What would be fascinating though (and Google actually could do it) is take all the content related to a game from every individual interacting with it on YouTube and come up with a brief review.

Yup, there are plenty of data services training AI to translate text to speech utilising various accents, intonations and inflections. This is almost certainly the future of such services, where they will have the AI transcribe video content, collate user generated content that meets their "sensitivity" standards, and then summarise based on the collected information.