We’ve all seen them: AI-generated pieces of content. Are they relevant? Do they deliver real business results? Do they bring in web traffic?
The short answer: yes, you can slash your SEO page production costs by a factor of 100—if you forget about ChatGPT and invest in a more powerful system.
Try to accomplish a complex task with a single AI assistant (like ChatGPT or Claude), and you’ll notice the result lacks depth. Your assistant responds in a way that’s too generic and superficial. Both your audience and Google’s algorithms pick up on it. The content ends up adding little or no value—or even actively harms your brand.
You wouldn’t expect a CEO to handle every single task in a company without any help from staff. The same is true for AI: if you want the best possible outcome, you need to build a team that works in a coordinated way toward a complex goal.
This is where workflows, AI agents, or multi-agent systems (MAS) become essential. These are more than just chatbots like ChatGPT; they allow you to build a chain of reasoning in which each link is specialized in a distinct sub-task.
Platforms like RelevanceAI or aimw make it possible to roll out these solutions. That’s exactly what I did for a client who wanted to automate their SEO content production.
Why use such a system for SEO content creation?
Creating SEO content is inherently complex. When a copywriter crafts a webpage or blog post, they have to complete several stages to produce optimized text that can compete in search rankings:
- SERP analysis
- Examining top-ranking pages
- Researching search intent
- Staying updated with the latest news
- Identifying secondary keywords
- Building an outline
- Writing and reviewing the text
- And so on…
Asking ChatGPT to generate your content—even with a solid prompt—will rarely deliver the depth you need, and Google won’t rank it highly. A chatbot on its own can’t effectively handle all of these tasks.
However, if you assign each part of the process to specialized agents or tools, your AI system can create content that matches (or even surpasses) what the best copywriters can do while cutting your production budget by a factor of 100.
The Components of My Multi-Agent SEO System
On RelevanceAI, as with most multi-agent systems, the architecture hinges on two key elements: agents and tools.
My SEO Agents
Agents are autonomous entities with their memory, each capable of completing specific tasks according to predefined guidelines. My system revolves around a Manager Agent that oversees the workflow, delegating tasks to more specialized agents:
- Research Agent: Gathers in-depth information on the topic.
- SEO Brief Agent: Synthesizes the research findings into a concise, SEO-focused brief for the writer.
- Writing Agent: Writes the content following the brief, using SEO-ready copywriting techniques.
- Email/Google Docs Agent: Handles final content delivery, whether via email or integration with Google Docs.
Each agent follows structured instructions:
- Role: What the agent does and its area of expertise.
- Objective: Clearly state the outcome we want.
- Context: Explains why the agent’s task matters in the overall process.
- SOP (Standard Operating Procedure): A step-by-step guide ensuring quality and consistency.
- Tools: Lists the resources and APIs the agent needs to complete its job.
- Output: Defines the expected final format and information.
- Notes: Flags any critical details and special instructions that must be followed.
My tools
The tools employed by my agents are specialized modules, which can include:
- Google search
- Competitor page analysis
- News and trend research
- SEO brief generation
- SEO outline creation
- Content writing
- Content review
- And more…
Each tool is built to take a precise input, perform one or more steps through dynamic variables (allowing information to be passed from one step to another), and send back an output that the agent can use to fulfill its objectives.
Optimizing and Managing AI Models
A major challenge in this system is integrating various AI models, each with its own strengths and limitations.
Affordable, Fast Models
Some models, like ChatGPT 4o, Claude 3.5 Sonnet, and others, aren’t designed for complex reasoning. They’re fine for simpler tasks but fall short when given detailed instructions or asked to produce richly nuanced content, especially for advanced SEO.
Models for Complex Tasks
For more demanding tasks, we turn to models like o1, DeepSeek R1, or Gemini Flash 2.0 Experimental Thinking. The o1 model processes large data volumes without blending context and instructions, though it’s more expensive to run. DeepSeek R1 and Gemini Flash 2.0 call for specific prompting tactics due to their different context windows and how they handle extended inputs.
What Does It Cost to Run My System?
A single complete run can cost anywhere from a few cents to several euros/francs, depending on the model you use. After switching its system to DeepSeek R1 (an open-source model), my client now pays around 30 to 50 cents per full run.
Quality Assurance and Validation
To ensure each agent’s work meets the required standards, I carry out a manual check at every stage. Whenever an agent or tool completes its task, I review the output to verify that it aligns with the overall workflow’s goals. This step-by-step verification helps catch any issues early so we can refine the instructions before moving on.
AI Workflow vs. MAS: Control vs. Flexibility
A multi-agent system can be incredibly flexible and adaptive—crucial qualities for tasks like SEO content creation. However, this same adaptability can also introduce unpredictability.
By contrast, a traditional workflow is more rigid and offers tighter oversight through fixed steps and linear execution. Depending on the project and the necessary level of detail, it can be wise to choose a flexible, dynamic approach that combines the freedom of a MAS with the discipline of a controlled process.
Conclusion
Using an AI-driven approach for SEO content creation is poised to become standard procedure, effectively replacing more manual methods. By segmenting the process into specialized tasks and leveraging targeted tools, you can produce top-quality content at a mere fraction—1%—of the usual cost. SEO departments will need to adapt or risk being left behind. If you’d like to automate your content production, feel free to reach out: clement@clementschneider.ai.
