What your post's reading level signals to LinkedIn's algorithm

360Brew reads your post the way a tired reader does: it estimates how much work each sentence costs. Posts written above a 10th-grade reading level give up roughly 35% of their reach. The mechanism is dwell time, not keywords. In our data, the gradient matters more than the threshold.

Reading Level Is a Direct LinkedIn Post Reach Signal

Posts written above a 10th-grade reading level lose roughly 35% of their reach compared with posts written at a 4th-grade level. That figure usually gets repeated as a threshold, as though grade 10 were a wall and everything below it were equally safe. It is not a wall. We correlated Flesch-Kincaid grade level against post impressions across B2B accounts in the SocialNexis dataset, and the shape we found was a gradient with a very uneven slope.

Posts at grade 7-8 do outperform posts at grade 10 and above. That part matches the conventional advice. But the largest single lift in our data sits between grade 7-8 and grade 5-6, not between grade 10 and grade 6. The practical consequence is counterintuitive for anyone who thinks of readability work as a cleanup pass. Rewriting a dense post into a merely professional one captures the smaller half of the available gain. The second edit, the one that feels almost too plain to publish, is where the reach lives.

This matters more now than it did two years ago because the baseline moved under everyone. LinkedIn organic reach fell roughly 50% year over year in 2026 following the 360Brew rollout. ContentIn's analysis of 100,572 personal-profile posts published between 2023 and August 2026 puts median impressions per post at around 412 to 860 depending on account size. When the median is that low, a 35% swing is not a rounding error on a good month. It is the difference between a post that gets read and a post that effectively did not happen.

Reading level is also one of the few signals in this entire system you can measure before you hit publish. Flesch-Kincaid grade level and the Gunning Fog index are both computable from the draft itself. Most word processors expose a readability score, and free calculators will take a pasted draft. Dwell time, save rate, and Stage 2 expansion are all outputs you learn about afterward, when it is too late to change anything.

One clarification, because this advice gets misread constantly. A low reading level is not a simple idea. It is a complicated idea carried by short sentences and common words. The failure mode we see most often in B2B accounts is what we call the dense-expert post: genuinely good thinking, delivered in stacked subordinate clauses with three abstract nouns per sentence. The thinking is fine. The delivery costs the author a third of their audience before anyone has decided whether they agree.

How 360Brew Reads Your Post Text for Content Ranking Factors

360Brew is a decoder-only foundation model of roughly 150 billion parameters, deployed across LinkedIn's feed ranking and recommendation surfaces in late 2025. The parameter count is the least interesting thing about it. What changed for writers is the input. Earlier ranking systems scored a post mostly by what happened to it after publication: likes, comments, clicks, the behavior of the people who saw it first. 360Brew reads the text itself and forms a view before the engagement data exists.

There is no reading-level penalty written into the model. Nobody at LinkedIn built a Flesch-Kincaid gate. The model infers reading difficulty from the features that produce it, sentence structure and vocabulary density, and uses those features to predict how long a reader will stay with the post. A low predicted dwell time becomes a lower distribution score. The penalty is real, the mechanism is indirect, and that distinction explains why readability advice framed as a style preference underperforms advice framed as a distribution input.

Profile-content alignment sits alongside it as a first-order signal. When your posts consistently match the professional expertise stated on your profile, 360Brew assigns a wider initial test audience, because the model has a confident guess about who should see your work. Alternating between unrelated topics narrows that Stage 1 distribution width. The model is not punishing range. It is admitting uncertainty, and uncertainty gets fewer impressions.

LinkedIn's own engineering blog has described using multitask learning across many simultaneous ranking objectives, which is the structural reason no single post attribute operates alone. Reading level interacts with topical authority, with format, with posting frequency. A grade 5 post from an account with no topical pattern does not perform like a grade 5 post from an account that has written about one subject for two months. Anyone promising you a single lever is selling you a simplification of a system that was explicitly built not to have one.

The most common implementation failure we see is the vocabulary swap. An operator runs through a draft replacing long words with short ones, checks the readability score, and finds it barely moved. The reason is that grade-level formulas weight sentence length as heavily as word length. If every sentence still runs three clauses deep, swapping utilize for use buys almost nothing. Cut the sentence in half instead. The score moves, and so does the dwell time it is standing in for.

The Dwell-Time Mechanism Behind LinkedIn Organic Reach

Dwell time is the number of seconds a user actively views a post in the feed, and it is 360Brew's highest-weighted distribution signal. Posts that sustain 30 or more seconds of active viewport engagement get amplified. Posts generating under 2% engagement in the first 60-minute window are deprioritized before Stage 2 distribution ever begins. Everything else in this guide is downstream of those two numbers.

The link between reading level and dwell time is mechanical rather than aesthetic, which is why it holds across industries and writing styles. A reader who processes a sentence in one pass keeps scrolling at a steady rate. A reader who hits a clause-heavy sentence either re-reads it or leaves. Re-reading sounds like it should increase dwell time, and occasionally it does, but in our data the far more common outcome is the exit. Scroll velocity is the thing 360Brew's dwell signal is actually sampling.

This is not only a LinkedIn phenomenon, and that independence is worth something. Peer-reviewed research on post-caption readability across social platforms found that lower Flesch-Kincaid grade level is associated with higher awareness metrics, including reach and impressions, with the Gunning Fog index showing similarly strong associations with engagement. That study was not written about 360Brew and predates it. When an academic finding and a proprietary ranking model point the same direction, the effect is probably in the reader rather than in the algorithm.

Our practical target for dwell time is grade 5-6, and the reason is specifically about marginal return. In the SocialNexis dataset that is where the reach gradient is steepest. Each grade level you remove in that range buys more impressions than each grade level you remove at the high end. If you only have the appetite for one editing pass, spend it getting from the sevens into the fives, not from the elevens into the eights.

There is a tempting shortcut here that does not work: buying dwell time with length. If 30 seconds of viewport time is the threshold, a longer post should hold the reader longer. What we see instead is the see-more trap, where a dense multi-paragraph post earns the expansion click and then loses the reader mid-scroll, producing the engagement signature of an abandoned post rather than a read one. Length without readability generates the click and not the seconds.

What Most LinkedIn Algorithm Signals Advice Gets Wrong

Most LinkedIn algorithm advice treats AI-generated content as a binary. Either a post is AI or it is not, and the model either catches you or it does not. That framing misses what 360Brew is doing. The model evaluates lexical diversity as a proxy for authenticity: human writers naturally vary vocabulary and sentence rhythm, while generated text falls into predictable patterns, repeated transitions and an unnaturally steady tone. Those patterns depress dwell time, and the downweighting follows from the dwell time rather than from a detection verdict.

Which means the interesting problem is not AI at all. It is structural predictability, and it is far more widespread than AI drafting. We have watched accounts where every post opens with a bold hook, moves to three bullet points, and closes with a question. The copy is entirely human. The thinking is good. The shape never changes.

In the SocialNexis dataset that pattern degrades engagement velocity over 4-6 weeks. Two things are happening at once. 360Brew's semantic model learns that this author's posts follow a predictable arc, and returning readers recognize the template before they have read a word. The second effect is the one that does the damage, because recognition without curiosity produces a scroll past, which produces no dwell time, which progressively shrinks the initial test audience the next post gets.

We call this template recognition, and it has a signature you can spot in your own analytics. Reach holds steady for a month or so, then declines across consecutive posts with no change in topic, length, or quality. Authors almost always misdiagnose it. They conclude the algorithm changed, or that their niche is saturated, and respond by posting more of the same shape more often, which accelerates the decline.

The fix is structural variety, and it is independent of reading level. Vary post length so consecutive posts do not occupy the same vertical space in the feed. Vary the lead sentence type: a flat claim, a number, a short story opening, a direct question. Vary paragraph count. None of this requires worse writing or a higher grade level. It requires that your last five posts not be interchangeable in silhouette. The most common bad advice on LinkedIn right now is find your format and stick to it, which is exactly the instruction that builds a template the model and your readers both learn to skip.

Do External Links Hurt LinkedIn Post Reach in 2026?

Yes. External links placed in the post body reduce reach by 50-70% under LinkedIn's current algorithm, with Forbes putting the figure at roughly 60% in July 2026. The reasoning is not mysterious. LinkedIn wants sessions to continue on LinkedIn, an outbound link reads as navigation intent, and the distribution penalty is the platform acting on that read.

An honest caveat, because the 60% number gets quoted with more confidence than it deserves. An independent analysis of 1.3 million posts measured the reach drop from a single external link in the body at 18.8%, which is a different order of magnitude. We treat the 50-70% range as a ceiling case rather than a typical one. The spread across studies probably reflects real variation by account type and link destination. The direction of the effect is consistent everywhere, so the operational conclusion survives the disagreement about its size.

The bigger change in 2026 was the workaround closing. LinkedIn's March 2026 Authenticity Update patched the first-comment link trick, flagging it as bridge behavior and applying the same reach penalty used for in-body links. That detail is missing from most of the algorithm guides currently ranking for this topic, which is a reasonable way to date their last real revision.

In our dataset the compression was sharp rather than gradual. Accounts that relied on first-comment links for demo booking CTAs saw reach-per-post compress within two weeks in mid-March 2026, which is the signature of a rule change rather than organic drift. Organic drift looks like a slow slope across months. This looked like a step. The signal LinkedIn is reacting to is not the presence of a URL string: it is the off-platform click-through behavior the model infers navigational intent from, which is why moving the same URL one comment down did nothing.

The practical answer is to stop fighting it. Move links to your profile section or a pinned post, and reference them in the post body as link in profile without putting a visible URL in the text. You will lose click volume. You will gain enough reach that the absolute number of clicks often holds. For B2B accounts where a single qualified reader is worth more than a hundred casual ones, that trade is usually favorable, and it is the one we recommend to accounts whose demo pipeline depended on in-feed links before March.

Saves Outweigh Likes: LinkedIn Feed Algorithm Engagement Signals Ranked

Saves carry approximately 5 times the algorithmic weight of likes in LinkedIn's distribution model, based on analysis of more than 3 million posts. That makes the save the highest-intent distribution signal LinkedIn tracks. A handful of saves will do more for a post's expansion than a much larger pile of passive reactions, because a save is a reader telling the platform they intend to come back.

To see why the weighting matters so much, you need the stage model. LinkedIn distributes in phases: Stage 1 runs from 0 to 60 minutes and shows the post to 2-5% of the author's network. Stage 2 runs from 1 to 6 hours and expands to 10-20%, but only if early engagement from Stage 1 exceeds 5-10% of the people who saw it. Every signal you earn in that first hour is being measured against a small denominator.

The consequence is that early engagement quality beats total engagement volume, and the gap is larger than most people assume. A post earning saves and substantive comments from a small group of genuinely relevant viewers clears the Stage 2 threshold. A post earning a broad spray of passive likes from a loosely connected network may not, even with a higher raw engagement count, because the composition of the signal is what gets evaluated rather than its size.

This is also why engagement pods stopped working. LinkedIn detects coordinated engagement with 97% claimed accuracy as of 2026 and penalizes it directly. The detection problem is easier than it sounds: pod engagement arrives in a burst, from the same accounts, with no accompanying reading behavior. Reactions with zero dwell time attached are a distinctive pattern, and the flag lands before Stage 2 distribution occurs, which means the pod does not just fail to help. It consumes your Stage 1 window.

Saves are a format problem more than a copy problem. What gets saved is a post a reader expects to need again: a checklist, a set of numbers, a process they will have to explain to someone else. In our experience the single most reliable way to raise save rate is to make a post referenceable rather than merely agreeable. Posts that provoke agreement earn likes. Posts that provide something reusable earn saves, and saves are what the ranking system is weighting.

Topical Focus and Posting Cadence Compound the LinkedIn Organic Reach Benefit

Accounts posting 3-5 times per week saw 47% better reach consistency than sporadic posters under the 360Brew topic-authority model. The follower-count implication is more striking: an account with 8,000 focused followers now routinely outperforms one with 80,000 unfocused ones. Audience size stopped being the asset. Audience coherence became the asset, because coherence is what lets the model predict who else should see the post.

Cadence works by conditioning the Stage 1 test-audience size. Accounts in our dataset that maintain a 48-72 hour posting rhythm for 6 or more consecutive weeks receive a noticeably wider Stage 1 window, roughly 15-20% broader initial exposure by our measurement, because LinkedIn's ranking model has built a reliability score for that creator's content. Going dark for 10 or more days resets the advantage. The first 2-3 posts after a gap perform below the account's historical baseline even when the content is objectively as good, which is the most demoralizing pattern in this entire system and the one that causes the most people to quit.

Reading level and topical focus compound each other, and this interaction is absent from every competing guide we have read on the subject. An account posting exclusively about B2B revenue operations at a grade 5-6 reading level will outperform an account posting about B2B sales, leadership, and personal growth at the same reading level. The 360Brew profile-content alignment score effectively amplifies or dampens the readability benefit. Low reading level helps dwell time, which gets you through Stage 1. Narrow topical focus determines how wide a Stage 2 audience LinkedIn considers relevant enough to surface the post to.

That ordering gives you a priority sequence, and it runs opposite to how most people sequence this work. Establish topical focus first. Then optimize reading level. Improving readability on a topically scattered account produces smaller gains than the same edit on a topically consistent one, because you are improving the multiplicand while leaving the multiplier at a fraction of its potential. Readability work on a scattered account is not wasted. It is just discounted.

The failure mode here is the relaunch. An account goes quiet for a few weeks, comes back with the best post it has ever written, watches it underperform, and concludes that quality does not matter on LinkedIn. The reliability score explains it. The post was fine. The account had told the ranking model it was unreliable, and the model priced that in before anyone read a word. If you are coming back from a gap, publish your second-best ideas for the first week and save the one you care about for when the window has reopened.

Measure and Lower Your Reading Level Before You Publish

Paste the draft into a Flesch-Kincaid calculator before you publish. The target is grade 5-6. If the score comes back above the sevens, there are two edits that reliably move it: split sentences at their conjunctions, and replace the longest word in each sentence with a shorter one. Do the splitting first, since sentence length carries more weight in the formula than syllable count, and most professional writing fails on length rather than vocabulary.

Format stacks on top of reading level rather than substituting for it. Document and carousel posts reach a median engagement rate of 6.60%, against 2.78% for video and 1.54% for text-only, largely because multi-slide swiping mechanically extends dwell time no matter what the slides say. A carousel at grade 7-8 will beat a text post at grade 7-8. A carousel at grade 5-6 beats both. Neither lever replaces the other, and the accounts that treat format as the whole answer end up publishing readable-looking carousels full of unreadable sentences.

Run a template audit at the same time, because it catches the problem the readability score cannot see. Look back at your recent posts and count how many opened with a bold hook, moved to three bullet points, and closed with a question. If most of them fit that shape, vary the next post's structure before you touch its reading level. Template recognition degrades reach over 4-6 weeks regardless of how plainly each individual post is written, and no amount of sentence-splitting fixes a silhouette your readers have memorized.

It helps to separate the signals you control from the ones you only observe. Reading level, external link placement, and post length are inputs. You set them before publishing, in a few minutes, with free tools. Dwell time, save rate, and Stage 2 expansion are outputs: real, heavily weighted, and not directly editable. Most LinkedIn advice spends its time on the outputs, which is why it reads as motivational rather than operational.

We build tools in this category, so the honest version of the recommendation includes its limits. Lowering your reading level is the cheapest reach improvement available on LinkedIn right now, and it is the one almost nobody in B2B is doing, which is the whole reason it works. It will not rescue a profile with no topical pattern, it will not undo a 10-day posting gap, and it will not survive a template your audience has already learned to scroll past. Fix the focus, hold the rhythm, vary the shape, then write it so a tired reader gets through it in one pass.

Frequently asked questions

What reading level should my LinkedIn posts be written at for maximum reach?

The practitioner-consensus target is approximately a 4th-grade reading level, with the largest reach gains occurring between grade 7-8 and grade 5-6. Posts above a 10th-grade reading level see roughly 35% less reach. The mechanism is dwell time: readers process plain sentences faster without losing comprehension, so they spend more time in the viewport, which feeds 360Brew's top-weighted distribution signal.

Do external links in LinkedIn posts hurt my reach in 2026?

Yes. External links in the post body reduce reach by 50-70% under LinkedIn's current algorithm. LinkedIn's March 2026 Authenticity Update also closed the first-comment link workaround, applying the same penalty to that 'bridge behavior.' The practical alternative is to place links in your profile section and reference them in the post body as 'link in profile' without creating a visible URL in the post text.

How does LinkedIn's 360Brew algorithm decide what posts to show?

360Brew is a roughly 150-billion-parameter foundation model that evaluates post text semantically, predicts dwell time from reading difficulty and content quality, checks profile-content alignment, and scores engagement signal quality. It distributes posts in two stages: 2-5% of the author's network in the first hour, expanding to 10-20% only if early engagement exceeds 5-10% of that initial audience.

What types of posts perform best on LinkedIn in 2026?

Document and carousel posts achieve the highest median engagement rate at 6.60%, compared to 1.54% for text-only posts. Multi-slide swiping mechanically increases dwell time regardless of content quality. Text posts at a low reading level still outperform text posts at a high reading level. Format and reading level are independent levers; optimizing both compounds the distribution benefit.

Does posting frequency affect LinkedIn reach? How often should I post?

Accounts posting 3-5 times per week see 47% better reach consistency than sporadic posters. Maintaining a 48-72 hour rhythm for 6 or more consecutive weeks produces a roughly 15-20% wider Stage 1 distribution window because LinkedIn's model builds a reliability score for consistent creators. A gap of 10 or more days resets this advantage, depressing the first 2-3 posts after the gap below historical baseline.

Does LinkedIn penalize AI-generated content?

360Brew does not apply a binary AI penalty. It evaluates lexical diversity as a proxy for authenticity. AI-generated text tends to fall into predictable patterns: repeated transitions, unnaturally steady tone, identical structural templates. These patterns depress dwell time and reduce distribution. The problem is template predictability, not AI origin. Human-written posts following the same template every time accumulate the same penalty over 4-6 weeks.

What is dwell time on LinkedIn and why does it matter for reach?

Dwell time is the number of seconds a user actively views a post in the feed. It is 360Brew's top-tier distribution signal. Posts sustaining 30 or more seconds of active viewport engagement are amplified into Stage 2 distribution. Posts generating under 2% engagement in the first 60 minutes are deprioritized. Reading level affects dwell time directly: lower reading level reduces processing friction, keeping readers in the viewport longer.

Why did my LinkedIn reach drop in 2026?

LinkedIn organic reach fell approximately 50% year-over-year in 2026 following the 360Brew rollout and the March 2026 Authenticity Update. The most common account-level causes are: posts at high reading levels depressing dwell time, external links in the post body or first comment triggering the reach penalty, topically inconsistent posting reducing the profile-content alignment score, and posting gaps resetting the consistency reliability score.

How do LinkedIn saves affect post distribution compared to likes?

Saves carry approximately 5 times the algorithmic weight of likes in LinkedIn's distribution model, based on analysis of more than 3 million posts. A post earning 10 saves will outperform a post earning 50 likes in Stage 2 expansion. Saves signal that a reader found the content worth revisiting, a high-intent behavior that 360Brew treats as a strong quality indicator relative to passive reactions.

Do LinkedIn polls and comments count as engagement signals for algorithm visibility?

Comments carry more weight than likes but less than saves, because they require effort and indicate the reader stayed long enough to form a response. Polls generate impressions and interaction data and are primarily useful for dwell time: users must read the question and options, extending viewport time. Neither polls nor comments substitute for the dwell-time signal that comes from low reading level and substantive content.

Sources and further reading

• LinkedIn's official feed relevance documentation
• LinkedIn Engineering Blog: multitask machine learning for content ranking
• peer-reviewed study on post-caption readability and social media engagement