You'll build a three-step workflow: first, use Apify to extract your LinkedIn post history into a structured dataset. Second, feed that data to Claude with a style analysis prompt that identifies your patterns, tone, and structure. Third, save that analysis as a custom Claude Project or instruction set you can reuse whenever you need to generate new posts that sound exactly like you.
This approach works because you're training the AI on your actual successful content, not asking it to imitate a generic "professional" voice. The result is new posts that match what already resonates with your audience.
What Is Training Claude on Your LinkedIn Writing Style?
Training Claude on your writing style means feeding it a collection of your existing LinkedIn posts and asking it to extract patterns in how you structure ideas, choose words, and express your perspective. You're not fine-tuning the model in the technical sense. You're creating a reusable instruction set that tells Claude how to mimic your voice.
The process relies on Claude's context window (200,000 tokens for Claude 3.5 Sonnet) to absorb dozens of your posts at once. You provide examples, Claude analyzes them, and you save that analysis as a custom prompt you can use repeatedly. Think of it as creating a style guide that an AI can follow instead of a human editor.
Apify handles the extraction piece. It's a web scraping platform that can pull your LinkedIn post history without manual copy-pasting. You'll get clean, structured data (usually JSON or CSV) that's ready to feed into Claude. The entire workflow takes about 2 to 3 hours to set up the first time, then maybe 15 minutes per month to maintain.
Why Training AI on Your Own Content Beats Generic Templates
Generic AI-generated posts sound generic because they're based on the average of millions of training examples. When you ask Claude to "write a professional LinkedIn post about leadership," it produces something that could have been written by anyone in any industry. Your audience can tell.
Posts trained on your specific style perform better because they maintain the quirks and perspectives that made your original content work. If you tend to open with questions, use short paragraphs, or reference specific frameworks, Claude will replicate those patterns. One user reported a 34% higher engagement rate on AI-assisted posts after implementing custom style training compared to generic prompts.
You're also teaching Claude which topics and formats already succeeded with your audience. If your how-to posts consistently outperform your opinion pieces, that signal gets baked into the training data. The AI learns not just how you write, but what works for you specifically.
This matters for personal brand consistency too. Publishing 3 to 5 times per week and maintaining a consistent voice without AI help? Exhausting. With a trained Claude skill, you can draft faster while sounding more like yourself, not less.
How to Extract LinkedIn Posts to Train Your AI Writing Assistant
You'll use Apify's LinkedIn scraper to pull your post history. Start by creating a free Apify account (the free tier includes $5 of platform credits, enough for roughly 500 to 1000 posts depending on the scraper you choose). Search the Apify Store for "LinkedIn Profile Scraper" or "LinkedIn Posts Scraper."
The scraper needs your LinkedIn profile URL as input. You'll also specify how many posts to extract. Start with your last 50 to 100 posts if you publish regularly, or go back 6 to 12 months if you post less frequently. More data isn't always better. Focus on your best-performing content.
Setting Up the Apify Scraper
Once you've selected a scraper, configure these settings in the Apify console. Set the profile URL to your own LinkedIn profile. Set max posts to 100 (you can adjust this later). Enable "Include post text" and "Include engagement metrics" if the scraper supports it. You want likes, comments, and shares data to identify your top performers.
Run the scraper. It'll take 5 to 15 minutes depending on how many posts you're extracting. Apify will show you a progress bar and notify you when it's complete. Download the results as a JSON file (CSV works too, but JSON preserves formatting better).
Cleaning and Filtering Your Post Data
Open the JSON file and look for posts with the highest engagement. You don't need to feed Claude every post you've ever written. Focus on the top 20 to 30 based on total engagement (likes plus comments plus shares). This ensures you're training on what actually worked, not your experimental failures.
Create a new text file and copy just the post text from your top performers. You can include engagement numbers as metadata if you want Claude to understand which posts resonated most. Format it like this:
POST 1 (245 likes, 32 comments):
[Full post text here]
POST 2 (189 likes, 28 comments):
[Full post text here]
POST 3 (167 likes, 19 comments):
[Full post text here]
Keep the total under 150,000 tokens to leave room for Claude's analysis. That's roughly 30 to 40 typical LinkedIn posts. If you're not sure about token count, paste your compiled posts into Claude and ask "How many tokens is this?" before proceeding.
How to Make Claude Write Like You on LinkedIn
Now you'll feed your compiled posts to Claude with a style analysis prompt. Open a new Claude conversation (use Claude 3.5 Sonnet for this, not Haiku). Paste your collection of posts, then add this prompt:
I'm providing 30 of my best-performing LinkedIn posts below. Please analyze them and create a detailed style guide that captures:
1. Sentence structure patterns (average length, use of questions, punctuation style)
2. Paragraph length and formatting preferences
3. Tone and voice (formal vs casual, use of humor, personal vs professional)
4. Common opening and closing patterns
5. Vocabulary choices and phrases I use frequently
6. How I structure arguments or narratives
7. Topics and angles that appear most often
Format your analysis as a reusable instruction set I can give you in future conversations when I want you to write new posts in my style.
[PASTE YOUR POSTS HERE]
Claude will generate a comprehensive style guide, usually 800 to 1200 words. It'll identify patterns you might not have noticed yourself. Save this output in a separate document. This is your reusable custom instruction set.
Creating a Custom Claude Project for LinkedIn Posts
Claude Projects let you save custom instructions and knowledge that persist across conversations. Create a new Project in Claude (available on Pro plans and above). Name it something like "LinkedIn Post Generator" or "My Writing Style."
In the Project's custom instructions field, paste the style guide Claude just created. Add this at the end:
When I ask you to write a LinkedIn post, follow this style guide exactly. Match my sentence structure, tone, formatting, and typical post length. Don't add flourishes or vocabulary I wouldn't use. Sound like me, not like a generic AI.
You can also upload your original posts as Project knowledge files. This gives Claude the full context to reference if needed. The combination of explicit instructions plus example posts creates surprisingly accurate style matching.
If you're on the free Claude tier and can't use Projects, save your style guide as a text file. Start every new conversation by pasting it with "Follow this style guide for all posts in this conversation." It's less convenient but works just as well.
Testing Your Custom Claude Writing Skill
Try generating a few test posts. Give Claude a topic and ask it to write in your style: "Write a LinkedIn post about the importance of testing AI outputs before publishing. Use my style guide." Compare the output to your actual posts. Does it match your paragraph length? Your tone? Your typical structure?
You'll probably need to iterate. If Claude's posts are too formal, add "Use more casual language and contractions" to your instructions. If they're too long, specify "Keep posts to 150 to 200 words maximum." Refine the style guide based on what you see. This is similar to how you might build AI agents that critique and improve work, except you're the critic in this loop.
How to Automate LinkedIn Content Creation with Claude AI
Once your style guide is working, you can streamline content creation. Set up a simple workflow: every Monday, brainstorm 3 to 5 post topics based on what's happening in your industry or what questions you're hearing from clients. Feed those topics to Claude in your custom Project. Review and edit the outputs. Schedule them throughout the week.
You should still edit every AI-generated post. Claude will get your style close, but you'll want to add current examples, adjust for recent events, or inject fresh perspectives. Plan to spend 5 to 10 minutes per post on edits rather than 20 to 30 minutes writing from scratch. That's where the time savings come from.
Some people batch this process: generate 10 to 15 post drafts in one sitting, save them to a content calendar, then edit and publish them over the next few weeks. Others prefer daily generation to keep content timely. Both approaches work. Pick what fits your workflow.
Track which AI-assisted posts perform well and add them back to your training data every few months. Your writing evolves, and your style guide should too. Re-run the Apify scraper quarterly, analyze your new top posts, and update Claude's instructions. This keeps the AI current with how your voice and topics shift over time.
Creating a Custom Claude AI Skill from Your Content
You can extend this workflow beyond LinkedIn posts. The same style analysis approach works for email newsletters, blog posts, or even Slack messages. Extract samples of any writing you do regularly, run the style analysis, and create task-specific Projects.
For example, you might create separate Projects for "LinkedIn posts," "Email newsletters," and "Long-form blog posts" because your style probably shifts between those formats. Your LinkedIn voice might be punchier and more casual than your whitepaper voice. Training Claude on each format separately gives you better results.
You can also train Claude on specific content types within LinkedIn. If you write both how-to posts and personal story posts, create two style guides. When you need a how-to, use that Project. When you need a story, switch to the other. This level of specificity takes more setup time but produces noticeably better outputs.
Look, the key is treating this as a skill you're building, not a one-time prompt. You're creating a reusable asset that gets better the more you refine it. Think of it like fine-tuning AI models without the full cost, except you're using prompt engineering and context instead of training runs.
Teaching AI to Match Your Writing Voice Over Time
Your style guide will drift out of sync if you don't maintain it. Set a calendar reminder every 3 months to re-scrape your LinkedIn posts, identify new top performers, and check if your style has shifted. Maybe you're writing longer posts now, or you've started using more data and fewer anecdotes. Update your instructions to reflect that.
You can also A/B test different versions of your style guide. Create two Projects with slightly different instructions (one more formal, one more casual) and see which generates posts that perform better. After a month, double down on the winner. This iterative approach mirrors how you'd test any content strategy.
Watch for Claude falling back into generic AI patterns. If you see phrases like "in conclusion" or "it's important to note" creeping into outputs, add explicit instructions to avoid them. Your style guide should include a "never use these phrases" section based on your personal pet peeves and your audience's preferences.
Some users find it helpful to include negative examples in their training data. Show Claude a few generic, poorly-performing posts and label them as "not my style." This gives the AI clearer boundaries. Small detail, but it can reduce the number of edits you need to make per post by 20 to 30%.
The workflow becomes second nature after a few weeks. You'll stop thinking about the mechanics and start using Claude as a genuine writing partner that knows your voice. That's when the real productivity gains kick in: you're not fighting the AI to sound like yourself, you're collaborating with a tool that already gets it.
Building this system takes a few hours upfront, but it pays dividends every time you need to create content. You'll write faster, maintain consistency across hundreds of posts, and spend your creative energy on ideas instead of sentence construction. Start with your top 30 posts, create your first style guide this week, and refine it as you go. The sooner you build this skill, the sooner you'll wonder how you ever managed without it.
Get a free AI-powered SEO audit of your site
We'll crawl your site, benchmark your local pack, and hand you a prioritized fix list in minutes. No call required.
Run my free audit