GCSE examinations play a powerful role in shaping what is taught in schools and how young people learn to think. As classrooms adapt to an AI-integrated world, there is growing emphasis on students being able to explain their reasoning, apply knowledge in new ways and demonstrate understanding that cannot be generated by digital tools. However, aspects of GCSE exam language may still encourage surface-level responses or rely on complex phrasing that makes it harder for students to show what they know. These linguistic choices matter: they can overload learners, obscure what kind of thinking is expected, and disadvantage students who process language differently. This project investigates how the wording of GCSE exam papers and mark schemes shapes the kinds of thinking students are able to demonstrate. It examines how different levels of cognitive challenge, from recall to explanation and reasoning, are constructed through language across subjects and awarding bodies. The study also explores patterns of linguistic complexity that may create unnecessary barriers to accessibility, particularly for students with additional learning needs or lower prior attainment. Using a large collection of publicly available GCSE materials, the project applies corpus-based methods to identify recurring features of exam language and the forms of thinking they prioritise. It also considers whether current assessment wording reflects policy ambitions for fairness, transparency and AI-era relevance, including the goal of assessing reasoning that cannot be automated. By providing corpus-based evidence of how language shapes assessment in practice, the project aims to support the development of GCSE examinations that are more accessible, equitable and fit for the future.