You paste a topic into ChatGPT. You get 300 words of fluffy LinkedIn copy. You delete it and write the post yourself anyway. Then you do it again next week. Most people are stuck in this cycle, and the problem is not the AI model. The problem is the prompt.
Generic prompts produce generic output. LinkedIn punishes generic output more than almost any other platform, because readers scroll fast and forgive nothing. A post that sounds like a corporate memo gets ignored in under five seconds. The gap between content that gets traction and content that disappears comes down to what you put into the prompt before you ask for anything.
This article gives you the structure, the templates, and the revision prompts to stop rewriting AI drafts from scratch.
The real gap
The gap between a weak prompt and a strong one is not about the AI model. It is about the information you give it before you ask for anything. Better input produces better output, every time.
Why LinkedIn prompts fail by default
Three structural failure modes that produce generic output
AI models train on massive amounts of LinkedIn content. Most of that content is mediocre. When you prompt vaguely, the model regresses to the average of its training data. You get the median LinkedIn post, which means hollow affirmations, bullet lists of obvious advice, and a sign-off about lessons learned.
There are three failure modes that cause this. Missing voice means the AI has no signal about how you actually write. Missing context means it has no raw material to work with, only a topic. Missing constraints means it fills every gap with its own defaults, and its defaults come from the worst of LinkedIn.
Before you look at solutions, diagnose which failure mode is costing you the most. Most people have all three.
The thought leadership trap
Asking AI to 'sound like a thought leader' produces the exact tone most LinkedIn readers scroll past. Thought leadership comes from specific experience, not style instructions. Tell the AI what you know, not how you want to sound.
The anatomy of a high-quality LinkedIn prompt
Five components that separate effective prompts from weak ones
Every high-performing LinkedIn prompt contains five components. Leave any one out and the AI fills the gap with a default. Those defaults are the reason your drafts sound like everyone else.
The five components are Role, Context, Constraint, Audience, and Output format. Each one does a specific job. Together they give the AI enough signal to produce a draft worth editing rather than a draft worth deleting.
Role
Who the AI is writing as. This anchors voice, experience level, and credibility. Be specific about years, team size, and industry.
Context
The raw material. Give the AI actual facts, results, and specifics to work from. A topic is not context.
Constraint
What to avoid. Explicit constraints override the AI's defaults. Name the formats and phrases you do not want.
Audience
Who reads it. Name the job title, company size, and the specific problem they care about right now.
Output format
Structure and length. Give exact word counts and structural rules. Vague length instructions produce bloated drafts.
From raw experience to a draft worth editing
Raw input
Your experience, data, or story
Role + context
Who you are and what happened
Constraint + audience
What to avoid and who reads it
Output format
Structure and word count
Draft
Ready for editing, not rewriting
Building your voice brief
A reusable document that anchors every prompt to your actual writing style
A voice brief is a 150-250 word description of how you write, not how you want to write. It comes from analyzing real posts, not aspirations. You paste it into every prompt, and it replaces the vague instruction to 'write in my voice.'
The brief works because it gives the AI specific patterns to replicate: sentence length, how you open posts, whether you use data or anecdotes, what you avoid. Without it, the AI guesses. With it, the AI has a template.
Build the brief once. Update it every six months or when your writing style shifts.
Pull your 10 best-performing posts
Go back 12 months and find the posts with the most comments and shares. Performance matters here, not your personal favorites.
Read them aloud and note the patterns
Listen for sentence length, whether you use questions, how you open posts, and whether you lean on data or stories.
Ask AI to describe the writing style
Paste 5 of your posts into the prompt below and ask for a style analysis. Do not ask it to describe your topics, only your style.
Edit the AI's description until it sounds accurate
Remove anything generic. If the description could apply to any professional on LinkedIn, cut it.
Save it as a reusable block
Store the final brief in a notes app or doc. Paste it into every prompt from this point forward.
Generate your voice brief
Claude / GPT-4Here are 5 LinkedIn posts I have written. Analyze the writing style and produce a 200-word style guide I can paste into future prompts. Focus on: sentence length and rhythm, how I open posts, whether I use data or anecdotes more, my tone (direct/warm/dry/etc), what I avoid, and any phrases or structures that repeat. Do not describe what topics I cover. Only describe how I write. [Paste your 5 posts here]
Prompt templates for the four main LinkedIn post types
Ready-to-use prompts for the formats that perform best in B2B
Four post formats cover roughly 80% of high-performing B2B LinkedIn content: the insight post, the story post, the contrarian take, and the how-to post. Each template below uses the five-component structure from Section 3.
These are starting points. Every draft needs editing. The templates cut the editing time by giving the AI a specific outcome to write toward rather than a topic to write about.
Insight post prompt
Claude / GPT-4You are writing as [NAME], a [ROLE] with [X] years in [INDUSTRY]. Voice brief: [PASTE VOICE BRIEF]. Write a LinkedIn post sharing one counterintuitive insight from [SPECIFIC EXPERIENCE OR PROJECT]. The insight should be specific enough that someone could disagree with it. Open with the insight as a direct statement, not a question. Follow with 2-3 sentences of evidence or context. Close with one sentence that tells the reader what to do with this information. No bullet points. No sign-off. Under 200 words.
Story post prompt
Claude / GPT-4You are writing as [NAME], a [ROLE]. Voice brief: [PASTE VOICE BRIEF]. Write a LinkedIn post about this specific situation: [DESCRIBE WHAT HAPPENED, INCLUDING THE PROBLEM, WHAT YOU DID, AND WHAT RESULTED]. Write it in past tense. Open with the moment of tension or the unexpected result, not background. Keep the narrative tight. End with one transferable lesson stated plainly. No inspirational framing. No 'this reminded me that.' Under 250 words.
Contrarian take prompt
Claude / GPT-4You are writing as [NAME]. Voice brief: [PASTE VOICE BRIEF]. Write a LinkedIn post that pushes back on this common belief in [INDUSTRY/FUNCTION]: [STATE THE BELIEF]. My actual position is: [STATE YOUR POSITION]. Support it with [DATA POINT or SPECIFIC EXAMPLE]. Write the opening line as a direct statement of your position. Do not soften it with 'I might be wrong' or 'this is just my take.' Anticipate one obvious objection and address it in one sentence. Under 220 words.
How-to post prompt
Claude / GPT-4You are writing as [NAME]. Voice brief: [PASTE VOICE BRIEF]. Write a LinkedIn post explaining how to [SPECIFIC TASK] for [SPECIFIC AUDIENCE]. This should be practical enough that someone could act on it today. Use a numbered list only if the steps must happen in sequence. Otherwise write in prose. Do not open with 'Here is how to' or 'In this post.' Open with the result the reader will get. Under 280 words. No filler closing line.
Outcome-first prompting
Each of these templates works because it gives the AI a specific outcome to write toward, not a topic to write about. Outcome-first prompting cuts editing time by roughly half. The AI knows what the post needs to accomplish before it writes the first word.
What to do when the first draft is wrong
Targeted follow-up prompts fix specific problems faster than rewriting from scratch
Most people either accept a weak draft or start over. Both waste time. A weak draft tells you exactly what is wrong. Use that information to write a targeted follow-up prompt.
Vague revision requests produce vague revisions. 'Make it better' gives the AI no direction. A specific instruction like 'the opening buries the point, rewrite it so the first sentence states the main claim directly' gives it a clear target.
Self-audit prompt for any draft
Claude / GPT-4Read this LinkedIn post draft and identify: 1. Any sentence that could appear in any industry without changing its meaning. 2. Any claim that is not supported by a specific fact or example in the post. 3. Any word or phrase that softens a point that should be stated directly. List each issue with the exact sentence and a suggested fix. [PASTE DRAFT HERE]
Common AI draft problems and what causes them
Why AI defaults to bullet lists: Training data on LinkedIn skews heavily toward listicle formats. List posts historically got high engagement, so they are overrepresented in the data the model learned from. Explicit constraints in your prompt are the only reliable fix.
Why it adds motivational sign-offs: Phrases like 'what would you add?' and 'drop a comment below' appear in thousands of high-engagement posts in the training data. The model associates them with successful posts. Name them explicitly in your constraints to block them.
Why it uses passive voice: Formal writing patterns in training data, especially content from corporate communications and press releases, bias the model toward passive constructions. Instruct it to use first person and active voice in every prompt.
Why it hedges claims: Safety tuning in models like GPT-4 and Claude makes them reluctant to state things definitively. For contrarian or direct posts, explicitly tell the model not to soften the main claim. You can always add nuance yourself in editing.
Prompting for a content series, not single posts
How to build a repeatable content engine instead of prompting post by post
Prompting reactively means you write a post when you need one. Prompting systematically means you build a content engine that produces consistent output with less effort each week.
The difference is a series brief. A series brief sits above individual post prompts and defines the audience, the core argument your content makes over time, and the post cadence. Every individual post prompt inherits from it. This keeps your content coherent across weeks and months, not just within a single post.
Your voice brief anchors the style. Your series brief anchors the strategy. Individual post prompts handle execution. Each layer does one job.
LinkedIn content system architecture
Voice brief
Anchors style across all posts
Series brief
Anchors strategy across all posts
Individual post prompts
Execution layer
Series brief prompt
Claude / GPT-4I am building a LinkedIn content series for [TIMEFRAME, e.g. 90 days]. Here is the context: - My role: [TITLE, COMPANY TYPE, YEARS OF EXPERIENCE] - My audience: [JOB TITLE, COMPANY SIZE, SPECIFIC PROBLEM THEY HAVE] - The core argument my content makes: [ONE SENTENCE, e.g. 'Most SaaS onboarding fails because it optimizes for activation, not habit formation'] - Post frequency: [e.g. 3x per week] - Post formats I want to rotate: insight, story, contrarian, how-to Voice brief: [PASTE VOICE BRIEF] Produce a 12-post content plan with: post type, the specific angle for that post, and the one claim or story it builds on. Do not write the posts. Only produce the plan. Each entry should be one sentence describing the specific angle, not the topic.
