I'm currently in the process of writing the proposal for my thesis work, which currently involves a vast quantity of reading. Because I do computational work, I've got even more literature to wade through. First is the computational stuff related to the kind of systems and calculations I'm interested in. However, in order to do good simulations, you need to have experimental data to compare to and validate your model. So every literature search I do, I pretty much end up doing twice.
Ultimately, you hope to find something that no one else has really done that somehow relates to your interests and preliminary work. That's really what the thesis is supposed to be: your novel contribution to science and human knowledge. You may know very early on in your research what that is. Sometimes, finding your little research niche comes about in more roundabout ways. It may be a conversation with faculty or fellow students, or an impulsive change of search terms, but when you find the gap, it feels awfully good.