Content at Scale: How to Train AI to Create Great Content

Struggling to keep up with content creation? Want to create more content in less time while maintaining consistency?

In this article, you’ll discover a practical strategy anyone can implement to create authentic content at scale with AI.

Audra's Background: Why Her Approach Matters

With more than two decades of experience running a marketing agency since 2009, Audra Carpenter has witnessed the evolution of content marketing firsthand. Her journey into AI began during the early days of Jasper (originally called Conversion AI, then Jarvis) when she recognized the potential of these emerging tools despite their initial limitations.

Today, Audra views AI as a golden ticket that enables marketers to build better systems, achieve greater consistency, and create comprehensive marketing ecosystems rather than siloed approaches. This perspective is particularly valuable for small businesses that have traditionally faced bandwidth limitations when trying to grow their products and services.

A Real-World Experiment: How to Build an AI-Generated Content Creation System to Double a Social Media Following in 90 Days

Despite her extensive experience helping clients with marketing, Audra had never focused on building an audience for her personal brand.

“I've gotten to a place where I'm recognizing that as we evolve in the next couple of years, it's not going to be so much about our services. It's not going to be so much about our products as it is going to be about the audience that we serve,” she explained.

This realization led her to embark on a personal challenge: Double her social media following in 90 days using AI tools to drive content creation.

Her comprehensive system, guided by Claude's advice, is something anyone can adapt to their content creation needs.

#1: Create a Strategic Plan

To ensure your AI efforts align with meaningful business objectives rather than simply creating content for content's sake, you need to establish a strategic approach with clear goals and metrics.

• Goal: To grow a social media audience from 25k to 50k
• KPIs to Measure Results: Track follower growth, engagement, and comments
• Initial Platform Focus: Start with LinkedIn

With this foundation in place, she created a Claude Project and added the following Project Knowledge to train it.

#2: Research Social Media Profiles for Successful Content Examples 

At the heart of Audra's approach is what she calls a training database – a collection of successful content examples that AI can learn from to generate better results. This data-driven method moves beyond relying solely on personal experience or intuition.

Audra spent 4-8 hours researching LinkedIn profiles with successful content to build her LinkedIn database. 

Audra didn't want to sound like everyone else in the marketing space; she wanted to create her own voice, so instead of looking at content only from other marketers, she also examined content across different industries, including coaches, service providers, and venture capitalists, looking for profiles with posts that performed well.

She looked at content from people with 40k-300k followers, avoiding mega-influencers such as Gary Vaynerchuk and Richard Branson because the volume of their audiences can skew the response to their content.

She ultimately settled on about 15 profiles.

#3: Assemble Your Training Database

Create and Populate a Structured Content Database

Next, Audra built a detailed Google spreadsheet with columns for:

• The actual post (copied exactly as posted, including emojis, formatting, and spacing)
• Date posted
• Number of likes
• Number of comments
• Links to the creators' profiles
• Links to any images used
• Content category (how-to posts, mindset posts, etc.)
• Target audience
• Intentional message/summary

She then copied and pasted 100 high-performing posts from her selected profiles into the Google sheet precisely as they appeared on LinkedIn, retaining emojis, formatting, etc., and filled in all fields except the last three columns.

Then, she pasted each post into Claude and asked it to analyze the content and provide notes on the final three additional fields. Audra then pasted those notes into the appropriate field in the spreadsheet.

You are an expert marketing strategist on LinkedIn. You know what a successful LinkedIn post looks like. I want to create a project to create LinkedIn content that will grow my audience from 25k to 50k in 90 days. I don't know how to create content for LinkedIn, but I have 100 examples of LinkedIn content that I think performs well. I want you to analyze these posts and note: 1. The content category (how-to posts, mindset posts, etc.) 2. The target audience 3. The intentional message/summary.

This structured approach gave her AI assistant a comprehensive understanding of what makes LinkedIn content successful and provided a framework for generating similar high-performing content.

Pro Tip: An important insight she gained during this research is that not all of her curated posts were mind-blowing. They were leaning into what the algorithm is rewarding, which makes studying what's actually working on a platform rather than simply creating what you think is good content will give you better results.

Detail Your Personalized Voice Guidelines

For AI content to truly represent you, it needs more than just your writing style – it needs to understand your experiences, expertise, and personality. Audra created detailed documentation that included:

• Standard brand voice guidelines
• Her professional experience
• Her personality, likes, and dislikes
• Her values and causes

For those who struggle to document their voice, Audra suggests:

• Recording a conversation with a friend or family member who interviews you
• Using ChatGPT's voice mode to have a natural conversation where the AI interviews you
• Transcribing these conversations to create reference material for your AI

This personalization ensures that the content the AI produces for you feels authentic rather than generic, which is crucial for building genuine audience connections.

#4: Upload the Training Data to Claude

Audra downloaded the training database and personalized voice guidelines as a PDF and uploaded it to her Claude Project. Though this took up about 60% of Claude's memory capacity, it gave the AI a comprehensive understanding of successful LinkedIn content patterns.

Audra initially attempted to build a fine-tuned OpenAI model with this data, but she encountered limitations because OpenAI only accepts three fields: the user field, the messaging, and the system prompt.

After reformatting and uploading her data, she was disappointed with the results.

This experience led her to prefer keeping her database within Claude, where she found more control over the output quality.

#5: Generate Content with Minimal Effort

With all this preparation complete, Audra can now create LinkedIn posts with remarkable efficiency–often as little as two minutes.

Good morning, Claude. Today, we're going to write a post on AI and Customer Service. 

Or

Good morning, Claude. Today, we're going to write a post about my interview with Michael Stelzner on the AI Explored podcast.

Because Claude has access to all her reference materials, the output is typically high-quality and authentic-sounding, and minimal follow-up prompting is required.

According to her, the results have been substantial:

• Significant follower growth
• More comments than she's ever had in 12 years on LinkedIn
• Invitations to speak at two events
• An invitation to be on Michael Stelzner's podcast

Creating Content for Other Social Media Platforms

Audra notes that what works on one platform may not transfer directly to another. When she tried using this LinkedIn Content Project to create content for Facebook, she didn't see the same growth results.

You'll need to create separate Projects for each platform to adapt its content training data to that platform's unique algorithm and audience behaviors.

Tips and Tools to Implement Advanced Content Creation Automation

For those ready to take their system to the next level, Audra has the following automation tools:

• N8N: An automation platform (similar to Zapier and Make but more technical) that connects various AI services
• Relevance AI: A tool for building AI agents that can perform specific tasks within a workflow.
• Open Router: A dashboard that provides access to multiple AI models (OpenAI, Claude, Perplexity, Gemini) through a single API key, allowing her to assign different tasks to different models. With Open Router, Carpenter can assign specific tasks to different AI models. Perplexity does her research, OpenAi does her product descriptions, and Claude writes her blog content.

These two tools allow her to create powerful automation, such as a LinkedIn scraper that automatically collects and adds new content from her selected creators to her content database spreadsheet.

• Content Hub OS: Audra’s platform allows users to take one piece of content and automatically create multiple assets. This tool promises to further streamline the content creation process by enabling true cross-platform distribution from a single source.

Other Notes From This Episode

• Connect with Michael Stelzner @Stelzner on Facebook and @Mike_Stelzner on X.
• Watch this interview and other exclusive content from Social Media Examiner on YouTube.

Where to subscribe: Apple Podcasts | Spotify | YouTube Music | YouTube | Amazon Music | RSS

✋🏽 If you enjoyed this episode of the AI Explored podcast, please head over to Apple Podcasts, leave a rating, write a review, and subscribe.

---

Stay Up-to-Date: Get New Marketing Articles Delivered to You!

Don't miss out on upcoming social media marketing insights and strategies! Sign up to receive notifications when we publish new articles on Social Media Examiner. Our expertly crafted content will help you stay ahead of the curve and drive results for your business. Click the link below to sign up now and receive our annual report!