It is definitely different from seeing something. I'd guess it activates a subset of the semantic level neurons that activate when you see a thing. There are no actual pixels. I also guess that something similar happens in dreaming.

When I commented earlier that when I visualize an apple it's one with that vertical stripy pattern some varieties have, that means I get the semantic vector of the stripes. It's not a stable image that I can draw - if I draw it, I'm recreating the pixels based on the semantic content. I think.