I think interactive is used for too many different experiences/actions and that that's what Manovich is getting at in some ways in this chapter. For the painting example, that relates to reader response and how all texts are percieved and how that perception creates an interaction. Then, there's the level where the reader/user can actually change the text itself - and these two blur because all of these occurs within perceptual limits.
I think part of the issue for Manovich is that it seemed like many early hypertext scholars were saying that hypertext was interactive in totally new ways, which most of them weren't. Manovich is then arguing that all art is interactive in fundamental ways, which reader response criticism also argues. Hypertext theory doesn't dispute this at all, it just argues that there's another level of interactivity because the user/reader's actions determine the story structure, and that makes the physical form of the story connect in different ways so that both the reading process and the text being read are interactive. People do read paintings in an interactive manner, but the painting itself doesn't change.
To argue whether or not it matters that the painting changes given the fact that the painting exists in the perceptions of the reader could be a valid argument for Manovich, but hypertext still does change and those changes are perceivable and quantifiable, so this is in some way interactive.