As others of said , the quality would be low. Partly because humans are terrible at understanding randomness and probability. Here's a simple though experiment.
If you flip a fair coin 10 times in a row once and only once and get 10 heads, that coin probably isn't fair and random, but instead biased.
If however you flip the coin 1000 times in a row, laws of probability say there will be like a 99% chance that a string of 10 heads in a row occurs somewhere. A memoryless event (outcome independent of the previous outcome) with probability 1/2 demand it.
Now if you DONT find 10 heads in a row (or 10 tails in a row) , that is now and indication that the coins are now biased and lost their memoryless independence property.
Now ask our human to create a string of 1000 random coin flips like tossing a fair coin would do. I would bet you a large sum of money that the human would go out his way to avoid calling 10 heads (or tails) in a row because that goes against the common layman understanding of randomness. In fact a lot of runs (e.g. 9 tails in a row) would have lower distribution than expected mathematically. But we just established this not what a uniform unbiased memoryless coin would do.
Humans have memory, especially unconscious memory and beliefs and bias, so it is very hard for them to mimic a memory unbiased random entity.
This reminds me when I was a smartass youngster and there was a huge lottery prize and my parents planned on buying multiple tickets and asked me to pick a random selection of 7 numbers from 40 numbers. I chose 1,2,3,4,5,6,7. My parents angrily complained that wasn't "random" and could never connive them that my selection had an equal chance of occurring as a 'random' selection of 3,15,9,24,29,41,33.