🧼 AI Hygiene
We gave everyone AI. Now we're drowning in suggestions.
AI tools like Copilot, ChatGPT, and Notion AI promised productivity boosts. But for many engineering teams, they’ve introduced a new problem: AI-generated clutter. From noisy pull request comments to walls of autogenerated documentation, the flood of helpful-but-low-signal content is real.
In this post, we look at three concrete examples where AI is creating more work — not less — and explore what teams can do about it.
1. 📄 Autogenerated Docs That Say Nothing
The scenario: You enable Notion AI or GitHub Copilot Docs and watch the documentation magically appear. Except… it mostly repeats the function names, adds generic phrases, and lacks context or rationale.
Why it’s harmful:
- False sense of completeness
- Bloats repos and wikis with irrelevant noise
- Engineers stop trusting the docs
What to do:
- Treat AI-generated docs as first drafts, not source of truth
- Use manual prompts: “Explain why this component exists” instead of “Describe this code”
2. 🧪 Test Files That Don’t Actually Test
The scenario: You use a GenAI test generator. It creates tests with green assertions — but they only verify that functions return something, not the right thing.
Why it’s harmful:
- Inflated test coverage without logic verification
- PRs filled with passing but shallow tests
What to do:
- Use test generators only to scaffold basic structures
- Always inspect assertions — are they checking logic, edge cases, or just existence?
3. 💬 Bot Spam in Code Reviews
The scenario: Lint bots, security bots, GenAI review bots. Every PR now has 10+ bot comments — some helpful, most redundant. Reviewers start skipping them.
Why it’s harmful:
- Review fatigue
- Missed real warnings in the noise
What to do:
- Gate bot comments by importance or file scope
- Let senior engineers customize bot rules
- Ask: Would a human say this?
🚧 Embracing AI Hygiene
AI tools are powerful, but without hygiene, they generate clutter:
- ✅ Curate what you enable
- ✅ Review AI outputs like you would intern drafts
- ✅ Train your team to critically assess AI-generated artifacts
AI should accelerate clarity, not dilute it.
Let’s not trade human confusion for machine verbosity. Let’s practice AI Hygiene.
Tags: genai, software engineering, productivity, ai tooling, code review