I played this yesterday and have been thinking about it ever since, although I didn't quite know what to say because it feels so personal and raw and I felt as if I was intruding on your life, except then I would remind myself that you chose to put this out there so it's ok, and you are expecting comments.
I am not fond of AI generated art but I think this is a rare example of a use case exception that works extremely well, especially since the generation is still not perfect and the imperfection (or, rather "something isn't quite right") is a good mirror for your state of mind in the narrative and the journey we experience as we go along with it. If you had made the art yourself it would have been perfect and wouldn't have had as much impact.
I like your specific choice of "New Yorker" style cartoons. (Not just the art style but because one often reads New Yorker cartoons and goes "...what?" - or maybe that's just me.)
The only thing I would change is in the cases where the options all lead to the same image, I would present the image up front so the "player" can select the caption that feels most right to them, giving them a bit of agency in the experience (or maybe that's contrary to the point of the story - one feels as if one has no agency?). With the options that lead to different images I would keep it as is, so the player picks the caption that seems most apt for the story and then the image appears as a surprise - and it may not quite be conveying what the person was expecting, which I think is very appropriate to the confusion and doubt that the narrative wants to convey.