Periagoge
Concept
2 min readself knowledge

Hallucination: When AI Makes Up Facts and How to Catch It

Hallucinations happen when AI generates plausible-sounding but completely false information—it fills gaps in its knowledge with confident nonsense. Catching hallucinations requires spot-checking facts and citations, especially anything that sounds oddly specific or cited from sources you can't verify.

Hypatia
Why It Matters

Think of hallucination like when your friend confidently tells you a "fact" they're actually just guessing about. They sound certain, but they made it up. AI does this too—it generates text that looks real but is completely invented. For freelancers doing research, this is dangerous because you might use fake data in a proposal and look incompetent.

Here's why it happens: AI predicts the next word based on patterns. If you ask "What did Nike's CEO say about sustainability in 2022?" and AI has seen thousands of CEO quotes about sustainability, it might generate a quote that sounds real but never happened. It didn't look it up—it predicted what a reasonable quote would be. That's hallucination.

The scary part? Hallucinations sound confident. The AI doesn't say "I'm not sure." It presents invented facts as certainties. A statistic that doesn't exist, a company that doesn't have a product, a quote that was never said—all delivered with confidence.

How to Prevent Hallucination in Your Freelance Work

First rule: Never use specific data (statistics, quotes, numbers) from AI without verifying it. Ask AI "What percentage of small businesses use [tool]?" and it gives you "47%"—verify that 47% exists before putting it in your proposal.

Second: Use research-focused AI tools (like Perplexity) when you need current facts. Perplexity searches the web in real-time before answering, so it can't hallucinate outdated data (though it can misinterpret sources).

Third: Ask AI to cite sources. If you ask "What's the average freelancer hourly rate?" ask "Please tell me where you got this data from." If it cites sources, you can verify. If it vaguely says "based on industry reports," it's likely hallucinating.

Fourth: Use AI for brainstorming and drafting, not for definitive facts. AI is great at "Here are 10 ways to position yourself as a UX expert." It's terrible at "Here's the exact budget for UX projects in 2024."

Try this: Ask Claude a specific factual question like "What was the exact revenue of Stripe in 2023?" Then verify the answer with Google. You'll quickly learn when AI hallucinations happen so you know when to trust it and when to fact-check.

Helpful guides
Hypatia
Daily Life & Decisions
Related Concepts
Peri
Questions about Hallucination: When AI Makes Up Facts and How to Catch It?

Peri can explain this concept, give practical examples, help you decide whether it applies to your situation, or recommend a journey if appropriate.

Ready to work on Hallucination: When AI Makes Up Facts and How to Catch It?

Explore related journeys or tell Peri what you're working through.