The short answer
The best AI prompts for LinkedIn posts start from your own experience, include a sample of your writing, ban stock phrases and ask for a specific format such as a story, how-to or opinion post. The 12 prompts here cover finding ideas, writing posts, first lines, editing out AI tells, repurposing transcripts, comments and a monthly calendar. They work in ChatGPT, Claude and Gemini; Anthropic's Claude is our first pick for matching your voice from examples.
What makes a good AI prompt for LinkedIn posts
The LinkedIn feed is full of posts that were clearly written by AI: the same openers, the same three-part lists, the same "Thoughts?" at the end. Readers spot them in a second and scroll past. The 12 prompts for LinkedIn posts on this page are designed to avoid that. They start from your own experience, give the AI a sample of your writing, ban the stock phrases, and ask for formats that people actually save and share.
The structure follows the prompting advice that both OpenAI and Anthropic publish. Give the model a role and a goal, provide the context it cannot guess (your audience, your story, your numbers), state your rules plainly, and show an example of the output you want. Anthropic's guide calls examples one of the most reliable ways to steer tone, and for LinkedIn the best example is a post you wrote yourself and were happy with.
The prompts work in ChatGPT, Claude, Gemini or Microsoft Copilot. They are grouped by job: finding ideas, writing posts, hooks and editing, and repurposing and engagement.
Prompts for finding LinkedIn post ideas
Most people run out of ideas because they think a post needs a big announcement. The best posts come from small things: a mistake, a question a client asked, a number from a project. The interview prompt makes the AI ask you ten questions one at a time, which pulls out stories you would never think to write about. It takes 15 minutes and usually gives a month of material.
The weekly notes prompt is the habit version: paste rough notes every Friday and get five ideas with formats and opening lines. Tell it what must stay private, because the AI does not know which client names or numbers are confidential. The opinion prompt helps you find a view you genuinely hold, then argues against it first. Posts that state a defended opinion tend to start better conversations than posts that agree with everyone.
Prompts for writing LinkedIn posts
All three writing prompts follow the same rule: the AI supplies the structure and you supply the substance. The story prompt asks you for what happened, what you felt and what changed, then starts with the most surprising moment rather than the background. The how-to prompt builds on your method and the step people usually skip, which is where the real value sits. The results prompt shows you how to share good news while crediting others and admitting what went wrong, which reads as confident rather than boastful.
Every prompt bans a few phrases. Add your own pet hates to the list. If you include a statistic, make sure you can back it up: never ask the AI to "add a statistic", because it may produce one that sounds right and is not.
Prompts for hooks and editing LinkedIn drafts
The first line decides whether anyone reads the second. The hooks prompt asks for ten openings in six different styles and ranks the best three for your audience, with a rule against clickbait the post cannot back up. Read the ten aloud and pick the one you would say to a colleague.
The "sounds like you" editing prompt is the one to use on any draft that started with AI. It removes the familiar machine phrases, including the em dash many readers now see as a tell, and marks [ADD EXAMPLE] wherever a general claim needs something real. The reader feedback prompt asks the AI to react as one specific person scrolling between meetings. It is not a replacement for real reactions, but it reliably catches posts that take too long to get to the point.
Prompts for repurposing content and engaging on LinkedIn
A single talk, report or podcast episode can feed two weeks of posts. The series prompt turns one source into five standalone posts in five formats and labels where each idea came from, so you can check it quickly. Paste the full transcript rather than a summary, because the specific quotes and numbers live in the detail. On ChatGPT's free plan the GPT Instant context window is 27K tokens, about 12 pages of input, so long transcripts may need to be split.
Commenting is how new accounts get seen, and it is where generic AI help does the most harm. The comment prompt gives you three options that add an example, disagree with a reason or ask a sharp question, and bans flattery and pitching. Always add your own detail before posting. The calendar prompt plans a balanced month and lists the stories and data you need to collect first, which is the step that keeps a posting habit alive.
How to adapt these LinkedIn prompts
- Keep a voice file. Save two or three of your best posts and a short list of words you never use. Paste them into the <example> or <my_writing> slots, or store them in a Claude Project or ChatGPT project so every chat starts with them.
- Be precise about the reader. "Founders of seed-stage B2B startups in Europe" changes the vocabulary, examples and length the AI chooses.
- Explain the rules. "Short paragraphs, because most people read LinkedIn on a phone" helps the AI apply the rule sensibly, as Anthropic's guide recommends.
- Give it real numbers or none. Supply your own figures and mark them as verified. Delete any figure the AI adds that you did not give it.
- Post your edit, not the draft. Change at least one sentence into something only you could have written. That is what makes a post recognisably yours.
Our prompt kits include a ready LinkedIn set, and how to make social media posts with AI covers the workflow end to end.
Which AI tool works best for LinkedIn posts, and why
Claude by Anthropic is our first choice for LinkedIn writing, because its strength is careful writing and matching a voice from examples. The free plan uses Claude Sonnet 5.5 and allows up to five Projects, enough for a voice file. Paid plans add Claude Opus 5.5. See Claude prompts for writing.
ChatGPT by OpenAI is the better pick for turning transcripts, recordings and slides into posts, since paid plans can take audio uploads directly and it handles files and images in the same chat. Perplexity is useful when a post needs a fresh, sourced fact, because every answer lists its sources. Grammarly works inside the apps and browser you already write in, which makes it handy for a final polish; its free plan includes 100 AI prompts a month.
A word on privacy: personal plans of ChatGPT, Claude and Perplexity can use your chats for training unless you switch it off, so do not paste client names or confidential numbers into a draft. For the full tool ranking, read the best AI for social media and ChatGPT vs Claude.
Prompts to copy
Mine your own work for post ideas
You are a LinkedIn ghostwriter interviewing a [job title] in [industry] to find post ideas. Ask me 10 questions, one at a time, about my work: mistakes I made, decisions I would make differently, things clients often get wrong, numbers from my projects I can share, and opinions I hold that peers disagree with. Wait for my answer before asking the next question. After the tenth answer, turn my answers into 10 post ideas, each with a one-line angle, the specific story or number from my answers it uses, and who it is for.
You get: Ten post ideas drawn from your real experience instead of generic topics
Turn a week of work into post ideas
Here is what happened in my work this week, in rough notes: <notes> [meetings, problems solved, questions people asked me, something I read, a number I saw] </notes> I post on LinkedIn for [audience, e.g. finance leaders at mid-sized companies] about [themes]. Find 5 post ideas in these notes. For each one give: the angle in one sentence, the format that suits it best (story, lesson list, contrarian take, how-to, behind the scenes), the opening line, and why my audience would care. Skip anything confidential: [what must stay private].
You get: Five ideas from your week, each with a format and an opening line
Find a sharp opinion you can defend
I work as [role] and my audience is [audience]. Here are three beliefs common in my field: [belief 1], [belief 2], [belief 3]. For each belief, give me the strongest honest case against it, using reasoning rather than invented statistics. Then ask me which one I actually disagree with. When I answer, help me write the argument as a 200-word post that states my view in the first line, gives one concrete example from my experience (ask me for it), acknowledges the best counterpoint, and ends with a question that invites people who disagree.
You get: A contrarian post you can stand behind, built from your own example
Write a story post from a real moment
You are a LinkedIn writer whose posts read like a conversation, not a press release. Write a post about this moment from my work: What happened: [the situation]. What I felt or thought: [reaction]. What changed: [outcome]. The lesson: [lesson]. Who it is for: [audience]. Rules: start with the most surprising moment, not the background. Short paragraphs of one to three lines. 150 to 250 words. No hashtags in the body, no emojis, no 'Here's the thing' or 'Let that sink in'. End with a specific question, not 'Thoughts?'. Here is one of my posts whose tone I like: <example> [paste] </example>
You get: A story post in your tone that starts at the interesting moment
Write a practical how-to post
Write a LinkedIn post that teaches [audience] how to [specific task] in [number] steps. I have done this [context, e.g. on 30 client projects], and the step people usually skip is [step]. Structure: a first line that names the result and a specific number or time saved I can back up ([number]), one sentence of context, the numbered steps with one line each, the mistake to avoid, and a closing line telling the reader what to try first. Under 220 words. Plain language, no jargon such as 'leverage' or 'synergy'.
You get: A useful, save-worthy how-to post built on your own method
Write a post about a result without bragging
I want to share [achievement, e.g. our team cut onboarding time from 6 weeks to 3] on LinkedIn without sounding like a humblebrag. Context: [what we did, who helped, what went wrong along the way]. Audience: [audience]. Write a 180-word post that puts the problem first, credits the people involved by role (I will add names), shares one thing that did not work, gives the result with the number, and ends with the one lesson others can use. Then give me two alternative first lines.
You get: A result post that is generous and useful, plus two alternative openings
Write 10 first lines and pick the best
Here is my LinkedIn post draft: <draft> [paste] </draft> The first line decides whether people read on. Write 10 alternative first lines, under 15 words each, using different approaches: a specific number, a surprising admission, a direct claim, a short scene, a question the reader has asked themselves, and a before and after. Do not use clickbait, false urgency or claims the post does not support. Rank your top three for [audience] and say why in one line each.
You get: Ten honest opening lines ranked for your audience
Edit a draft so it sounds like you, not AI
<draft> [paste draft, possibly written with AI] </draft> <my_writing> [paste two posts you wrote yourself] </my_writing> Edit the draft to sound like my writing samples. Remove phrases that sound machine-written, such as 'fast-paced world', 'level up', 'here's the thing', 'let that sink in', groups of three adjectives and every em dash. Replace general statements with specifics from the draft, and if none exist, mark [ADD EXAMPLE]. Keep it under [number] words. Return the edited post, then list what you changed.
You get: A draft cleaned of AI tells and matched to your real voice
Get honest feedback from your target reader
Act as [specific reader, e.g. a head of HR at a 300-person company who scrolls LinkedIn between meetings]. Read this post: <draft> [paste] </draft> Answer as that reader: Would you stop scrolling at the first line, yes or no, and why? Where would you stop reading? What is the one thing you would remember? Would you comment, and what would you say? What would make you save or share it? Then step out of the role and suggest the two edits with the biggest impact.
You get: A reader's-eye reaction and the two edits most likely to improve the post
Turn a long piece into a short series
<source> [paste article, talk transcript, podcast transcript or report] </source> Turn this into 5 LinkedIn posts for [audience], to be published over two weeks. Each post must stand alone, use a different specific idea, quote or number from the source, and use a different format (story, list, opinion, how-to, question). Put the source location of each idea in brackets after the post. 120 to 200 words each. Do not add facts that are not in the source.
You get: Five standalone posts from one asset, each traceable to the source
Write thoughtful comments on other people's posts
Here is a LinkedIn post by someone in my field: <post> [paste] </post> I am a [role] with experience in [area]. Write three possible comments, each under 60 words: one that adds a specific example from my experience (ask me for it if needed), one that respectfully disagrees with a point and gives a reason, and one that asks a sharp follow-up question. No 'Great post!' or flattery, and nothing that pitches my company.
You get: Three substantive comments that start conversations rather than add noise
Plan a month of posts around your themes
I post on LinkedIn [number] times a week for [audience]. My three themes are [theme 1], [theme 2], [theme 3]. My goal is [goal, e.g. inbound consulting enquiries]. Things happening this month: [events, launches, holidays]. Create a 4-week calendar as a table with columns: week, day, theme, format, working title, the personal story or data it needs from me, and the call to action. Balance the themes and formats, keep at most one promotional post in four, and list the material I need to collect before week 1.
You get: A balanced four-week calendar and a list of what to gather before you start