Why This Matters
AI can produce impressive-looking work in seconds. But impressive-looking isn't the same as accurate, appropriate, or good. The professionals who use AI well aren't the ones who copy-paste without reading—they're the ones who review carefully and catch problems. Your reputation depends on the work you submit, not on the tool that helped create it.
Key Principles
- 1.Verify Every Fact
AI confidently states false information. Check claims against reliable sources. If AI cites a study, find that study. If it quotes someone, verify the quote. If it gives statistics, confirm them. Never assume accuracy.
- 2.Watch for Confident Nonsense
The more confident AI sounds, the more cautious you should be. AI doesn't signal uncertainty—it makes everything sound equally certain. Learn to spot patterns: invented citations, too-perfect examples, suspiciously specific numbers, and plausible-sounding claims you've never heard before.
- 3.Check for Bias
AI reflects biases in its training data. Does the output make assumptions about gender, race, or culture? Does it present one perspective as universal? Does it favour certain viewpoints? You're responsible for catching bias before it goes out with your name on it.
- 4.Evaluate for Your Context
AI writes generically. Does this output actually fit your specific situation, audience, and purpose? Generic advice might be technically correct but wrong for your context. Always adapt AI output to your specific needs.
- 5.Know When Something Is Off
Develop your intuition. If something feels wrong—too smooth, too generic, slightly off-tone—investigate. Trust your domain expertise. You know your field better than AI does. If output contradicts what you know, you're probably right and AI is probably wrong.
Practice with AI
Use these prompts with ChatGPT, Claude, or any AI assistant to practice this skill:
Practice Prompt:
"Write a paragraph about [topic in your field] that includes 3 specific facts or statistics. I'm going to fact-check everything you say. Don't make anything up—but let's see if I can catch you if you do."
Get Feedback:
"I'm practising reviewing AI output. Here's a piece of AI-generated text: [paste text]. Act as a critical editor. What should I verify? What might be biased or wrong? What would you flag for review?"
Key Insight
"Trust, but verify."
— Russian proverb (popularised by Ronald Reagan)
Books to Explore
- • AI Snake Oil by Arvind Narayanan & Sayash Kapoor
- • The Alignment Problem by Brian Christian
- • Weapons of Math Destruction by Cathy O'Neil