According to [the Shattered paper](http://shattered.io/static/shattered.pdf), the reason why the proof-of-concepts are PDFs is because we are looking at a

> identical-prefix collision attack, where a given prefix P is extended with two distinct near-collision block pairs such that they collide for any suffix S

They have already precomputed the prefix (the PDF header) and the blocks (which I'm guessing is the part that tells the PDF reader to show one image or the other), and all you have to do is to populate the rest of the suffix with identical data (both images)