Ok - let's say I tell you a number between 1 and 100 inclusively, which is random - how could you prove it?
Of course, if I say 1, 100, 99, 42 or 13, most people will claim it wasn't random, because their first thoughts where about the same numbers, but if these numbers where excluded, then it wouldn't be a fair draw from 100 numbers, but from 95.
So from a single number, you can't distinguish a biased number from a random number. I can tell you a number, and you can't prove it isn't random, but you can't prove it is as well.
So I could tell you my technique, how to generate the number, and convince you. If most people could learn the technique, and you coulnd't tell before, what number the people are producing, and more: if the repeated usage of the technique would lead to an uniformly distribution over the numbers from 1 to 100, we would call this a 'Yes, we can'.
Ok. So I take a sheet of paper, write the numbers 1 to 100 on small pieces of it, put them in a bag, and pull a number - mission accomplished! :)
That is cheating!
Is it? Ok - we then need to see how to sharpen the rules, I guess? No tools allowed, like paper and bag, RNG and dices?
Ok - I pull a lot of hairs out of my head - much more than 100, let's say 500 to 1000 hairs, and then I count them, and take the modulo. :)
That is cheating!
Is it really? Well, of course the technique doesn't scale. It can't be repeated very often; not often enough to check, whether an equal distribution is reached. And you could tell, that the rules prohibit any material as tool.
Ok. Then I'm nearly to the end of my wisdom. For a random result, I need some unpredictable input, and of course, my brain isn't a good source for such an input. I could name a measurement, which would work for a small amount of small random numbers (1-10), and it works similar to the hair example:
I think about a song I know, and the first song which comes to my mind, I count the characters. I can't predict from the first song - "I am the walrus" how many characters it has. The number is pretty big, compared to the number range (1-10), and I take the modulo, and have a random result.
So this is a proof of concept, and the next time, I would, of course, have to choose a different song, and so the number of experiments is limited to the number of songs or poems I know, and it is a time consuming procedure.
If the rules say 'name a number spontaneusly - in 2 seconds', I'm pretty sure nobody can produce repeatedly random numbers. But given enough time, you can.