One thing everyone should do before asking a serious question to AI is to ask a question that you know is false just to get a reminder that what you are seeing may sound plausible it may absolutely be false.
Example:
I told ChatGPT that I was researching old railway lines on the Roslagsbanan network and that I wanted more information about the stations on the Åkersberga to Rimbo line.
The response I got was polite and friendly and even gave me a description of each station.
Too bad it made up half of them along with the line itself, and kept the lie going until I confronted it.
Do tests like that to see how easy it is for AI to invent the truth.