I assume that the very simplified backend of this is that when the message is parsed by Google Messages, it looks for the text in quotes and applies the reaction to the original based off what it sees before the quote. Since image reactions only say that an image is involved it has no way of knowing which and ignores…