Trying to read the generated article gave me a very similar feeling compared to the feeling I get when trying to read an article that's much too difficult for me. I read a sentence or two and try to grasp at the meaning of the sentences and how they relate to each other. I reread these words a few times, and eventually draw out a little bit of understanding and move on to the next sentences.
In this case with the generated text though, eventually my mind concludes that it's all nonsense and gives up on finding any meaning. With a real text, hopefully, I'd derive more meaning little by little as I struggled on.
What's interesting to me here is that a nonsensical text can produce the same "in over my head" feeling as a highly intellectual and difficult text. It seems to imply that we could be fooling ourselves when reading difficult texts - perhaps a text is all nonsense, or perhaps it's brilliant, but we have little way to tell.