SEO consultant turns personal knowledge into paid AI audits

From manual audits to AI pipelines

SEO consultant Marie Haynes recently shared a detailed breakdown of how she transformed her consulting business using Google's Open Knowledge Format (OKF). For years, Haynes manually diagnosed why websites lost traffic, a process that often took weeks of intensive work. She now uses a system that drafts these reports significantly faster and at a lower cost than her previous manual methods.

The core of her new workflow is a 'brain' consisting of roughly 600 markdown files. These files contain her entire body of work, including blog posts, specific playbooks, and a news archive she began compiling in 2012. By organizing her expertise into these small, labeled text files, she allows an AI to 'look up' her specific knowledge rather than relying on general training data.

As PPC Land reported, Haynes is using these files in conjunction with Google's Gemini 4 Argon model. This setup allows her to generate traffic drop reports and helpfulness assessments for clients. She notes that the AI's output is often as good as her own manual work, allowing her to serve smaller clients who previously could not afford her high-figure consulting fees.

How the OKF brain functions

The system operates on a principle Haynes calls 'more specified' retrieval-augmented generation (RAG). In standard RAG, a model pulls from a large index, which can sometimes lead to hallucinations or incorrect information. By using OKF, Haynes breaks her knowledge into distinct files that an agent can choose between based on specific labels and tags.

The structure of the knowledge base was actually designed by an AI agent using Google's Antigravity platform. It includes a master index that acts like a library card catalog, helping the model see what documents are available before it opens them. This prevents the model from getting lost in irrelevant data and ensures it pulls the correct playbook or entity for a specific client's needs.

To maintain accuracy, Haynes performs a periodic cleanup process she calls 'lint.' When she notices the AI making a mistake or using the wrong terminology, she corrects it and uses a specific command to record that change. This feedback loop ensures that the AI learns her specific preferences and nuances over time, refining the quality of every subsequent report.

The economics of AI consulting

Haynes emphasizes that the goal of this system is to sell access to knowledge rather than the knowledge itself. She argues that selling raw files hands over intellectual property, whereas an AI-powered interface allows clients to interact with her expertise while she retains control of the underlying data. She currently offers three versions of this access, including a chat interface and an automated report pipeline.

The cost of running the system is remarkably low for her current scale. She reports spending about $5 or $6 a day on the Gemini 3.8 Flash model for her chat interface, with individual queries costing only pennies. Because she stores her files in Google Cloud storage buckets, she currently pays $0 for storage, making the business highly profitable compared to her old manual model.

While she mentions that this specific setup might not scale for massive enterprise needs, it works for her current client base. She believes that as the technology matures, she may eventually be able to automate the reports entirely. Currently, she still manually checks the information to ensure it is correct and not hallucinating before it reaches the client.

Future of knowledge as a skill

This project highlights a shift in how professional expertise can be monetized in the age of AI. Instead of selling hours of human labor, experts can package their life's work into standardized formats like OKF. This allows them to create tools that can serve hundreds of customers simultaneously while maintaining a high level of personalization.

Haynes points to new Google features as a potential way to expand this model further. She suggests that a panel for connecting external servers could allow consultants to sell their knowledge as a skill inside the Gemini Business environment. This would let users interact with an expert's 'brain' without the expert having to build a custom app or hand over their private files.

For the broader marketing and consulting industry, this represents a move toward 'knowledge as a service.' It suggests that the most valuable asset for a consultant is no longer just the ability to do the work, but the ability to structure that work into a format that an AI can navigate accurately. This transition allows for a more scalable way to provide high-quality, specialized advice to a wider audience.