AI chatbots can produce an answer that sounds authoritative and is simply wrong. These errors are often called hallucinations. Knowing why they happen helps you decide when to trust a response.
The model aims for plausible text
A language model is trained to continue text in a likely way. It is not looking facts up in a verified database unless it is connected to one. When the right answer is rare, unclear, or absent from its training, it can still generate something that fits the pattern of an answer.
Situations that raise the risk
- Questions about obscure people, recent events, or niche details.
- Requests for exact quotations, citations, or statistics.
- Long chains of reasoning where one early mistake carries forward.
- Prompts that assume something false, which the model may accept and build on.
Habits that reduce the problem
- Ask the model to say when it is unsure, and treat vague confidence with caution.
- Provide the source material and ask it to answer only from that text.
- Verify names, numbers, and quotations against a trusted source.
- Use tools that cite sources, then check the sources themselves.
A practical rule
Use AI freely for drafting, brainstorming, and explaining. When the answer will be acted on, such as a medical, legal, or financial decision, treat the output as a starting point for checking, not as the final word.
Continue with the book
Explore the full treatment, worked examples, and exercises in AI Made Simple.
Adapted from AI Made Simple, published by Choral Software. Article text and original examples © 2026 Choral Software. All rights reserved.