If I had used ChatGPT in the way you described, I would have likely verified that the numbers sourced from search results it found were used correctly when it was "doing some numbers," since LLMs don't really do numbers at all. At that point, why not just search and do the numbers myself?
Answers to your questions below:
> After reading TFA, did you not wonder, just a bit, what the box could be worth?
No, but admittedly I have a lot of experience selling rare collectibles of several different kinds myself. So I knew enough to know that the only remotely accurate way to estimate the value of something like this (rare, not previously known, possible collector base is very small but dedicated) is to put it up for auction. I recognize that not everyone may understand this, though.
If you browse auctions for collectibles, and dig up the rarest of the rare items, you'll notice that even if the auction house provides an "estimate," they are usually wildly off compared to the actual hammer price, despite coming from the presumable experts in that field. Sometimes the estimates are high, sometimes low; it just depends on the whims of the market.
> Would you have taken the time to respond to my post if I hadn't told the source and just said it's "probably" worth x, instead of saying "might, according to"?
Yeah, I quite possibly would have, and I likely would have responded with an answer similar to my answer to the previous question.