds. Your next send gets scored partially on how your last 30 did.

Operational rules we're holding teams to:

• Cap connection requests at 20-25/day on established accounts, 10-15 on anything under 90 days of activity.
• Space messages at least 60-90 minutes apart; never fire in parallel batches.
• Keep follow-up sequences to 3 touches maximum within a 14-day window. Four or more now reads as automation signature.
• Pause outbound entirely for 48 hours after any account warning or unusual throttling.

The teams keeping reply rates in the 8-12% band aren't sending more. They're sending less, into tighter ICPs, with sender profiles that match the message content on 360Brew's reading of both.

• 360Brew scores sender profile + message + recent activity as one prompt. Generic templates fail because the model sees the mismatch between what you claim and what your profile proves.
• The Interest Graph now drives 50% of distribution, connections only 10% — topic coherence between your posts and your outreach ICP matters more than network size.
• Engagement-to-outreach ratio is the new sender trust metric. Target 15-40+ weekly engagement actions depending on send volume.
• Rewrite templates to remove compliment-pitch openers, fake mutual-interest hooks, and C-suite direct asks — these are the three patterns we see demoted most reliably.
• Cap sends at 20-25/day on aged accounts, 10-15 on new ones, and run a 7-day warm-up before scaling any flagged profile.