1.g3 is by some databases one of the most successful opening moves, yet it is also much rarer than 1.e4, 1.d4, 1.c4 and 1.Nf3. Why?
Related: Why do masters open 1.d4 more often than 1.c4? and Why is 1. Nf3 not nearly as popular as 1.e4 and 1.d4? which ask about 1.c4 and 1.Nf3.
1.g3 and 1.Nf3 are very similar moves and are extremely likely to transpose into each other. So the main answer to this is that people prefer the move order starting with 1.Nf3. One reason to prefer Nf3 is that it takes away the option of 1. ... e5, the other reason is that it's more flexible because you can choose whether to fianchetto the bishop or to instead develop it in a different way. Basically 1.Nf3 leaves you with more options and your opponent with fewer options so it's hard to come up with a compelling reason for 1.g3, but it really doesn't make a big difference which is why the stats are fine either way.
The answer to your question may come from game statistics or knowledge of openings.
With regard to statistics, the "universe" of games can generate different responses if they are Grandmasters, FIDE, international, national or amateur tournaments. The former have more in mind that the order of plays allows them to reach the position they want to achieve, while the latter generally want a game in familiar territory.
With a base of 2.19 million games reported by "The Week in Chess", which I imagine is representative of players in major tournaments, 0.6% of the games started with g3 and scored 56.4% (42.9% - 27% - 29.9%, wdl). While all the games obtained 54.0% score (40% - 28% - 32%, wdl). The difference of 2.4 percentage points in score equals a difference of 0.17 pawns (Pawns = 4 * log (Winning percentage / (1 − Winning percentage))).
With my very limited knowledge of openings, I can tell you that g3 and Nf3 are part of the sequence of plays of various openings. Considering only two that comes to mind, when starting with Nf3, ..e5 is avoided while g3 does not. Thus with 1 g3 e5 2 Bg2 d5, the Benko opening is reached with a 49.7% score, while 1 Nf3 d5 2 g3 Nf6, leads to the King's Indian Attack of the Reti opening, called the Barcza System, with a score is 58.8%. Between both approaches, there is already a difference of 9.1 percentage points, which is still too early to indicate a favorite to win, since the errors in the middle game and even more during the end game, determines who will win. Let's just remember the aphorism: The player who makes the last mistake loses.
I asked a GM and apparently the reason for top GMs not playing 1. g3 is because it's not as good as the main moves at pressing for an advantage. This is confirmed by the cloud analysis on Lichess:
Of course, 1. g3 is playable and if you believe chess is a draw (virtually everyone does) then it's just as good as the main moves. But you are still less likely to reach a position where you can pressure the opponent.
