It's actually counter-intuitive. I was going to argue on your side, and then I wrote up a quick program that proved me wrong[0]. Let's go through the scenarios, with Goat A, Goat B, and Car C. In the scenario where Monty picks a door purposefully, always selecting a goat, the scenarios are:
You picked A, Monty showed you B, and you switch to get C.
You picked B, Monty showed you A, and you switch to get C.
You picked C, Monty showed you either A or B, and you switch to get the other goat.
So a 2 / 3 chance of getting the car if you always switch.
If Monty is choosing randomly, we have the following scenarios:
Initial Choice | Monty's choice | Remaining Door
A | B | C
A | C | B
B | A | C
B | C | A
C | A | B
C | B | A
But we know in the problem statement that Monty hall showed us a goat, so we can eliminate possibilities 2 and 4 to get:
Initial Choice | Monty's choice | Remaining Door
A | B | C
B | A | C
C | A | B
C | B | A
Whether you switch or not, you have a 50/50 chance.
I'm not great with probabilities, but the major difference I can see is that in the first scenario, if you pick the car, Monty will either show you the goat A or B with equal probability as a part of the same 1/3 scenario. So you have really:
1/3: You picked A, Monty showed you B, and you switch to get C.
1/3: You picked B, Monty showed you A, and you switch to get C.
1/6: You picked C, Monty showed you A, and you switch to get B.
1/6: You picked C, Monty showed you B, and you switch to get A.
But in the second scenario, each of those options is actually 1/4, because he was choosing randomly. Most importantly, each option was a 1/6, but two options where you selected a goat were eliminated because those were ones where Monty selected the car.
[0]: https://gist.github.com/Taywee/2ba202b1bf7af40293ecffb01c2ab...