Chapter 3: What AI Gets Wrong (And How to Catch It) — Rachel Holt

In this chapter

What Hallucination Actually Means
The Patterns: When AI Is Most Likely to Be Wrong
A Simple Verification Habit That Takes 30 Seconds
Tools That Reduce the Risk

Tools mentioned in this chapter

Perplexity

Cites sources for every claim — best for verification

Claude

Tends to say when it's uncertain rather than guessing

ChatGPT

AI assistant by OpenAI

Google Gemini

Grounded in Google Search — good for current facts

Microsoft Copilot

Cites web sources inline

Prompts to use

Click any prompt to expand it, then copy and paste directly into your AI tool. Replace the bracketed placeholders with your specifics.

You just told me [paste the AI's claim]. Is this accurate? Show me sources. If you're not confident, tell me that.

I'm going to use this information for [purpose]: [paste AI-generated text]. Flag anything that should be verified before I rely on it. Mark your confidence level on each claim.

What did you just tell me that you're least confident about? What could you be wrong about in your previous response? Be specific.

Give me three strong arguments against this conclusion: [paste AI's conclusion]. I want to pressure-test it before I use it.

For every factual claim in your next response — any number, name, date, statistic, or legal reference — include a source or clearly mark it as "unverified."

What kinds of questions or tasks are you most likely to get wrong? Give me a practical guide to when I should double-check your output.

Can't find the right prompt?

Ask your AI tool to write one for you:

"I need to [describe your goal in one sentence]. Write me a clear, specific prompt I can use to get the best possible result from an AI assistant."

View full prompt library →

Video tutorials

Curated videos to go deeper on the topics in this chapter.

Get the book

Productivity for Women

By Rachel Holt — Available in Kindle and paperback

Kindle — $9.99 Paperback — $17.99

← Chapter 2 Chapter 4 →