That is true, their conversion understands text regions and various other things. However, what makes Scribd's viewer more sophisticated is that it will actually use structured HTML to render the document content. This is more than just putting on a layer that specifies regions in the document, it will actually just be a normal HTML document, made of divs, text, images, etc.

Plus, it will maintain the fidelity of the document -- meaning that even PDFs with complicated layouts will be rendered properly in HTML. No trivial task.