Metadata Engineering & Knowledge Management
Implement a "Source of Truth" architecture using metadata and specialized folders. Build a centralized "Memory Document" to proactively train your personal AI companion.
Key Takeaways
- Using Drive Description fields as semantic hints
- Filename optimization for retrieval probability
- The @ mention system for targeted retrieval
- Building your Biz Gem Memory Document
The Metadata Wrapper
Every file in Google Drive has a hidden superpower: the Description field. This field provides high-level semantic hints to the vector index, dramatically improving retrieval accuracy.
Optimizing the Description Field:
- •Add a 1-2 sentence summary of file contents
- •Include key entities and topics covered
- •Add temporal context (e.g., "Q1 2026 planning document")
- •Note relationships to other documents
- •Include relevant keywords for your use cases
Filename Optimization
Filenames are the first signal AI systems use for relevance scoring. "IMG_2026.jpg" has near-zero retrieval probability. "Vendor_Invoice_Q1_2026.pdf" is immediately categorizable.
Filename Best Practices:
- •Use descriptive names: Project_Budget_Analysis_Q1_2026.xlsx
- •Include date context when relevant: Meeting_Notes_2026-01-15.md
- •Add category prefixes: INVOICE_, RECEIPT_, CONTRACT_
- •Use underscores or hyphens, not spaces
- •Avoid generic names: Document1, Untitled, Final_FINAL_v2
Filename Obscurity Cost: A file named "IMG_2026.jpg" has 10x lower retrieval probability than "Vendor_Invoice_Q1_2026.pdf" for a query about expenses.
Targeted Retrieval with @ Mentions
The @ mention system in Gemini allows you to restrict search scope. Instead of searching all your data, you can point to specific folders or files.
@ Mention Strategies:
- •@[FolderName]: Restrict search to a specific folder
- •@[FileName]: Reference a specific document for context
- •Combine with queries: "Based on @ProjectAlpha folder, what's our timeline?"
- •Create "context folders" for AI conversations
- •Use for sensitive topics to avoid unrelated data exposure
The Biz Gem Strategy: Memory Documents
A Memory Document is a centralized file containing your brand voice, core philosophy, operating principles, and key context. It serves as a permanent anchor for all your AI conversations.
What to Include in Your Memory Document:
- •Brand voice and tone guidelines
- •Core business philosophy and values
- •Key terminology and definitions
- •Project naming conventions
- •Preferred communication style
- •Frequently referenced facts (company size, key dates, etc.)
- •Template structures for common outputs
Pro Tip: Start conversations with "Using @MemoryDocument as context, help me..." This primes the AI with your preferences before every interaction.
The Inbox Indexing Protocol: JSON-LD for Orders
Standard JPG invoices are invisible to aggregation queries because they lack semantic structure. The solution is JSON-LD Order markup in email confirmations.
{
"@context": "http://schema.org",
"@type": "Order",
"merchant": {
"@type": "Organization",
"name": "[Brand Name]"
},
"orderNumber": "123456789",
"orderDate": "2025-12-15T10:30:00-08:00",
"priceCurrency": "USD",
"price": "150.00",
"acceptedOffer": {
"@type": "Offer",
"itemOffered": {
"@type": "Product",
"name": "Product Name"
}
}
}When merchants include this markup in order confirmation emails, Gemini can accurately aggregate your spending across all purchases.
Practitioner assets
Turn this lesson into a repeatable GEO workflow
Use the checklist, sources, templates, and assessment prompts to move from theory to a client-ready diagnostic or implementation step.
- highDefine the prompt set, user intent, market, persona or vertical scenario for this lesson.
- highCapture current AI answer evidence with provider, date, excerpt, citations and competitor mentions.
- highIdentify the likely root cause: content gap, authority gap, technical access, source inconsistency, review signal or policy risk.
- mediumCreate the visible page, proof block, profile update, policy clarification or report artifact that resolves the gap.
- mediumAssign owner, due date, expected impact and remeasurement window before calling the work complete.
- Google Search Central: Creating helpful, reliable, people-first contentGoogle Search Central · 2025
- Google Search Central: Intro to structured dataGoogle Search Central · 2025
- Google Workspace Admin Help: Control data sharing with GeminiGoogle Workspace Admin Help · 2025
- Metadata Engineering for Private AI Work Product TemplateA repeatable worksheet for applying Metadata Engineering for Private AI to a real brand or client account.
- Before/After Answer ProofA reporting format for showing how AI answer quality changed after the improvement shipped.
This lesson includes 5 assessment questions to reinforce the concepts before you apply them to a real GEO audit.