GenAI & LLMs · 1 min read · Updated 2026-10-05
AI hallucinations
Why it happens
The AI guesses likely words. When it does not know a fact, it can still write a smooth sentence. It might invent a book title, a quote or a number.
How to reduce it
Give it the facts to use. Ask for sources and check them. Tell it it is fine to say 'I don't know'. Use a calculator for sums. Newer models make fewer mistakes, but none are perfect.
When to be extra careful
Double-check anything about health, money, law or safety. Also check exact details like dates, prices and version numbers.
Key takeaways
- AI writes what sounds right, which is not always true.
- Giving it sources and checking its claims helps.
- Always verify important details.
Quick questions
Can it be fixed completely?
Not yet. It can be reduced but not removed.
Why does it invent sources?
It knows what a source looks like and can make one up that seems real.
How can I spot one?
Look up the specific claim. Made-up details are often very exact but cannot be found.
Keep learning
Retrieval-augmented generation (RAG)RAG lets an AI look things up before answering. The system finds the right pages in your documents and hands them to the…1 min readRead →What is a large language model (LLM)?An LLM is a computer program trained on a huge amount of writing. Its main skill is guessing the next word, again and…1 min readRead →