How to Prompt AI to Write a LinkedIn Post

To prompt AI to write a LinkedIn post that does not read like it came from a template, feed it a specific real detail, a result, a mistake, an observation from your week, instead of a topic. Tell it explicitly what to avoid: rhetorical-question hooks, "here's what nobody tells you" framing, one-sentence-per-line formatting piled on for its own sake, and forced humility. Give it a line or two describing your actual tone. Most AI LinkedIn posts are recognizable in the first sentence because the prompt handed the model a topic and let it invent the hook, the story, and the lesson on its own. Fix what goes in, and the output stops sounding like everyone else's feed.

Why AI-written LinkedIn posts are so easy to spot

You can spot the pattern in one line. "I got fired from my first job. Here's what it taught me about leadership." A confessional hook. Then a wall of single-sentence paragraphs, each one alone on its own line, building to a lesson that sounds suspiciously tidy. Then three bolded fragments with an emoji in front of each. Then a line asking you to drop a comment.

None of that is an accident. It is what happens when a model gets asked to write "a LinkedIn post about leadership" with nothing else to go on. The platform's most-shared posts over the last few years lean hard on a handful of structural tricks, and those patterns show up constantly in training data. When a model has no specific input, it defaults to genre conventions instead of your actual voice.

The specific tells, spelled out:

• Generic hook-bait first lines, a sentence built purely to stop the scroll, unconnected to anything the writer actually experienced.

• Forced humble-brags: "I don't usually share this, but..." followed by a result clearly meant to impress.

• One-sentence-per-line formatting applied to the entire post, whether or not any individual line needs the emphasis.

• Fake vulnerability, a confession that reads like it was reverse-engineered from "vulnerability performs well" rather than something the person felt.

• "Unpopular opinion:" openers attached to opinions that are, in fact, popular.

• Rhetorical questions as hooks, like "Ever wonder why your team keeps missing deadlines?"

None of these are wrong because they are formulaic. They are wrong because they are generic, interchangeable across any topic and any person. That is the problem a better prompt can actually fix.

The real fix: stop asking for a topic, feed it a result

Most weak prompts look like "write a LinkedIn post about the importance of good onboarding." That is a topic, not information. The model has nothing to draw on except every other LinkedIn post about onboarding it has seen, so it writes something close to the average of those, which is exactly the genre-conventions problem above.

A prompt that produces a specific post gives the model a specific thing that happened: a number, a decision, a mistake, a before-and-after that a stranger could not write about anyone else. This is the same idea that runs through prompt engineering fundamentals generally: specificity beats instruction. If you have noticed this same flattening in other kinds of AI output, not just LinkedIn posts, it is usually the same root cause.

That root cause is covered in more general terms in how to prompt AI without getting generic answers: the input has no texture, so the output cannot have any either.

A prompt structure that actually works

Here is a template. Fill in the bracketed parts with your real situation, then paste the whole thing in.

Write a LinkedIn post based on this specific thing that happened to me:

[Describe the actual event, result, or lesson in 2-4 sentences. Include a real number or detail if there is one.]

Rules:
- Do not open with a rhetorical question.
- Do not use "here's what nobody tells you" or similar reveal framing.
- Do not manufacture a confession or vulnerability that isn't in what I described.
- Write it the way I would actually say it out loud to a colleague, not the way a LinkedIn post "should" sound.
- No emoji. No hashtags.
- Short paragraphs are fine, but do not force every sentence onto its own line.
- End on the actual point, not a call for engagement or a "thoughts?" prompt.
- Keep it under 150 words.

My usual tone is [describe it: e.g. "direct, a little dry, not big on exclamation points"].

Before and after: the same idea, two prompts

Same underlying story both times: a freelancer switched from hourly billing to flat-rate project pricing and it worked out better than expected. Here is what a generic prompt produces next to what the structured version produces.

Generic prompt: "Write a LinkedIn post about switching from hourly billing to flat-rate pricing as a freelancer."

What that produces:

Ever feel like you're trading time for money and getting nowhere?

I used to think hourly billing was the only way to freelance. I was wrong.

Here's what nobody tells you about pricing your work:

Hourly billing punishes speed.
Flat-rate pricing rewards value.
Clients respect confidence.

I made the switch to flat-rate pricing six months ago and it changed everything.

My income went up. My stress went down. My clients trust me more.

Unpopular opinion: if you're still billing hourly in 2026, you're leaving money on the table.

Are you ready to make the switch?

Drop a comment if you want my pricing framework.

Structured prompt, filled in with the actual detail: "I switched three clients from hourly billing to flat-rate project pricing in March. Revenue on those same three accounts is up about 20 percent since then, mostly because hourly billing was quietly punishing me every time I got faster at the work. Two of the clients told me they preferred knowing the total upfront."

What that produces:

I switched three clients from hourly to flat-rate pricing in March.

Revenue on those same three accounts is up about 20 percent since then. Not because I raised my rates. Because hourly billing was quietly punishing me every time I got faster at something.

The bigger surprise was the clients. Two of them told me they actually preferred knowing the total upfront, even when it landed a bit higher than they expected an hourly estimate to be. Nobody likes an open-ended number.

I still quote hourly for anything I can't scope well. But for repeat work, flat-rate has been the better deal for everyone, not just me.

Notice what changed. The second version has a real number, three clients, March, 20 percent, instead of a vague timeframe. It does not open with a question. It does not claim a universal lesson for every reader, it just describes what happened to one person. No emoji, no bolded fragment list, no ask for comments. It reads like something a person would say to a colleague over coffee, which is the actual test.

Voice constraints worth putting in every prompt

The template above bakes in five constraints. Worth understanding why each one matters, since you may need to adjust them for your own voice.

• No rhetorical questions as hooks. They are a tell because they are addressed to nobody in particular. Real observations tend to start mid-thought, not with a survey question.

• No "here's what nobody tells you" framing. Nine times out of ten, somebody has told people. The phrase exists to manufacture false exclusivity, and most readers have learned to discount it on sight.

• No forced vulnerability. If the thing you are describing is not actually a vulnerable moment, do not let the model invent one. A humble-brag reads like a humble-brag even dressed up as honesty.

• Write it the way you would say it out loud. This one instruction does more work than any formatting rule. Read the output aloud. If you would not say it to a colleague at a desk, rewrite it.

• Formatting follows content, not the reverse. One-sentence-per-line is a real technique for genuine emphasis. Applied to every sentence in every post, it stops meaning anything.

If what you actually want is a consistent, recognizable version of your voice across many posts rather than getting one post right, that is a slightly different problem, covered in how to prompt AI to match your brand voice. This post is more about one post's honesty than long-term consistency.

The editing pass: what to catch after AI drafts it

Even a good prompt needs a five-minute pass before you post it.

• Cut the first line if it is still doing "hook" work instead of just starting. Often the real first sentence of the post is buried second or third.

• Check for any claim you cannot back up. If the draft invented a number, a timeframe, or a quote you did not give it, fix or remove it. A made-up statistic in a post with your name on it is a credibility problem, not a style one.

• Read it out loud once. Anything that makes you wince is usually the model reaching for a LinkedIn-post convention instead of your actual sentence.

• Strip stray hashtags and emoji the model added out of habit, unless you deliberately want them there.

• Trim the ending. Drafts tend to close with a sentence that just restates the post. You usually do not need it.

Reusing this approach for other post types

The prompt-and-example structure above is not specific to lessons-learned posts. It works the same way for an announcement, a hot take, or a how-to breakdown, because the fix is always the same: give the model something specific instead of a category, and tell it which conventions to skip. The same principle is what makes prompting AI for a cold outreach email work better than a generic "write me a sales email" request. Specificity replaces genre convention regardless of the format.

One more thing worth deciding upfront: whether you are drafting the post yourself and pasting it in manually, or considering some kind of scheduled or automated posting. If it is the latter, it is worth reading up on what is actually safe when an AI agent posts directly to your accounts before wiring anything up, since drafting and auto-publishing carry different risks.

A related but separate problem is when the model drops a constraint you did give it instead of inventing one you didn't; why AI ignores parts of your prompt covers why that happens and how to stop it.

FAQ

Can AI write a LinkedIn post that doesn't sound like AI?

Yes, but only if the prompt gives it something specific to work with. A generic topic prompt produces generic, genre-conventional output almost every time, because that is what the average LinkedIn post in its training data looks like. Feed it a real detail, a number, an actual before-and-after, and tell it explicitly which conventions to skip, and the result reads much closer to something a person actually wrote.

What's the best ChatGPT prompt for a LinkedIn post?

There is no single best prompt, but the structure that works consistently has three parts: a specific real event or result rather than a topic, explicit rules about what to avoid, and a short description of your actual tone. The template earlier in this post covers all three.

Should I let AI use emojis in a LinkedIn post?

Generally no, unless emoji are genuinely part of how you write. Emoji used as bullet points is one of the clearest AI-writing tells on the platform, mostly because it shows up constantly in generated content and rarely in how people actually write to each other.

How long should an AI-written LinkedIn post be?

Short enough to read in the time someone spends scrolling past it, usually under 150 to 200 words for a single observation or lesson. Length is not really the tell people react to, though. A short post can still read as obviously AI-written if it is built from genre conventions instead of a real detail.

Do I need to fact-check AI-written LinkedIn posts?

Yes, always. Read the draft against what you actually gave it and remove or fix anything the model added on its own: a number, a timeframe, a quote that was not in your original input. This matters more on LinkedIn than most places, since the post carries your name and reads as a first-person claim.