This is a question I've pondered myself. One major problem with heatmaps that I can think of is that I often move my cursor out of the way of what I want to look at, whether they be images, video, or text. So aside from when I'm actually moving the mouse to click on a link or button, I usually place my cursor in an empty area like the page margins to not block any content.
There's also little evidence that people will move their cursor towards non-clickable page elements which they're interested in or are focused on.
So it's most-likely only going to reveal what your most-often-clicked elements are, and maybe also highlight the typical path that the cursor travels to go between these elements. So if you have a CTA and a login button that are very popular with visitors, the heatmap might show a region between the two that is highly trafficked because it's the most direct route between two frequently clicked items. But that doesn't mean users are particularly drawn to the content in that region.
Maybe cursor-movement-generated heatmaps take these issues into account some how (I believe some marketing materials for mouse-tracking heatmap services do show side-by-side comparisons between eye-tracking-generated heatmaps and their own to show that their heatmaps are a reliable substitute). But if they don't, the results could be very misleading for people who try to interpret the heatmaps the same way as eye-tracking heatmaps.