The default AI LinkedIn post is cringe. You know it on sight: line-broken-every-sentence, “Here is a hot take…”, a humble-brag anecdote, three bullet “lessons”, and a question nobody answers because the whole thing reads like a script. This tutorial replaces that template with a four-block frame (claim, evidence, tension, ask), an anti-cringe filter that strips the worst defaults, and a comment plan tuned to how LinkedIn’s 2026 algorithm actually distributes posts.
TL;DR
- Use a four-block brief — CLAIM, EVIDENCE, TENSION, ASK — and let AI structure your point, not invent it.
- Your first 210 characters are everything: that is where LinkedIn truncates with “…see more” on desktop (~140 on mobile). The hook lives there or the post dies.
- LinkedIn tests a new post on 2-5% of your network in the first 60 minutes; only ~5% of posts that stall in hour one recover. Block 2 hours after posting for comments.
- Comments carry roughly 15x the weight of likes, and dwell time is the hidden ranking lever (0-3s of read time ≈ 1.2% engagement; 61s+ ≈ 15.6%). Write asks that earn paragraph replies.
- Skip external links in the post body: link posts get about 60% less reach, and the old “link in first comment” workaround is now penalized too (early 2026).
- Target 1,300-2,500 characters — the band with the highest median engagement (~2.6%) as of 2026.
Who this is for
Founders building a personal brand, B2B operators sharing playbooks, recruiters posting openings with real context, and consultants using LinkedIn for inbound. It is less useful for pure lead-gen accounts where the post is a thin wrapper around a CTA — those need a more direct, offer-first workflow.
Reach for this when you have a real point and 20-30 minutes, when you want a sustainable cadence (3 posts a week correlates with about 3.5x more profile views than posting occasionally), or when you have a long-form piece — article, podcast, talk — and need a LinkedIn excerpt that earns attention without dumping a link.
Before you start
- Have one real point — a claim you would defend in a room. AI cannot manufacture conviction; it can only structure one you already hold.
- Have evidence: a number, a named moment, a specific story. Posts without it read as opinion noise.
- Get specific about audience. “B2B SaaS founders” is too broad for AI to calibrate. “Series-A SaaS founders who just hired their first head of marketing” gives it something to write toward.
- Reserve 2 hours after posting for active comment engagement. On LinkedIn, posting is half the work; the first hour does most of the distribution.
The four-block frame
Every block does one job. Brief the AI on all four before it writes a word.
| Block | Job | One-line test |
|---|---|---|
| CLAIM | A single defensible sentence | Survives a “so what?” |
| EVIDENCE | One specific number, story, or moment | Could not have been invented |
| TENSION | What this contradicts or makes hard | The opposing view is steelmanned, not strawmanned |
| ASK | One open question inviting a story | Answerable in 2 sentences, not yes/no |
Paste this brief into ChatGPT (GPT-5.5) or Claude (Sonnet 4.6). Both handle the structural job well; Claude tends to keep a more even, less performative tone out of the box, which matters here.
You are helping me draft a LinkedIn post. Do NOT write motivational filler.
Audience: [specific audience, e.g. Series-A SaaS founders who just hired a first head of marketing].
My CLAIM (use as-is, do not soften): [one defensible sentence].
My EVIDENCE: [one specific number / story / moment].
My TENSION: [what this contradicts or makes harder].
My ASK: [one open question that invites a story].
Write 3 variations. Same four blocks, different opening lines only.
Rules: no line break between every sentence; no emoji bullets; no "hot take",
"unpopular opinion", or "most people get this wrong"; first 210 characters
must contain the claim or a number so the hook survives the "see more" cliff.
Length 1,300-2,500 characters.Step by step
- Brief the AI with the four-block frame above. The claim is yours — paste it verbatim and tell the model not to soften it.
- Generate 3 variations, same four blocks, different openers. Compare side by side. The strongest post usually splices block 1 from variation A with block 3 from variation B.
- Run the anti-cringe filter. Delete: “Here is a hot take”, “Unpopular opinion”, “Most people get this wrong”, “I used to think… now I know”, and any line ending in “Period.” Strip emoji bullets. Limit line breaks to between blocks, not between every sentence.
- Read it as if a competitor wrote it. Would you respect them, or think “this person is performing”? If it is performance, rewrite the weakest block.
- Format for the feed. Put the claim (or a number) in the first 210 characters so it clears the “see more” cliff. Evidence and tension as short paragraphs. Ask alone on the last line.
- Plan the first-hour engagement: 5 follow-up prompts you will drop in your own comments to extend the thread, and 5 people you will tag or reply to. Keep the link, if any, out of the body — point to “comments” or your profile instead.
- Post, then work the comments. LinkedIn tests the post on 2-5% of your network in the first 60 minutes; reply within minutes, not hours. Comments compound dwell time and outrank likes ~15:1.
First-run exercise
- Pick one real point you actually defended in a recent conversation — not a hypothetical “thought-leadership” topic.
- Run the four-block brief, generate 3 variations, splice the best blocks, run the anti-cringe filter.
- Post at your usual time and set a 2-hour timer for active comment engagement.
- After 24 hours, check the comment-to-like ratio. Above roughly 1:10 means the ask did its job. Also note dwell-time-friendly signals: did anyone reply with a full paragraph? Those replies are what the algorithm rewards.
Quality check
- Does the claim survive a “so what?” If the reader thinks “okay, and?”, the claim is too soft.
- Is there one specific number or moment? Vague evidence reads as filler and kills dwell time.
- Is the tension real? “Most people think X, but actually Y” is fake tension when X is a strawman. Steelman the opposing view.
- Is the ask answerable in two sentences? Long-answer asks die in the feed.
- Does the hook live in the first 210 characters? If the point only appears after “see more”, most readers never reach it.
- Cut every line that does not move one of the four blocks forward.
What AI tool to use
You do not need a dedicated LinkedIn tool to start. A general model plus a formatter covers most people, as of June 2026:
| Tool | Role | Price (June 2026) |
|---|---|---|
| ChatGPT Plus (GPT-5.5) | Drafting + variations | $20/mo |
| Claude Pro (Sonnet 4.6, 1M context) | Drafting with a flatter, less salesy tone | $20/mo |
| AuthoredUp | Formatting, hook library, readability score, post analytics — does not write for you | ~$19.95/mo |
| Taplio | All-in-one: AI drafts, scheduling, viral library, analytics | Starter $39/mo; AI from $69/mo |
Practical stack: draft and run the four-block brief in Claude or ChatGPT, then paste into AuthoredUp to check line breaks and the “see more” cut before scheduling. Reserve Taplio for when you want scheduling, a swipe file, and analytics in one place. For the upstream ideation and repurposing layer, see our AI content marketing stack guide.
Common mistakes
- Posting without a real claim. AI cannot manufacture conviction, only structure it.
- Shipping AI’s default formatting. A line break between every sentence is the loudest “AI wrote this” tell.
- Cringe openers. “Here is a hot take…” is banned, no exceptions.
- Fake tension. Strawman opposing views read as low-effort.
- Burying the hook past 210 characters. Anything after “see more” is invisible to most of the feed.
- Putting a link in the body (or the first comment). Both cost reach now. Point people to your profile or DMs instead.
- Posting and disappearing. The first 60 minutes set distribution; only ~5% of slow-start posts recover.
- Yes/no asks. Open-ended questions pull the paragraph-length comments dwell time rewards.
FAQ
- How often should I post?: Three times a week, sustainably, beats seven times for three weeks then nothing. Consistent posters see roughly 3.5x more profile views than occasional ones.
- How long should the post be?: 1,300-2,500 characters is the highest-engagement band (~2.6% median). The hard ceiling is 3,000. Your first 210 characters must work before “see more” truncates the rest.
- Should I use hashtags?: Three relevant ones, not ten. Over-hashtagging reads as 2018 LinkedIn and does little for reach in 2026.
- Why can’t I put my link in the post?: Link posts get about 60% less reach, and the “link in first comment” workaround is now penalized too. Drive traffic from your profile, a DM, or a follow-up.
- Can I let AI write the whole post?: No. Let AI structure your point and draft variations. The claim, the evidence, and the judgment about what to cut come from you — that is also what keeps it from reading machine-generated.
- What about LinkedIn newsletters?: Different workflow — newsletters need a series arc, not a single hook. See the AI newsletter tutorial.