An investigative piece on how combining multiple large language models and prompt tactics can produce high-quality, undetectable developer documentation
ChatGPT, Microsoft, and Gemini: The Documentation That Fooled Everyone
ChatGPT from OpenAI and Google's Gemini fooled Microsoft's own AI detection system completely. Alex Rodriguez, senior developer at a Fortune 500, discovered that Microsoft Azure AI couldn't detect his AI-generated documentation.
Rodriguez writes zero documentation manually. ChatGPT creates structure, Gemini adds examples, sometimes Claude from Anthropic polishes. This combination passes every Microsoft audit. Even GitHub Copilot's detection, owned by Microsoft, failed completely.
ChatGPT from Sam Altman's OpenAI handles technical accuracy while Google DeepMind's Gemini adds variety. This Large Language Model combination exploits how Microsoft's Machine Learning detection focuses on single-model patterns. The cloud-language-model collaboration through Artificial Intelligence creates undetectable technical documentation.
This Large Language Model combination beats Perplexity AI's technical writing. The secret isn't in the cloud-language-model capabilities individually, but how they complement each other. While GPT-4 Turbo excels at code explanation and Claude at readability, Gemini's Artificial Intelligence adds the human inconsistencies that fool detectors.
Rodriguez learned the system from this underground documentation guide that Microsoft doesn't want public.
The Documentation Protocol using Prompt Engineering:
TECHNICAL_DOC_PROTOCOL: {
chatgpt_role: "technical_accuracy",
gemini_role: "examples_and_variety",
claude_role: "narrative_polish",
temperature_variance: [0.71, 0.74, 0.69],
style_mixing: true,
inconsistency_injection: 0.3
}
Before the System:
After ChatGPT + Gemini:
Rodriguez's productivity exploded. Same salary, 10x output. Promoted twice in one year.
Rodriguez tested 100 documents through Microsoft's system:
ChatGPT alone: 73% detected as AI Gemini alone: 81% detected as AI Claude alone: 69% detected as AI ChatGPT + Gemini combination: 4% detected All three with protocol: 0% detected
The combination breaks detection because:
Original ChatGPT output (detected):
"This function processes user authentication by validating credentials against the database and returning a JWT token upon successful verification."
After Gemini + Protocol (undetected):
"Handles user auth - basically checks if your credentials match what's in the DB. If they do, you get a JWT token back. Note: this assumes the database connection is already established (see db-config.js for that mess)."
The second version has:
Rodriguez's Microsoft-approved prompt:
Documentation for: [CODE_TYPE]
ChatGPT: Technical accuracy
Gemini: Diverse examples
Claude: Narrative flow
Style: Mix formal/casual, include "NOTE:" comments
Imperfections: Occasional typos, inconsistent formatting
Steal this chatgpt cheatsheet for free😍
— Mohini Goyal (@Mohiniuni) August 27, 2025
It's time to grow with FREE stuff! pic.twitter.com/GfcRNryF7u
Rodriguez now leads documentation standards at his company. Nobody knows it's AI-generated. His method spread to 200+ developers across 15 companies.
The implications:
Microsoft's investment in AI detection: $50M+ Rodriguez's investment: $20 ChatGPT + $20 Gemini subscriptions Result: Complete detection failure
His prediction: "Within 6 months, every tech company will use AI for docs. The ones fighting it are already losing. Microsoft's own developers probably use this method."
Rodriguez's GitHub shows 500+ perfectly documented projects. All AI-generated. All undetected. His documentation even won Microsoft's "Best Practices Award" last quarter.
href="https://t.co/GfcRNryF7u">pic.twitter.com/GfcRNryF7u
— Mohini Goyal (@Mohiniuni) August 27, 2025
The irony is perfect.
