Exams provoke stress. Among researchers there is no consensus if stress enhances or decreases human task performance, especially in exam-like situations. We propose an alternative way of capturing experienced stress through measuring electromagnetic waves captured by changes in the action potentials of a basil plant. Thus, we conducted a lab experiment where 30 participants individually attended a short exam that does not require any special prior knowledge. Additionally, we tried to capture unbiased signals and build a machine learning model that uses the human electromagnetic waves to predict exam performance and estimate stress. Toward that goal we used the plant SpikerBox which captures the electrical signals of a basil plant positioned near the human experiencing stress. We also try to answer the question to which extent stress influences the exam performance of students. We developed machine learning models that predict preliminary grades of students with over 90% accuracy right after the exam has been taken, based on the signals of the basil plant.