Which LinkedIn post formats compound over 90 days

Most LinkedIn format advice treats the choice as styling: carousels for reach, text for authenticity, video for presence. That framing misses the mechanism. In our client data, the format you lead with sets the ceiling your whole profile can reach, because after 90 days, profile-level signals drive over half your distribution.

The format you pick at launch sets the ceiling

The format you lead with over the next three months matters more than the engagement rate of any single post you publish. Here is the number that explains why. After 90 days of consistent topical posting, profile-level factors, meaning your follower history, your posting frequency, and your topical consistency, account for 50.1% of a post's eventual reach. The individual post's own engagement signals account for only 29.5%. Most of what decides whether your next post travels is settled before you hit publish, by the profile you spent the previous three months building.

Carousels give you a running start on building that profile. On personal profiles, carousel and document posts average 1,387 impressions each. Text-only posts average 589. That 3.7x gap does not switch on once the algorithm learns who you are. It compounds from the first post, because every carousel feeds a larger reach figure into the profile-level signals that will dominate your distribution three months later. A text-first account is compounding from a smaller base every single week.

SocialNexis client data shows the two paths separate in a specific window. Carousel-first accounts hit a steep reach inflection between days 35 and 50, the point where the algorithm's topic classification of the profile solidifies. Text-only accounts posting at the same frequency show a shallower curve that plateaus earlier and at a lower absolute ceiling. Same cadence, same effort, different ceiling. The format you pick at launch does not only set your per-post engagement rate. It sets the height of the curve you spend the quarter climbing.

Carousels earn 7.00% against 4.50% for text, roughly 3 to 6 times the rate of text-only posts. Most format guides rank by that engagement rate and stop, as if the decision were cosmetic. The lasting advantage is structural. A higher-reach format builds a higher-reach profile, and the profile is what carries you once the first three months are behind you.

So the honest answer to which format compounds most is carousel, and the reason is not that carousels are prettier. It is that they push more reach into the profile-level signals early, and those signals are half your distribution by day 90. The rest of this guide is the mechanism behind that answer, plus the cadence and voice mistakes that flatten the curve before it can compound.

How the algorithm builds a fingerprint for your profile

LinkedIn's feed no longer ranks posts by raw engagement volume. It runs on an LLM-powered generative recommender model that processes more than 1,000 past interactions per reader and ranks each candidate post by relevance and expertise match, a shift LinkedIn confirmed in its own engineering blog. That architecture is the structural reason profiles compound. The model is not scoring your post in isolation. It is scoring your post against everything it already believes about your topic and your reader.

Every topically consistent post you publish sharpens the model's classification of your profile. Post about one subject for weeks and the recommender builds a confident fingerprint: this account is about this thing, show it to people who care about this thing. Scatter across unrelated subjects and the fingerprint stays blurry, which means the model has nothing sharp to match against when it decides who sees your next post.

Time works in your favor here in a way it did not two years ago. Posts now resurface in the feed for 2 to 3 weeks, where they used to fade in 1 to 2 days. A steady weekly cadence therefore keeps several posts in circulation at once, so your effective presence in the feed is larger than your publishing calendar suggests. You do not need to post daily to stay continuously visible. You need to post consistently enough that older posts are still circulating when the new one lands.

Outbound engagement changes the golden window for your next post, and the timing is specific. In SocialNexis client data, accounts that comment on 5 to 8 posts from high-reach accounts in their niche in the 30 minutes before publishing consistently see first-hour impressions 40 to 60% higher than accounts that publish cold. The mechanism is simple: outbound activity primes the algorithm to surface your next post to an already-warm, topically relevant slice of the audience. You are telling the recommender which room you are standing in right before you speak.

Put those three together and the compounding mechanism is legible. The recommender fingerprints your topic, extended resurfacing keeps your back catalog working, and pre-publish engagement warms the audience for each new post. None of it requires more volume. It requires consistency and sequence, which is exactly what most posting strategies get backwards.

The carousel is the format to build around

Rank the formats by 2026 engagement rate and the hierarchy is clear. Across 1.3 million posts, document and carousel posts lead at 7.00%, followed by multi-image at 6.45%, video at 6.00%, image at 5.30%, text at 4.50%, and link posts last at 3.25%. That spread is not trivia. It is the difference between a format that seeds your profile with reach and one that quietly caps it.

Document posts lead all major formats with 14% year-over-year engagement growth. Polls are the only other format showing meaningful growth, and from a much lower base, 4.40% against 7.00%. When a format's engagement rate is climbing across a dataset this size, the algorithm is rewarding it more each cycle, which means a carousel published today is competing in a category the feed is favoring.

Polls deserve a specific mention because they are overhyped. Poll engagement has doubled to 4.40% since 2023, which sounds impressive until you set it against document posts. Polls trail carousels by a wide margin, 4.40% against 7.00%. Within the poll category, 3-option polls running 7-day durations outperform shorter formats. Treat polls as a tactical supplement for a slow week, not a primary format you build a compounding strategy around.

Carousels and newsletters compound different channels at the same time, and they reinforce each other rather than compete. SocialNexis client data shows newsletter open rates are highest among readers who first discovered the creator through a carousel post. The carousel builds reach inside the feed; the newsletter converts a slice of that reach into guaranteed delivery outside it. Running both does not split your audience. It builds two audience segments that feed each other, with the carousel doing the discovery work and the newsletter doing the retention work.

Your primary format sets your ceiling, so make it the one the algorithm rewards most and the one that generates the dwell time signals covered in the next section. For nearly every account trying to compound reach in 2026, that primary format is the carousel. Everything else, polls, the occasional text post, the newsletter, plays a supporting role around it.

Video is the format going backwards

Video is the one major format where reach is going backwards. Across 1.3 million posts, video views dropped 36% year-over-year, and they did it while video posting frequency rose 36%. More creators posted more video and collectively earned fewer views. That is the signature of a saturated format: supply climbed, the reach pool did not, so reach per post fell.

To understand why carousels hold up where video slips, look at the first hour. The first 60 to 90 minutes after posting determine almost all of a post's eventual reach. Posts that fail to clear roughly 500 impressions in that first hour receive minimal further distribution. And the primary quality signal the algorithm reads in that window is not likes. It is dwell time, how long readers stop and stay on your post.

The dwell numbers are stark. Posts that hold reader attention for 61 or more seconds achieve a 15.6% engagement rate. Posts viewed for 0 to 3 seconds achieve 1.2%. That gap is enormous, and it is driven by attention alone. It also explains the structural edge of the carousel. A carousel earns dwell time mechanically: every swipe is another few seconds of held attention, and a reader working through a full deck is generating exactly the signal the feed rewards.

Video has to earn that same dwell time inside the first few seconds of playback, before a viewer decides whether to keep watching or scroll. In a feed where video supply is up 36% and views are down 36%, that opening threshold is getting more unforgiving every quarter. The videos that clear it are excellent. The median video does not, which is why the format's average reach keeps sliding even as more people pour effort into it.

This is not an argument to never post video. It is an argument about where the compounding lives. If your goal is a profile that reaches more people every month, you want the format that reliably generates dwell time in the golden window, and right now that format is the carousel, not video.

Posting twice in a day loses reach, it does not add it

The most common cadence mistake is treating more posts as more reach. It is the opposite. Accounts that post 2 or more times in a single calendar day see a median per-post reach drop of over 40%. Doubling up does not expand your reach for the day. It splits the same reach pool across more posts while flagging your account as a high-frequency publisher, which the feed treats as a downgrade signal.

That penalty is not uniform across formats, which is a detail almost no cadence guide mentions. Text posts take the sharpest hit when doubled in a day. In SocialNexis client data, carousel posts show only about 20 to 25% suppression against the 40%-plus that text absorbs. The likely reason is dwell time again: a carousel's swipe-driven attention sends a strong enough quality signal to partially offset the cadence flag. The cadence ceiling is real, but it is format-dependent, not a fixed number that applies to everyone equally.

The other half of the cadence problem is topic, not timing. Domain-specific expert content earns roughly 3.7x more reach than generalized content. Combine that with the fact that profile-level signals drive 50.1% of reach at 90 days, and the lesson is blunt: topical consistency is not a nicety, it is the input the compounding runs on. Posting about an adjacent but unrelated subject just to fill a slot does not merely underperform. It blurs the topic fingerprint you are trying to sharpen, eroding the profile-level signals that consistent posting exists to build.

So the failure mode to name and avoid is the fill-the-calendar reflex: you committed to a cadence, you are short an idea, so you post something off-topic to keep the streak alive. That post costs you twice. It does not compound on its own, and it dilutes the fingerprint that makes your on-topic posts travel. A skipped day is cheaper than an off-topic post.

The functional ceiling for most accounts is one substantive post per day, and a steady weekly cadence on a single tight topic usually beats daily output. It avoids the double-post penalty, it keeps the fingerprint sharp, and it leaves room for the pre-publish engagement that warms each post. Volume is not the lever. Consistency on one topic is.

Subscribers compound; followers don't

Feed reach and newsletter delivery obey different rules, and confusing them is why a lot of creators plateau. Feed followers receive your posts only when the algorithm decides to serve them. Newsletter subscribers get delivered to every time. LinkedIn sends push, in-app, and email notifications for each newsletter issue, and those land at 40 to 50% open rates against 20 to 25% for traditional email. Newsletter subscriber bases grew 150% year-over-year platform-wide in 2026. A subscriber is a guaranteed delivery; a follower is a lottery ticket the algorithm draws from.

Inside the feed, the ranking signals shifted in a way that matters for what you write. Saves now outrank likes as the most important feed ranking signal after LinkedIn's LLM-powered algorithm update in early 2026. A post with 200 saves will outperform one with 1,000 likes in distribution. Saves are a stronger signal because they encode intent to return: a reader saving your post is telling the algorithm this is worth coming back to, which is a far better proxy for quality than a reflexive like.

That shift has a direct bearing on AI-assisted writing, and it gives you an early-warning metric most creators miss. In SocialNexis client data, voice drift under AI-assisted drafting shows up in the saves-to-impressions ratio before the creator notices anything is wrong. The first signal is a decline in saves, not in likes or comments, and it typically appears in weeks 3 to 5 of an AI-assisted cadence. What is happening: readers register that the content is structurally familiar but tonally off, so they stop saving it while still occasionally liking it out of habit.

The practical instruction is to audit save rate, not overall engagement rate, when you introduce AI into your drafting. Overall engagement is a lagging, noisy signal that can hold steady while the thing that actually compounds, the reason people return, quietly erodes. Save rate degradation is the canary. If your saves-to-impressions ratio starts sliding a few weeks into an AI-assisted stretch, your voice has drifted, even if the likes still look fine.

Newsletters accumulate; feeds reset. Every feed post starts a fresh negotiation with the algorithm. A newsletter subscriber base only grows, and it delivers regardless of feed volatility. That is why the compounding play is to build feed reach with carousels first, then convert a slice of it into a subscriber base that no longer depends on the algorithm serving you at all.

The 90-day plan

Three phases, one tight topic, held constant throughout. The plan assumes carousels as the primary format. Topic constancy is not optional; it is the input the whole thing runs on.

Phase one, the opening weeks: establish the topic fingerprint. Publish carousels on a single narrow topic at a steady weekly cadence, and comment on 5 to 8 posts from high-reach accounts in your niche in the 30 minutes before each post goes live. Watch first-hour impressions closely. If posts consistently fail to clear roughly 500 impressions in the first hour, your topic is too broad or your audience is not warmed yet. Narrow the topic before you add any volume; more posts on a blurry topic will not fix a distribution problem.

Phase two, days 35 to 50: let the compounding run and change nothing. This is the window where carousel-first accounts hit their reach inflection in SocialNexis client data, and it is the exact moment creators sabotage themselves by increasing frequency or pivoting to a hotter topic. Do neither. The profile-level signals that will drive 50.1% of your eventual reach need consistency to solidify. Because posts now circulate for 2 to 3 weeks each, several of your posts are in simultaneous distribution during this phase without you publishing any more often. Stability is the strategy here.

Phase three, around week 8: add the non-feed channel. Launch a LinkedIn newsletter once your cadence has established enough algorithmic recognition, which tends to accelerate around week 8. Your first subscribers will come disproportionately from the carousel audience you spent two months building, which is exactly why the newsletter goes third and not first. From here, the two channels run in parallel: carousels keep compounding feed impressions, and the newsletter compounds guaranteed delivery to a subscriber base the algorithm cannot throttle.

The whole plan is one idea executed patiently: pick the format that seeds reach, hold one topic long enough to be fingerprinted, warm each post before it publishes, and only diversify channels once the base exists. None of the steps are clever on their own. The compounding comes from doing them in order and not breaking cadence when the curve is still shallow, which is the point where most accounts quit.

Frequently asked questions

Which LinkedIn post format compounds the most reach over 90 days: carousel, video, or text?

Carousel (document) posts compound the most. SocialNexis client data shows carousel-first profiles hit a reach inflection point between days 35 and 50 that text-only profiles miss entirely, even at identical posting frequency. Carousels average a 7.00% engagement rate and 1,387 impressions per post in 2026, versus 589 for text-only posts. Video is the one major format where reach is declining: views dropped 36% year-over-year despite a 36% increase in video posting frequency.

At what point in a 90-day posting cadence does LinkedIn's algorithm start distributing content more broadly?

The inflection typically arrives between weeks 5 and 7 for accounts posting 3-4 times per week on a consistent topic. LinkedIn's LLM-powered generative recommender model processes 1,000+ past user interactions per reader; by week 5-6, the profile has generated enough topically consistent signal for the algorithm to build a reliable expertise match. Carousel-first accounts see a steeper acceleration in this window than text-only accounts posting at the same frequency.

Why are LinkedIn video views declining while carousel engagement rates keep rising in 2026?

Video views dropped 36% year-over-year across 1.3 million posts analyzed, even as video posting frequency rose 36%; supply grew but algorithmic reach per post contracted. Carousel posts generate dwell time structurally through swipe interactions, and dwell time is the primary quality signal in LinkedIn's feed ranking window. Video must earn dwell time within the first few seconds of playback, and in a more competitive feed, fewer videos clear that threshold before viewers scroll past.

What is the difference between LinkedIn reach growth and follower growth, and which should I optimize for first?

Reach (impressions per post) and follower count are driven by different mechanisms and frequently diverge. Reach compounds through topical consistency and formats that generate strong dwell-time signals; follower growth comes from profile visits triggered by readers discovering your posts in someone else's network. Optimize for reach first: consistent posting with a focused topic builds the profile-level signals that drive broad distribution, and follower growth follows as that distribution brings new readers to your profile repeatedly.

How does posting frequency affect per-post reach on LinkedIn, and is there a cadence ceiling?

There is a cadence ceiling. Accounts posting twice in a single calendar day see a median per-post reach drop exceeding 40%. That ceiling is format-dependent: text posts show the sharpest suppression, while carousel posts show approximately 20-25% suppression because dwell time partially offsets the cadence flag. One substantive post per day is the functional ceiling for most accounts. Posting 3-4 times per week often outperforms daily posting because it prevents reach dilution while maintaining the consistency signals the algorithm requires.

What signals does the LinkedIn algorithm use to build a topic fingerprint for your profile?

LinkedIn's LLM-powered generative recommender model draws on three signal categories: identity signals (your connection graph and stated expertise), content signals (topic classification, format, and keyword relevance to each reader's interests), and activity signals (your posting consistency, reader dwell time on past posts, saves, and topical focus over time). After 90 days of consistent topical posting, these profile-level signals account for 50.1% of a post's eventual reach, per published research on sustained LinkedIn posting cadences.

How do LinkedIn newsletter subscribers compound differently than feed followers over time?

Newsletter subscribers compound guaranteed delivery: LinkedIn sends push, in-app, and email notifications for each issue, bypassing feed algorithm volatility entirely. Feed followers receive your posts only when the algorithm serves them. Newsletter subscriber bases grew 150% year-over-year platform-wide in 2026, with open rates of 40-50% versus 20-25% for traditional email. The two channels grow separately; SocialNexis client data shows newsletter open rates are highest among readers who first discovered the creator through carousel posts.

What happens to LinkedIn reach if you take a two-week posting break, and how long does recovery take?

A two-week break begins degrading the profile-level signals that drive 50.1% of post reach. Posts will no longer circulate for 2-3 weeks as they did during active cadence because distribution weight has decreased. Recovery typically takes 3-6 weeks of consistent posting on a focused topic to rebuild the algorithmic fingerprint. Returning at high frequency after a break to compensate tends to trigger the cadence penalty; a steady 3-4x weekly cadence on a consistent topic rebuilds the fingerprint faster than volume spikes.

How much does dwell time affect LinkedIn feed distribution compared to likes and comments?

Dwell time is the primary quality signal in LinkedIn's golden window (the first 60-90 minutes after posting), outweighing likes and comments in determining whether a post receives further distribution. Posts that hold reader attention for 61+ seconds achieve a 15.6% engagement rate, versus 1.2% for posts viewed 0-3 seconds. Saves now outrank likes as the most important ongoing feed ranking signal following LinkedIn's 2026 algorithm update, making dwell time and saves the two signals most worth optimizing for sustained reach.

Should I post carousels or newsletters to build a long-term LinkedIn audience in 2026?

Both, in sequence. Lead with carousels for the first 90 days: they average 1,387 impressions per post versus 589 for text-only, compound feed distribution through the algorithm's topic fingerprint, and build the initial audience that will convert to newsletter subscribers. Add a newsletter around week 8-9 when your posting cadence has established algorithmic recognition. From that point, carousels grow feed reach and the newsletter grows guaranteed delivery outside the feed, and the two formats reinforce each other's retention rather than compete.

Sources and further reading

• LinkedIn's engineering blog on the LLM-powered generative recommender model
• SocialInsider LinkedIn benchmarks across 1.3 million posts and 16,645 pages
• LinkedIn's engineering documentation on dwell time as a feed ranking signal