The LinkedIn algorithm in 2026 is a ranking system that distributes content based on attention quality, not engagement volume. 360Brew — the 150-billion-parameter ranking model LinkedIn described in a January 2025 paper — reflects a broader shift away from simple engagement counting. It evaluates posts the way a human editor would: is this worth anyone's time?
Posts that genuinely hold a reader's attention, generate substantive comments, and consistently signal topic expertise reach far beyond a creator's follower count. Posts that don't, regardless of how many likes they collect, get suppressed.
Most of what you've read about the LinkedIn algorithm is still giving you advice from 2023. Post at 9am on Tuesday. Use 5-7 hashtags. Put your link in the first comment. These tactics either don't move the needle anymore or actively hurt reach.
The 360Brew shift — what changed and why
Until early 2025, LinkedIn's algorithm was a rough engagement counter. More likes and comments meant more distribution. That model had a well-known exploit: get your network to react quickly after posting, and the algorithm read it as quality. Engagement pods were a direct response to that mechanic.
360Brew broke the exploit. LinkedIn moved to a decoder-only model fine-tuned on its own professional data, described in a January 2025 paper. The model doesn't count interactions; it evaluates content much closer to how a person reads it.
The questions it asks:
- Does the post make a specific professional claim?
- Does it stay in a consistent topic area?
- Do the people engaging with it have relevant expertise?
The downstream effect on reach was significant. Average post reach declined sharply year-over-year across the platform. Reach now concentrates heavily at the top, and the best-performing posts vastly outperform the median — a gap that didn't exist at that scale before 360Brew.
One other structural shift explains why your posts stopped reaching your own followers. LinkedIn moved from a Relationship Graph (show content from people you know) to an Interest Graph (match content to people who care about the topic, regardless of connection). A growing share of the feed is now driven by topic relevance rather than who you're connected to.
The five signals that actually drive reach now
LinkedIn's ranking broadly runs in stages — an initial quality filter, an early engagement test, then relevance and expertise ranking. These are the signals that matter across those stages.
1. Dwell time. How long a reader stays is the signal the ranking model is built around — expanding "see more", swiping a whole carousel, finishing a video. Pausing on a post without clicking anything still counts in your favour.
2. Meaningful comments. A like is the weakest signal available: it costs a reader nothing and tells the model nothing about whether the post was worth reading. A comment that adds a perspective is worth substantially more, and a thread where several readers trade replies is worth more again. Five real replies beat ten posts with 30 likes each.
3. Saves and sends. LinkedIn added both to the post analytics panel in late 2025 — signalling in plain sight what it values. A save is the strongest single signal a reader can give you, and a private send is close behind. Both come from durable reference value: a framework, a process, a contrarian take backed by a real outcome.
4. Creator expertise signals. The system rewards consistency of themes over time, and cross-references your profile positioning — job title, about section, featured work — against what you post about. Misalignment reduces distribution; alignment earns a credibility multiplier.
5. Automated quality filtering. Before any human sees a post, an NLP filter sorts it into spam, low quality, or clear. Spam catches excessive hashtags, repetitive promotion, context-free links. Low quality catches unedited-AI patterns: generic openers, no personal voice, flat sentence-length variation.
That last filter runs before the golden-hour test. A post that fails it is suppressed before your network ever gets the chance to save it.
What the algorithm actively suppresses
| Behaviour | What happens | What to do instead |
|---|---|---|
| External links in the post body | Cuts reach. LinkedIn ranks off-platform traffic below content that keeps readers on LinkedIn, and the old first-comment workaround is detected and penalised too. | Keep most posts self-contained. Where a link is genuinely necessary, put it in the body and accept the reach cost. |
| Engagement pods and coordinated activity | LinkedIn detects clusters of accounts that reliably engage with each other within minutes of publishing. The detection is pattern-based: a tight, repeating cadence reads as coordinated rather than organic. Flagged accounts lose reach; repeat offenders face harsher, lasting penalties. | Nothing substitutes for it. Earn replies from people who have an actual opinion about the claim you made. |
| Low-dwell-time AI content patterns | Suppression here works through the reading signal, not an AI detector. What gets measured is that nobody finished the post — viral-template patterns like the one-line hook opener, the predictable three-paragraph rhythm and the "Agree?" close are recognisable in two seconds, so readers scroll. | Vary structure and sentence length, and lead with something only you could have written. The suppression here arrives through the engagement signal, not an AI detector. |
The insight the mega-blogs can't give you
Every major piece of LinkedIn algorithm advice ends in the same category of recommendation: post at 9am, use 3-5 hashtags, put links in comments, reply within 60 minutes. Buffer and Sprout Social are good at that advice. They're writing for brand accounts and marketing teams, where tactical timing still matters somewhat.
What nobody else is saying: in 2026, gaming the algorithm and sounding like a credible human are the same thing. That's the actual logic of 360Brew.
The version of this with no shortcut in it: the algorithm now rewards exactly what makes for good professional writing. Specific claims. Real experience. Consistent expertise. A recognizable voice. These are also the things that are hard to fake at scale.
For a practical guide to building writing habits that hold up to this standard, read how to write LinkedIn posts that don't sound like AI.
What this means practically for a busy professional
You don't need to post more. You need more signal per post.
Own 1–2 adjacent topics. Topic authority builds over weeks within a theme. One post a week in a consistent niche outperforms three a week across five unrelated topics. The worked examples in LinkedIn post ideas for software engineers show how one domain becomes a steady stream of specific posts.
Write for the "see more" click. LinkedIn hides everything past the first couple of lines behind a "see more" link, so the sentence sitting at that cut is the most important one in the post. Draft the opening last, once you know what the post actually delivers.
Accept the external link tax. If a link belongs in the post, put it in the body and accept the reach reduction. If it doesn't belong, don't add it. The "link in first comment" workaround is gone.
Prioritise depth over frequency. One post earning 30 seconds of read time and five substantive comments does more for your reach trajectory than five posts scanned in 2 seconds each. A hard reframe if you've been told to post daily, but the data is consistent.
Skip the viral templates. Hooks like "I just realized…" or "Hot take:" or a solitary number on a line are the patterns the NLP filter flags. They also tell human readers you're posting content rather than thoughts.
For individuals holding a consistent presence without losing hours to it, a tool that keeps your writing style intact across drafts removes the biggest practical barrier. ThoughtFuel gives you up to five Writing Style profiles trained on writing samples you provide, so drafts start from your actual voice rather than a blank slate.
If you're evaluating options, our head-to-head comparisons with Taplio and AuthoredUp cover where each one fits.
The LinkedIn algorithm rewards what good writing already does. Invest there, and the distribution follows.
FAQ
Does posting time still matter in 2026?
Timing is a minor secondary signal, not a primary driver. The consensus remains Tuesday through Thursday, late morning or late afternoon in your audience's timezone — full data tables in our guide to the best time to post on LinkedIn.
For an individual posting once or twice a week, voice and depth outweigh timing by a wide margin: a well-written post at 7pm Friday reaches further than a template post at 9am Tuesday.
How many hashtags should I use?
Three to five, used as topic categorisation rather than reach amplification. LinkedIn demoted hashtags from a reach driver to a lightweight classification signal in 2025. They help the algorithm confirm your topic area; they don't boost distribution on their own. More than five starts to look like spam to the quality filter.
Can LinkedIn detect AI-generated posts?
Not directly. LinkedIn's NLP classifiers detect patterns associated with unedited AI output — generic sentence structures, viral-template formatting, low variation in sentence length — because those correlate with near-zero dwell time.
The suppression is engagement-based: an AI-patterned post gets seen, generates no real attention, and distribution stops. A well-edited AI-assisted post that reads like you, with specific claims and natural variation, doesn't trigger those classifiers.
Is the “link in first comment” trick still safe?
No. LinkedIn now detects the workaround and applies a reach penalty, though typically smaller than the penalty for an in-body link. If your goal is maximum reach, keep links out of both the post body and the first comment, and put them in a later reply if needed. For most posts, the content should be self-contained.
Do engagement pods still work in 2026?
No. LinkedIn detects coordinated engagement: groups of accounts with consistent mutual engagement in a narrow window after posting. It suppresses reach for flagged accounts, and repeat offenders face harsher, lasting penalties. The detection is good enough that this is no longer a usable strategy.
What actually generates meaningful comments?
Specific, contestable claims. A general observation ("leadership is about trust") generates agreement in two words. A claim with evidence ("we reduced onboarding time by 40% by cutting the kick-off call entirely") gives readers something to answer at length.
A meaningful comment requires content that needs more than a reaction. That is what the algorithm is measuring: whether your content generates genuine professional exchange.