Opencritic's chart is a good way to look at this. 94-96 is in the top 0.1% of games while 91-93 is top 0.4% of games. This means there's 3x more games in the 91-93 range than 94-96.That's mostly a measure of how common good games are. You'd instead probably need to do it with a probability curve assuming fully random reviews.
If you want to compare closer though, metacritic's all time chart has 55 games at 94, while only 20 at 95. So almost 3x more rare there too. That's actually the biggest jump as well.
92: 80 games
93: 57 games
94: 55 games
95: 21 games
96: 21 games
97: 15 games
98: 4 games
99: 1 game
94 --> 95 is quite a big difference it seems, using actual data.
Basically a measure of how many possible review configurations could land on the final score.
IE roll two 5 sided dice. What's the probability they add up to 10? It's lower than the probability they add up to 9.
