Best LinkedIn post length: data analysis of characters and lines

The best LinkedIn post length is not one character count. LinkedIn publishes a ceiling: a feed post can contain up to 3,000 characters. It does not publish the cut where the feed hides the rest behind "see more." Public datasets from 2026 then disagree with each other about whether longer text earns more reactions. They disagree because some compare every post on the internet with every other post, and some compare a writer with that same writer.

Boykli does not score length, and it does not draw the fold. Create posts asks you to pick a band before the draft exists. Short is labeled 45–400 characters, medium 400–1,400, long 1,400–3,000. Those labels are writing targets. The ceiling on the way out is still LinkedIn's 3,000, including spaces and line breaks. A caption on a document post uses the same ceiling. How to look at the first lines on a phone and on a desktop column, after the draft exists, is LinkedIn post preview.

Logged-in visitors can browse Radar and Algorithm with no plan. Writing, carousels, Voice Agent, publishing, and scheduling need a paid seat. BYOK is $29 a month. Managed is $88 a month. Prices are the public list on boykli.com as of 27 October 2026.

The only length LinkedIn states as a rule

LinkedIn Help on posting, read 27 October 2026, says posts can contain up to 3,000 characters. The same page says that if a post exceeds 3,000 characters, write an article on LinkedIn's publishing platform instead. The help page does not say whether the count is Unicode code points or UTF-16 code units. It does not give a character count for "see more."

A second LinkedIn page is advice, not a measurement. On 30 June 2026, LinkedIn's marketing blog told writers aiming at AI search to keep posts between 200 and 300 words, and to treat articles as a longer form at about 800 to 1,200 words (How to maximize AI visibility for your LinkedIn posts). The post gives no engagement table beside that range. Two hundred to three hundred words is roughly 1,200 to 1,800 characters if you use the 5.9 characters-per-word median Depost measured in its own library. It lands inside Boykli's long band, and at the top of medium.

Treat the fold as a line problem that other people have measured.

Boykli's editor will not insert a "see more" marker, and it will not refuse a draft for being under 1,200 characters. The floor on a core idea for short, medium, and long is 10 characters. Ten characters is a guard against an empty box. It is not a post. The first line of the draft is the hook. A separate switch can write one to five alternate openings from your material. That job is the LinkedIn hook generator.

Characters, lines, and the fold

A character count and a line count answer different questions. The character count tells you whether LinkedIn will accept the post, and which public study band you are in. The line count tells you what a stranger sees before they tap.

A phone column is often described as about three lines, near 140 characters, before "see more." A blank line spends one of those lines on nothing. Two short sentences and a blank line can put the number you cared about under the fold, even when the whole post is under 400 characters. A desktop column near 210 characters can show the number that the phone hid. That is why a single badge of "under 210" is not a preview.

LigoSocial's September 2026 length report measured line shape on the same 134,522-post set it used for length. Posts whose lines averaged under 40 characters did no better than posts whose lines averaged 70 to 110 characters (minus 1 percent versus that second group). The length of the first line barely moved the result: under 40 characters on line one was plus 7 percent versus a first line of 40 to 70. The same report says a first line that opens a moment beat the same author's bold-claim openers by 31 percent, and question openers were minus 6 percent. Line breaks are not a growth tactic. They are how the claim survives three lines. Where those breaks should fall, and which invisible characters fake a gap, is the LinkedIn text formatter.

Count a draft in plain text before you trust a word-processor total. A blank line is two line-break characters if you pressed Enter twice. Those characters are inside the 3,000. A post that looks short because every sentence has its own line is spending the budget on structure. That is usually the right trade. Spending it twice, once on the sentence and once on a Unicode copy of the same sentence, is not.

What the 2026 datasets measured

Four public write-ups are the ones buyers paste into Slack. They are not four measurements of the same thing.

MagicPost's ladder and Depost's flat ratios can both be printed on the same day and still describe different questions. MagicPost's medians rise because the posts in the long bands, across all accounts, collect more likes. Depost's ratios stay near 1.00 because each writer is scored against their own typical post, so a large account's long posts do not set the baseline for a small account. Ligo is the awkward third: it is within-author, and it still finds a premium, but the baseline is one character band (600–900), and the outcome is "engagement," which that page does not define as Depost's reactions-plus-comments-plus-reposts. Do not average plus 24 percent and 1.05 into a single "best length."

Ligo's own caveat matches the failure mode that matters on a writing product. Longer posts are often the ones with a real story, a number, and an edit. Posts under 300 characters are often a link, a reshare caption, or an announcement. Padding a thin note to 2,400 characters does not buy the band. LinkedIn's 2026 distribution notes are about that padding when it reads as generic. The action list is LinkedIn algorithm penalties. How retrieval and ranking use dwell, separate from a character target, is the LinkedIn algorithm guide.

Pick a band from the job, then stop

Use the studies as a veto, not as a quota. The veto from Depost is the 50–149 word post: long enough to hide the point, short enough that the tap finds nothing. The veto from Ligo is the under-150-character post used as a habit, when the note actually had a number in it. The veto from LinkedIn Help is anything over 3,000 characters in the feed composer.

Boykli's other modes are not length sliders. YouTube, article, and document start from a source. Series is 2 to 8 posts. Tone, a hook switch, and a call to action are separate. The draft comes back as plain text. The app does not add emoji or hashtags. If the source note is thin, a long band will not invent the missing number. The four-line check before any of these bands is the thought leadership framework.

A practical conversion, so the tables above stay honest: Depost's library median was 5.9 characters per word. MagicPost describes 200–400 words as roughly 1,200–2,400 characters. Both are English prose with spaces. A post of one-line sentences and blank lines runs more characters per word. Recount the draft. Do not convert in your head and then discover the composer is over.

A counting pass before the post is queued

Budget twenty minutes the first time, and five minutes once the cuts are familiar. Boykli publishes the text you approved. It does not shorten it on the way out.

1. Write the note in plain words before you open a length band. One incident that happened, one figure, one decision you will live with. If the note is a topic ("thoughts on post length"), stop. A long draft of a topic is the generic case. Why the draft has to start from material you supplied is the founder account.
2. Decide whether the stranger must tap. If the sentence is complete in under about 140 characters and two lines, use short and do not add a paragraph to look serious. If the sentence needs the objection and the figure, use long and put both in the first two lines anyway.
3. Generate one draft in that band. Read the character count in the editor you will actually publish from. Include the line breaks. Stay under 3,000 with room for a last sentence. If the draft is over, cut a second idea. Do not switch the overflow into styled letters to "save space." Styled letters are still characters, and a UTF-16 count charges many of them twice.
4. Lay the lines in plain text. One idea, then a break you typed. No braille blank, no no-break space, no zero-width space in a URL. The first two lines must still make sense if the reader never taps. A question is allowed when the post already stated your side. A question that is the entire opening is the pattern Ligo's first-line cut marked down.
5. Read it at phone width and at desktop width. Boykli does not simulate either column. The method, including the case where a three-line clamp hides the method, is LinkedIn post preview. If a blank line is inside the opening, delete it and see whether the figure moves above the fold.
6. Publish that text. If a queue will send it, save the same characters you just checked. A later "tidy" in another editor can put the blank line back.
7. After eight posts, split your own analytics by band. Record impressions, reactions, and comments at seven days, and write down the character count you posted. LinkedIn's impression figure is an estimate. Eight posts will not crown a winner. They will show whether your 50–149 word posts are the ones with nothing under the fold. Compare text with text. A PDF in the long band is a format test, not a length test.

Do not run this pass by asking a model to "make it 1,500 characters." The model will add a second example you did not have. Length that arrives without a new fact is the padding Ligo warned about, and it is the generic text LinkedIn described in 2026.

FAQ

What is the LinkedIn post character limit?

3,000 characters, including spaces and line breaks, on LinkedIn's posting help. The same page sends you to an article when the post is longer. Boykli's long band stops at that ceiling. It is not a goal.

How many characters show before "see more"?

LinkedIn does not publish that count. The measurement people repeat, including AuthoredUp's, is about 210 characters on desktop and about 140 on mobile, often described as about three lines. A blank line, a larger phone font, or a narrower column moves it. Test the lines on both clients. A counter with a single badge is not the feed.

Do line breaks count toward the 3,000?

Yes. They are characters in the post. A blank line is two breaks if you pressed Enter twice. Ligo's line-shape cut found almost no engagement gap between very short lines and ordinary paragraphs. Breaks are for the fold and for scanning. They are not a second algorithm.

Is 1,300 to 1,600 characters the best LinkedIn post length?

It is a band several cross-author charts pass through. AuthoredUp's quoted range starts at 1,301 characters. LinkedIn's June 2026 advice of 200–300 words sits near it. Depost's within-author scores in that neighborhood are about 1.01, almost the writer's own median. Use it when the incident fills it. Do not pad a 400-character note up to it.

Do longer posts get more impressions?

MagicPost's synced subset shows median impressions rising from 508 under 25 words to 999 at 400 words or more. That is across authors. Depost did not score impressions, because impressions are visible to the author, and its within-author engagement ratios stay near 1.00. Ligo reports a within-author engagement premium for posts over 2,400 characters against that author's 600–900 character posts, and calls it correlation. Your own eight-post split is the number that applies to your profile.

Which Boykli length should a founder use?

Match the note. Short (45–400 characters) when one sentence is the post. Medium (400–1,400) when the story is one scene. Long (1,400–3,000) when you owe the reader the objection and the figure. The 50–149 word middle, which overlaps the top of short and the bottom of medium, is the band to leave on purpose: either finish the thought above the fold, or put a real body under it.

Will Boykli tell me the length that will rank?

No. The product does not read your past impressions, and it does not estimate "see more" rate. You pick the band. You count the draft. You check the lines on a phone. The studies on this page are public samples through October 2026. They are not a coefficient for next Tuesday.