One cannot talk about mastering services or mixing, for that matter, without speaking about AI. It is a good question. And I have evaluated many plugins and services like this.
Overall, it’s a quick fix if you don’t know how to do a master yourself and want a standard result that’s loud. AI is good at repetition and automation of routine tasks. Automated online audio mastering services can do a basic job using their algorithms, making the project sound loud and reducing some peaks and resonant frequencies. However, using such processes is like using a preset over all your hard work. Moreover, essential details will likely be overlooked, and you have probably tried your hardest not to sound like everyone else.
Unsurprisingly they are not considered creative and capable of identifying complicated issues that may require more abstract thinking encompassing artistic and engineering decisions. They cannot differentiate by genre if that vocal was meant to be upfront or pushed back just for one section. It would not be able to make a creative judgement of when corrective EQ should be applied or maybe if automation should be applied elsewhere but not for one bar because the artist did something deliberate.
It is a shortcut, and if you are at this stage, chances are you are looking to release your project commercially. So, make sure you give it the best possible chance of success by having a real person listen and make changes (if they are needed). And then communicate specific revisions if you need to do so. You can’t get that with AI. It cannot react emotionally to a song to hit that sweet spot.