How to Build an AI Agent That Writes LinkedIn Posts in Your Voice

Prerequisites

Before you start, you need three things: a Claude account (free tier works, but Pro gives you longer conversations and Projects), five to ten LinkedIn posts you have already published that represent your best writing, and a clear idea of who you are writing for. That last one matters more than people think. An agent that writes for everyone writes for no one.

Step 1: Define Your Voice Profile

Most people skip this step and jump straight to prompting. That is why most AI-generated LinkedIn posts sound identical. Your voice profile is the foundation. Without it, you are building on sand.

Open a new Claude conversation and paste five to ten of your best LinkedIn posts. Then use the following prompt:

I'm going to share several LinkedIn posts I've written. Analyze them
and create a detailed Voice Profile document that captures:

1. TONE: Am I formal, conversational, provocative, reflective?
   Give me a spectrum, not a single word.

2. STRUCTURE: How do I typically open? Do I use short paragraphs?
   Do I ask questions? How do I close?

3. VOCABULARY: What words and phrases do I gravitate toward?
   What do I avoid? Do I use jargon or plain language?

4. SIGNATURE MOVES: What patterns make my writing recognizably
   mine? (e.g., opening with a contrarian statement, using
   numbered lists, ending with a question)

5. TOPICS & ANGLES: What themes recur? How do I typically
   approach a topic — through personal experience, data,
   analogy, or opinion?

6. ANTI-PATTERNS: What does generic AI-written LinkedIn content
   look like, and how does MY writing differ from that?

Format this as a structured reference document I can reuse.

Here are my posts:

[PASTE YOUR POSTS HERE]

Claude will return a structured document. Read it carefully. Does it actually sound like you? If something feels off, push back. Say: “That’s not quite right. I’m not really conversational. I’m more direct and dry. Think less motivational speaker, more engineer explaining a system.” This refinement is part of the process.

Save the final voice profile. You will use it in every prompt going forward.

Step 2: Build the System Prompt

The system prompt is the instruction set that turns Claude from a general-purpose AI into your personal LinkedIn writing agent. Think of it as the agent’s job description. Here is a production-ready template:

You are a LinkedIn content agent for [YOUR NAME], a [YOUR ROLE]
at [YOUR COMPANY/INDUSTRY].

## Voice & Style
[PASTE YOUR VOICE PROFILE FROM STEP 1 HERE]

## Audience
Primary: [e.g., Mid-career professionals in B2B SaaS]
Secondary: [e.g., Founders exploring AI for operations]
What they care about: [e.g., practical frameworks, not hype]
What they scroll past: [e.g., motivational platitudes, humble brags]

## Content Rules
- Never use these words/phrases: [e.g., "game-changer", "excited
  to announce", "thrilled", "leverage", "synergy"]
- Never start with a question unless it's genuinely provocative
- Keep posts between 150-300 words unless the format demands more
- Use line breaks for readability (LinkedIn is a mobile-first platform)
- End with something that invites genuine response, not engagement
  bait like "agree?"

## Output Format
For each post, provide:
1. The post text, ready to copy-paste into LinkedIn
2. A one-line hook analysis (why this opening works)
3. One suggested visual or media pairing if relevant
4. Three hashtag suggestions (industry-specific, not generic)

Where to put this prompt: If you are using Claude Pro, create a Project and add this as the Project instructions. This means every conversation inside that Project automatically uses your agent configuration. If you are on the free tier, you will need to paste it at the start of each new conversation, or save it in a text file for quick access.

Step 3: Structure Your Input

The power of a well-built agent is that your input can be minimal. You should not need to write a brief every time you want a post. Here is the input template that makes this work:

Write a LinkedIn post based on this:

TOPIC: [One sentence describing what the post is about]

ANGLE: [How you want to approach it: personal story, hot take,
framework, lesson learned, observation, or contrarian view]

TRIGGER: [What made you think about this? A meeting, an article,
a conversation, a mistake?]

FORMAT: [Short-form thought, numbered list, story with punchline,
or before/after comparison]

Here is a real example of this template in action:

TOPIC: Most companies are using AI to do the same work faster
instead of rethinking the work itself

ANGLE: Contrarian observation from my own team's experience

TRIGGER: Watched a colleague spend 2 hours getting AI to write
a report that shouldn't have existed in the first place

FORMAT: Short-form thought, under 200 words

Notice how little effort this requires from you. Four lines. The agent handles the rest: the opening hook, the structure, the voice, the close. Your job is to have the insight. The agent’s job is to articulate it in your voice.

Step 4: Create a Feedback Loop

This is where most AI content workflows fail. People generate a post, publish it, and never close the loop. The agent never learns what worked. Here is how to fix that.

After Every Post You Publish

Come back to your agent conversation (or Project) with a quick debrief:

Post performance update:

POST: [Paste the post you published, including any edits you made]

EDITS I MADE: [What did you change before publishing and why?]

PERFORMANCE: [Impressions, likes, comments, reposts — whatever
you track]

WHAT WORKED: [Did a specific line get quoted in comments? Did
people DM you about it?]

WHAT FELL FLAT: [Low engagement? Wrong audience? Felt forced?]

Update your understanding of my voice and preferences based on
this feedback. Tell me what you'll adjust going forward.

Claude will internalize this feedback within the conversation (or Project) and adjust. Over time, the agent develops a nuanced understanding of what works for your specific audience, not just LinkedIn in general.

Monthly Voice Recalibration

Once a month, run a recalibration prompt:

Review all the feedback I've given you over the past month.
Update my Voice Profile with any new patterns you've noticed:

- What types of posts consistently perform best?
- What openings get the most engagement?
- What topics does my audience respond to most?
- What should we stop doing?
- What should we experiment with next?

Give me an updated Voice Profile document I can use going forward.

This updated voice profile replaces the one in your system prompt. Each month, your agent gets sharper.

Step 5: Advanced Patterns

Once the basic loop is running, here are three ways to push it further.

Content Series Generator

Instead of one-off posts, ask the agent to plan a series:

I want to create a 5-post LinkedIn series on [TOPIC].
Each post should stand alone but build on the previous one.
The series should have a narrative arc:

Post 1: Surface the problem (make people feel seen)
Post 2: Challenge conventional thinking about it
Post 3: Share a framework or mental model
Post 4: Show a real example or case study
Post 5: Call to action or open question for discussion

Draft all five with hooks, body, and closes.

Repurposing Engine

Turn one piece of content into multiple formats:

I just published this blog post / gave this talk / had this
meeting. Extract 3-5 LinkedIn post ideas from it.

For each one, tell me:
- The core insight worth sharing
- The best format (story, list, hot take, framework)
- A draft of the post in my voice

[PASTE SOURCE CONTENT]

Engagement Response Drafts

Use your agent to draft thoughtful replies to comments on your posts or to write commentary on other people’s content:

Someone left this comment on my post: "[COMMENT]"

Draft a reply that:
- Acknowledges their point specifically (not generically)
- Adds something new to the conversation
- Stays in my voice
- Is under 50 words

The Complete Workflow at a Glance

Step	Action	Frequency
1	Build Voice Profile	Once (update monthly)
2	Set Up System Prompt	Once (refine as needed)
3	Input Idea → Get Draft	2–5 times per week
4	Edit, Publish, Debrief	After every published post
5	Monthly Recalibration	Once per month

Why This Works (And Why Most AI Content Doesn’t)

The difference between this approach and asking ChatGPT to “write me a LinkedIn post” comes down to three things.

Specificity. A generic prompt produces generic output. Your voice profile, audience definition, and content rules constrain the agent in the right ways. Constraints are not limitations. They are what make creative output focused and effective.

Memory. By using Claude Projects or maintaining a continuous conversation, the agent accumulates context about your preferences, your audience’s responses, and what works. This compounds over time in a way that starting fresh each time never will.

Feedback. The debrief loop is the entire game. Without it, you are using a static tool. With it, you have a system that adapts. After two months of consistent feedback, the agent will produce first drafts that require minimal editing.

What to Build Next

This blueprint covers LinkedIn posts, but the architecture is portable. Once you have your voice profile and system prompt working, you can adapt it for:

1. Email newsletters — Same voice, longer format, add a section for curated links
2. Twitter/X threads — Same voice, tighter constraints, add a hook-per-tweet rule (See Twitter / X Blueprint guide)
3. Internal comms — Same structure, different audience definition, more context on company culture
4. Thought leadership articles — Same voice profile, add a research/evidence requirement to the system prompt

Each of these is a separate blueprint. If you want to see them, let us know in the comments which one to build next.

---

Frequently Asked Questions

Q: Can an AI agent really write LinkedIn posts that sound like me?

Yes — if you give it enough examples of your writing, a clear prompt, and a review step. The strongest results come when the agent is trained on your past LinkedIn posts, your preferred structure, your favourite phrases, and the opinions you tend to repeat. It will not magically become you, but it can get close enough to save time while keeping your voice recognisable.

Q: What tools do I need to build this LinkedIn writing agent?

At minimum, you need an LLM for writing, a place to store examples of your past posts, and a simple automation layer that passes your instructions to the model. Many people also add a spreadsheet, database, or note-taking system for storing voice examples, plus an editing or approval step before anything is published. You can keep it lightweight at first and expand later.

Q: How do I train the AI on my writing style?

Start by collecting a strong sample of your best posts, then identify patterns in tone, structure, sentence length, hooks, and calls to action. Turn those patterns into explicit instructions, and include a few examples in the prompt or retrieval system. Then test the output, edit what feels off, and feed those corrections back into your style guide so the system improves over time.

Q: Will LinkedIn penalise AI-generated content?

What matters most is quality, originality, and usefulness — not whether you used AI somewhere in the workflow. If the posts are generic, repetitive, spammy, or clearly mass-produced, performance can suffer because people will ignore them. But if you use AI as a drafting assistant and make the final post insightful, relevant, and personal, you are far less likely to run into problems.

Q: How long does it take to set this up?

A simple version can be set up in an afternoon if you already have examples of your writing and know which tools you want to use. A more polished system — with prompt tuning, example libraries, approval steps, and platform-specific formatting — can take several days to refine properly. The first version is quick; the best version is iterative.

Q: Can I use this same system for Twitter/X or other platforms?

Yes. The core idea stays the same: teach the system your voice, define the content goal, and adapt the output format to the platform. For X, you might optimise for brevity and stronger hooks; for newsletters, you may want more depth and narrative flow; for blog posts, a clearer structure and SEO considerations matter more. One voice system can support multiple channels if you add channel-specific rules. (See Twitter / X Blueprint guide)