Unfortunately the AI didn’t feel Human enough. I found a technique where he will only place it down in one of two places once you begin, from there I can make a second move to make it impossible for the AI to win. It needs to learn from its mistakes, or appear to. I think it looks like it’s judging the best move based upon some sort of rating system, but it is only predicting one turn ahead, maybe it needs to go two ahead.
What a lot of AI’s I have experienced do is look at all possible moves ahead and have a certain number of “iterations” allowed which is usually linked to difficulty, since if an AI can see every single move possible then they will never lose since they don’t have a margin for error, however if you lower their maximum iterations they are constricted as to how much they can search through.