Have you ever heard people complain about the quality of reviews on the internet? “A bot can do better than this!” they say. Well, let’s put that to the test and find out. I recently ran an experiment where I fed comments from Kotaku Australia into a machine learning model. This model was based on GPT-2, a free online language model available through Shortly–although more powerful AI models like GPT-3 are now available for purchase via API access. Surprisingly enough, the elements related to experience and emotion turned out quite well in comparison with what we expected from such a tool. However, it seemed as though at every opportunity it adopted a “fake it till you make it” approach; introducing details that were wrong or taking names of developers from different projects altogether! The conclusion: AI is not yet up to par when it comes to reviewing video games accurately. Nevertheless, it’s still interesting to see just how far AI generated content has come over time! We may even reach the point where some news agencies start using GPT-3 generation for press release material due to its sheer volume–even if readership might suffer as result of its lack of human touch. At least when it comes to reviews, nothing beats having another person look at something and assess which elements matter most! While AI will one day be able get there eventually – today is not that day!