AI Tools: I knew AI + Notion would be good, but this… wow.

Watch on YouTube: https://www.youtube.com/watch?v=ofzHbyxkpVg

Channel: Jeff Su

Teaching AI Your Notion Conventions: A Rulebook Approach That Works Across Platforms

Most professionals who use Notion have developed personal systems — naming conventions, icon choices, database structures, property tags, and page layouts that reflect how they think and work. The challenge arises when bringing AI into the mix: connectors for Claude, ChatGPT, and Gemini understand Notion’s API, but they don’t know your conventions. Jeff Su’s demonstration shows how a single markdown rulebook solves this, letting any compatible AI operate your workspace exactly as you would.

What the Video Covers

The core problem is straightforward: external AI tools can read and write to Notion via API, but they lack the context of your personal workflow preferences. The solution is a Notion Conventions file — a markdown document that captures your rules for naming, icons, database selection, templates, properties, and page structure. Once this file is mounted in an AI project, the assistant reads it before every Notion action and applies your standards automatically.

Key takeaways from the demonstration:

• AI connectors know Notion’s API but not your personal conventions; a written rulebook bridges this gap.
• The same conventions file works identically across Claude, ChatGPT, and Gemini as long as the folder is mounted.
• The method scales to dozens of databases using a prompt template that has the AI analyze properties, templates, and recent entries, then ask clarifying questions.
• Complex multi-database workflows — such as creating a live Q&A workshop project with a project page, North Star document, deliverables database, tasks database, and notes — can be executed in a single command.
• The broader concept, called Jeff OS, positions a central AI context folder (holding conventions for Notion, Gmail, Calendar, and other tools) as the orchestrator, with Notion AI reserved for in-app tasks like copy review and bulk migrations.

Step-by-Step Walkthrough

The implementation follows a clear sequence that can be replicated in under an hour:

1. Download the starter template. Jeff provides a Google Doc formatted as a Notion Conventions starter. Download it as a .md file.
2. Create a local folder. On your desktop (or preferred location), create a folder named Notion Ops and place the conventions file inside.
3. Connect Notion to your AI. In your chosen platform (ChatGPT, Claude, or Gemini), enable the Notion integration/connector.
4. Create an AI project and mount the folder. Start a new project called “Notion Ops” and mount the Notion Ops folder so the AI can read and write the conventions file.
5. Teach task and note preferences. Copy the link of an ideal task page and an ideal note page from your Notion workspace. Use the provided prompt template (linked in the video description) to have the AI analyze those pages and save your structural preferences to the conventions file.
6. Verify with a fresh chat. Start a new conversation with the folder mounted and ask the AI to create a test task or note. Confirm the output matches your conventions.
7. Scale to each database. For every database you want the AI to manage, use the database analysis prompt template. The AI will inspect properties, templates, and recent entries, then ask clarifying questions before writing the rules to the conventions file.
8. Execute complex workflows. Once conventions are captured for the relevant databases, you can issue a single high-level prompt — e.g., “Create a live Q&A workshop project with North Star, deliverables, tasks, and notes” — and the AI will create and link all pages per your rules.

Tools, Numbers, and Scale

• Platforms: Notion, Claude, ChatGPT, Gemini, Notion AI
• Starter asset: Google Doc template for Notion Conventions (download as .md)
• Prompt templates: Provided for teaching task/note preferences and for database analysis (linked in video description)
• Workspace scale: Demonstrated across 20+ databases in Jeff’s personal Notion workspace
• Jeff OS: A folder system for multi-tool AI context (Notion, Gmail, Calendar, etc.), taught in the Co-work Academy course
• Claim: “90% built second brain” upon downloading the Jeff OS system (course promotion)

Caveats and Hype Check

The video promotes Jeff’s paid Notion course and the Co-work Academy (which teaches the Jeff OS system). The “90% built the moment you download it” claim is a marketing hook for the course. However, the core methodology — creating a conventions markdown file, mounting it in an AI project, and using prompt templates to teach preferences — is demonstrated live and is fully reproducible for free using the provided templates. Notion AI remains useful for in-app tasks such as copy review and bulk migrations, but the external AI becomes the cross-tool orchestrator. No vendor lock-in: the same conventions file works across all three major AI platforms.

Any AI with a connector to Notion already knows how to use Notion, but that’s not enough. It also needs a written rulebook with your preferences.

We spend a little bit of time up front teaching AI our preferences, and it saves them in a rule book for that tool. So that the next time we ask it to use that tool, the AI can read the rule book and do the work the way we like.

Concrete Next Actions

• Download the Notion Conventions starter Google Doc as markdown (.md).
• Create a Notion Ops folder on your desktop and place the conventions file inside.
• Connect Notion to your preferred AI (ChatGPT, Claude, or Gemini).
• Create a new project in the AI named “Notion Ops” and mount the folder.
• Copy the link of an ideal task page and an ideal note page in Notion; use the provided prompt to teach the AI your structure.
• Verify by starting a fresh chat with the folder mounted and requesting a test task or note.
• Repeat the database analysis prompt for each database you want the AI to manage.
• If you manage multiple tools beyond Notion, explore the Jeff OS folder system for a unified AI context.

The upfront investment is modest — roughly 30 to 60 minutes to set up the folder, mount it, and teach the AI your core databases. After that, every Notion interaction via AI follows your conventions without further prompting. For teams or individuals running complex, multi-database workflows, this rulebook approach turns an otherwise generic AI connector into a reliable operator that works the way you do.