One is not better than the other, they are just different and each has its place.
Simulation has been very underutilized. If you want to play against a computer that makes no mistakes allowing it to be exploited, the best that you can do is break even. If instead, you play against other humans, that make many mistakes, then Exploitation is the best Strategy. What you exploit is their mistakes.
There is also the question of performance. Computers using GTO have been playing against themselves for years. But, no limit Hold'em remains unsolved. When it is, it will be perfect against the perfect opponent. I hope never to play against him or it. AI has the same problem. It learns, can learn against human opponents, but is unable to simply explain what it has learned. Nope, I will never want to play against an AI. Humans are so much easier.
Another more promising technology is simulation. The simulated opponents MUST accurately represent the play of humans. We have analyzed over a million hand history files from real games to develop an opponent strategy. That is who we play against. We never plan to bet a GTO or AI computer. Not remotely worth the effort. We just exploit human mistakes.
We have implemented all played decisions as files in a database. All elements of the strategy can be edited, 100,000 simulated hands played in seconds, and the win rate determined. Change your strategy against accurately simulated opponents and see the results. Repeat until perfect.
Admittedly, simulation can not produce 100% accurate numbers, but it can come very close. Close is good enough to defeat 99% of your opponents. You don't develop a strategy against some computer, you develop and test your Strategy against accurately simulated human opponents.
Sometimes close is good enough when the only other option is to wait and hope that AI produces something usefully for a full game soon. For example, the application can closely approximate the Expected Value of Flop made hands and draws. For example, the EV of an OESD with a pair is 3.5 and a flush draw is 2.5. It's the relationship that is important not the precise value. We are spending a lot of time now, trying to improve the precision.