How To Build a Content Team of SEO AI Agents (n8n, OpenAI, Aidbase)

AI-Powered Blog Automation: A Deep Dive

Key Concepts:

• AI Agents
• Keyword Research
• Content Generation
• Internal Knowledge Base (RAG)
• Thumbnail Generation
• Automation Workflow (NAN)
• SEO Risk Assessment
• GPT-4
• Perplexity Sonar
• 8base
• Replicate API
• Flux

1. Introduction: The AI Blogging Experiment

The video explores the automation of blog content creation using AI agents for the presenter's products: Feed Hive, Link Drip, and Abbase. The goal is to create daily blog posts with minimal human intervention, covering research, writing, thumbnail creation, and publishing. The presenter acknowledges the potential risks associated with AI-generated content, particularly concerning Google's stance on low-quality, unhelpful content.

2. Google's Stance on AI Content

Google's policies state that they don't penalize content solely for being AI-generated. However, they do penalize low-quality, unhelpful content created primarily to manipulate search rankings. The key is to create helpful, high-quality, and relevant content, even if AI is used in the process.

3. Requirements for High-Quality AI-Generated Content

To avoid penalties and create valuable content, the presenter identifies several key requirements:

• Careful Research: Thorough research and fact-checking.
• Quoting Sources: Referencing other articles and sources to support claims.
• Unique Internal Knowledge: Incorporating proprietary data and insights not available elsewhere.
• Custom On-Brand Thumbnails: Creating visually appealing thumbnails that align with the website's design.

4. Keyword Research and Content Idea Generation

The video outlines two approaches to keyword research and content idea generation:

• Manual Keyword Research: The user performs keyword research and provides the AI with precise instructions. This offers more control but requires manual effort.
• AI-Driven Content Ideas: Using OpenAI's search model (GPT-4) to generate content ideas based on a description of the product and brand. This is fully autonomous but offers less control over specific keywords.

Tools Mentioned:

• SER API: An API version of Google that returns search results in JSON format.
• Perplexity Sonar: An AI search model that retrieves and synthesizes information from the internet.
• OpenAI's Search Model (GPT-4): Similar to Perplexity Sonar, used for finding content ideas.

5. Setting Up the Automation Workflow in NAN

The video provides a step-by-step guide to setting up the automation workflow using NAN (n8n.io), a no-code workflow automation platform.

Steps:

1. Create a NAN Account: Sign up for an account on NAN.
2. Add OpenAI Credentials: Connect NAN to your OpenAI account.
3. Create a New Workflow: Start a new workflow in NAN.
4. Add a Schedule Trigger: Configure the trigger to schedule the workflow execution (e.g., daily, weekly).
5. Add AI Nodes (OpenAI): Use the "Message a Model" node to interact with OpenAI's models.
   • Search Topic Ideas: Use GPT-4 to generate content ideas based on a system prompt describing the brand.
   • Prepare Topic: Use GPT-4 to prepare the topic, including title, description, and outline, in JSON format.
6. Prevent Duplicate Content:
   • Retrieve summaries of previous blog posts from Strapi (or your CMS).
   • Feed these summaries to the "Search Topic Ideas" AI agent to avoid generating duplicate content.

6. Enhancing Content Quality with Research and Internal Knowledge

The video emphasizes the importance of adding statistics, facts, references, and unique insights to the blog posts.

Steps:

1. Add a Research AI Agent: Use GPT-4 search or Perplexity Sonar to research the specific topic and find relevant facts and statistics.
2. Prepare an Internal Knowledge Base (RAG): Use 8base to create a knowledge base with private data and insights.
   • Sign up for 8base: Create a free trial account on 8base.ai (Pro plan required for API access).
   • Add Knowledge Sources: Add your website, help desk, and YouTube channel as knowledge sources.
   • Train AI Models: Train the AI models on the knowledge sources.
   • Create a Chatbot: Create a chatbot in 8base and add the trained knowledge to it.
   • Get Chatbot ID and API Key: Note down the chatbot ID and create an API key.
3. Integrate 8base with NAN:
   • Create a Generic Credential: Create a generic credential in NAN for 8base using the API key.
   • Add an HTTP Request Node: Use an HTTP request node to send a POST request to the 8base chatbot API.
   • Configure the Request: Set the URL, authentication, and body of the request.
4. Add an AI Agent to Retrieve Internal Knowledge: Use a chat model (e.g., GPT-4) to query the 8base chatbot and retrieve relevant internal knowledge.

7. Writing the Blog Post

After preparing the topic, researching, and gathering internal knowledge, the video moves on to writing the actual blog post.

Steps:

1. Add an AI Agent (OpenAI): Use a chain-of-thought model (e.g., GPT-3.5 Turbo) to write the blog post based on the prepared topic, research, and internal knowledge.
2. Provide a System Prompt: Give the AI agent a description of its task and instructions.
3. Provide a User Message: Provide the AI agent with all the necessary information to write the post.

8. Thumbnail Generation

The video explores two approaches to thumbnail generation:

• Flux (Replicate API): Using Flux on Replicate to generate thumbnails based on the image description created by one of the AI agents.
  • Create a Replicate Account: Sign up for an account on replicate.com.
  • Create an API Token: Create a new API token.
  • Create a Generic Credential: Create a generic credential in NAN for Replicate using the API token.
  • Add an HTTP Request Node: Use an HTTP request node to send a POST request to the Replicate API.
  • Configure the Request: Set the URL, authentication, and body of the request.
• Custom Thumbnail API (Vibe Coding): Creating a custom API using NodeJS and the Canvas library to generate thumbnails with a specific theme.

9. Publishing the Blog Post

The final step is to publish the blog post to your CMS (e.g., Strapi, WordPress, Webflow).

Steps:

1. Add an HTTP Request Node: Use an HTTP request node to send a POST request to your CMS API.
2. Configure the Request: Set the URL, authentication, and body of the request.
3. Automate Social Media Sharing: Use a trigger in Feed Hive (or your social media management tool) to automatically share the blog post on social media.

10. Risk Assessment and Conclusion

The presenter emphasizes that this automation workflow is experimental and potentially risky, especially for businesses that heavily rely on SEO. If SEO is a critical channel, it's recommended to carefully consider the risks before fully automating blog content creation. However, for businesses where blogs are not a primary focus, this approach can be a game-changer.

Key Takeaways:

• AI can be used to automate blog content creation, but it's crucial to ensure high quality and relevance.
• Google penalizes low-quality, unhelpful content, regardless of whether it's AI-generated.
• Incorporating unique internal knowledge and insights is essential for creating valuable content.
• Careful planning, research, and prompt engineering are crucial for successful AI-powered blog automation.
• The presenter offers access to Abbase and Feed Hive, as well as the entire workflow, through a lifetime purchase on Founderstack.pro.