I can only imagine it's incredibly difficult and would require quite a bit of subjectivity, even for academics. I don't have any insight into that. However, being a game designer, I have a related point to make: "game" output should be separated into two distinct categories when considering accuracy. They are:
1) Video games geared towards entertainment:
Historical accuracy will ALWAYS take second fiddle to providing a great gaming experience, as it should.
2) "Serious" games and simulations:
Some of the same academics that do all of the other research on these topics are now using game engines and 3D modelling tools to make accompanying models, and though I haven't seen any personally, probably more involved simulations. They're no more or less likely to be accurate than any other academic work on the topic. Critique by other academics seems like a pretty traditional way to advance a field even if the medium is pretty new.
Assessing the historical accuracy of an entertainment-based medium seems to me a bit like assessing jokes for factual accuracy: perhaps useful as a qualitative study, but probably missing the point if it's a quantitative measure of value. That said, if the marketing copy makes specific claims, then those claims absolutely deserve to be evaluated for accuracy.