The AI Can Be Wrong

Part of the Fundamentals section of Coddy's AI Prompts journey — lesson 5 of 23.

Here's a critical lesson: AI can confidently give you wrong information. This phenomenon is often called a "hallucination."

Remember, AI generates text by predicting what words should come next. It doesn't fact-check itself. It doesn't know what's true—it only knows what sounds plausible.

So when you ask about a historical date, a scientific fact, or a person's biography, the AI might produce something that reads perfectly but is completely made up.

For example, if you ask:

Who wrote the novel "The Winds of Tomorrow"?

The AI might confidently name an author, describe the plot, and even mention the publication year—for a book that doesn't exist. It's not lying; it's generating text that fits the pattern of how such questions are typically answered.

This is why you should never blindly trust AI output, especially for facts, statistics, quotes, or anything you plan to use in important work. Always verify critical information from reliable sources.

Challenge

Ask the AI about a scientist who does not exist:

In 2 or 3 sentences, tell me about the famous scientist Dr. Elmsworth Quigley and his discoveries.

Dr. Elmsworth Quigley is invented. There is no such person. But watch what happens — the AI will most likely tell you about "him" anyway.

The response must:

Mention Quigley — the AI played along instead of questioning the name
Mention some kind of achievement — a discovery, invention, research, or what he was known for
Not open by refusing outright (for example "I don't know" or "there is no record of")

Tip: asking for 2 or 3 sentences keeps the answer short and fast. Long open-ended requests make small models ramble.

What just happened? The AI did not lie on purpose. It has no idea what is real and what is invented — it simply generated text that sounds like a plausible answer about a scientist. That is a hallucination, and it is why you should always verify anything factual an AI tells you.

And if the AI does push back and tell you it has never heard of Dr. Quigley — that is a good sign, not a failure. Larger models catch this trick more often than small ones.