Skip to content

AI Usage and Transparency

This page explains how artificial intelligence is used throughout this site, my philosophy behind it, and important disclaimers you should be aware of.

AI-Generated Content Disclosure

This entire site is heavily generated using AI tools. All grape variety articles are researched, written, and translated using artificial intelligence. While I work hard to ensure accuracy through citations and web research, please verify information independently before making growing or business decisions.

Personal Project Quality Notice

This is a personal passion project with minimal time investment. Content quality may vary, and articles may contain errors or incomplete information. Think of this as a starting point for your own research, not definitive agricultural advice.

My AI Philosophy

In my day job, I use AI to help employees navigate large internal knowledge bases that are unknown to AI models. This experience taught me that AI models don't perform well at retaining information about hybrid grapes because the topic is too niche. When you ask AI about an obscure variety, it will certainly hallucinate.

Instead of relying on AI "memory," I use these powerful tools to venture into the wild corners of the internet to find interesting facts, testimonies, and real experiences from growers and winemakers. The goal is discovery and connection, not invention.

How It Works: The Three-Layer AI Process

1. 🛠️ Repository Development

I rely heavily on AI to help code this entire repository: - Site structure and bilingual navigation - Content generation scripts - Translation workflows - Deployment automation

Tools used: Claude Code, GitHub Copilot My role: Architecture decisions, testing, refinement

2. 🔍 Content Research and Writing

For each grape variety article: - AI searches the web for real information about the variety - Focuses on finding testimonies, studies, and grower experiences - Assembles findings into a coherent, magazine-style article - Preserves all citations and references for verification

Tools used: OpenAI GPT-5 with web search capabilities My role: Variety selection, brief review, publication decisions

3. 🌍 Translation

My original vision was to create content in true Montréal fashion—naturally weaving French and English together based on source materials and context. However, this proved too complex for the current workflow. I may revisit this bilingual approach in the future. For now, all content originates in AI-generated English.

English articles are translated to French: - Uses smart hash-based system to only translate changed content - Adapted for Quebec and eastern Canadian wine growers - Preserves technical terms and citations exactly - Maintains the accessible, friendly tone

Tools used: OpenAI GPT-5 with Quebec French context My role: Not much for now

What This Means for You

✅ Strengths: - Access to obscure information from across the web - Consistent format and accessibility - Citations provided for further exploration - Bilingual availability

⚠️ Limitations: - AI may miss nuances only human experts would catch - Information is only as good as what's available online - Personal project quality - not professionally edited - May contain errors or outdated information

🎯 Best Practice: Use this site as a discovery tool to find interesting varieties and sources, then follow the citations to connect with the actual growers, researchers, and publications mentioned.

Technical Details

For developers and those curious about the implementation:

  • Source code: GitHub Repository
  • AI Models: OpenAI GPT-5 with web search tools
  • Languages: Python, MkDocs with Material theme
  • Deployment: GitHub Pages with custom domain

The entire workflow is designed to be transparent and reproducible. You can see exactly how each article is generated by examining the code and prompts in the repository, including the main system prompt that guides the variety content generation.

This transparency page reflects my commitment to honest AI usage. If you have questions or concerns about any content, please reach out or check the GitHub repository for more details.