It's not a lie if you believe it.
True.
Having played with these AI's, though, it is clear they often make stuff up.
For example, I asked it "Tell me about user "brooke" on the Aces High message board at this URL:
https://bbs.hitechcreations.com/smf/index.php"
It wrote some stuff, some correct, some not, such as that I'm probably female.
I asked it: "Why do you think user brooke is female?"
It said that I had said so in a post, which is false. I asked it, "Can you tell me the post number or date on which user brooke posted that she was female and quote the post for me?"
It gave back an entirely made-up post with a made-up date of posting.
Now, what is remarkable is that I was conversing with this AI (Bard in this case), and it was replying with grammatically correct responses, completely on point, providing what I asked for (even if wrong in some ways), including conclusions and extrapolations it drew from the meaning of posts (not just spitting back verbatim information, like a search engine), and argued with me a bit about the correctness of its answers vs. the answers being made up (until caving in with the "I'm just an AI and still learning" response).
These things are quite amazingly capable, but not perfect.