I tried searching around and couldn't find anything similar, but I have this idea I want to test and see what people think about it. Given a non-spatially registered image that must be georeferenced, I can calculate the scale of the map (Given a scale bar or the scale value) and use this value along with actual paper space on the map (Inches on a piece of paper, whether it be 8.5x11 or a full sized plate) to determine the distance that this point is from any other point on the map in actual physical space (We'll just call this actual space-in contrast to paper space).
So in theory, if you have 1 visual point you can add this anchor point to (Referred to as a control point or link in the georeferencing toolbar) regardless of whether it be a topographic, geological, or any other feature, a python or vb script takes this control point and it's coordinates (actual space) and converts the paper space coordinates (most likely in inches or pixels converted from inches) to actual space in ArcMap by automatically adding 4 or however many more needed control points to the image based on the calculated map scale and the anchor point?
I have done this manually, so I know the scale factor works surprisingly well. Even without the actual physical map, you can measure the scale bar of a map in Adobe Acrobat in paper space (Dimensions of the paper map copy) and calculate the scale value this way. Then you can measure an X distance and a Y distance (I like to measure to the maximum extent of the map-so usually the boarders) from the visual anchor point (Control point) you have. You multiply the paper dimensions X and Y by the scale value, and it gives you the actual physical distance that point should be from your anchor point. So in Arc, you can measure the distance from your anchor point (making sure of course it's a straight line) and place a point or marker to use to manually add the control point to.
I've attached various images which to help support my explanation of doing this manually. If I could have some help writing something to automate this process, that'd be great.
P.S. I've gone back to remove/crop/blur some things from the images being georeferenced to try and protect the identity of everything. So try to ignore that stuff. This was just meant to be an example.