How it 'thinks' (and why it sometimes makes things up)
You don't need to understand the technology to use it well — but two ideas will save you a lot of confusion.
1. It predicts likely words, it doesn't 'look things up'
Under the hood, the AI is predicting what text should come next, based on patterns it learned from enormous amounts of writing. Most of the time that produces genuinely useful, correct answers. But because it's generating rather than retrieving from a database, it can occasionally produce something that sounds perfectly reasonable and is simply not true. This is often called a hallucination.
For anything that matters — a medical detail, a legal point, a statistic, a quote, a person's name — treat the AI like a smart friend giving you a tip, not like an official source. Verify before you rely on it.
The word 'hallucination' for AI confidently inventing things only became common around 2022–2023. It stuck because it captures the feeling exactly: the answer looks real, but there's nothing behind it.
2. Each new chat starts with a blank memory
Within a single conversation, the AI remembers everything you've said so far — so you can say 'make that shorter' and it knows what 'that' means. But start a brand-new chat and it's a clean slate again. That's not a bug; it's how you keep different tasks separate. If something matters, include it in the conversation you're in.