In virtual laboratories, various cognitive aids are used to improve students' performance and assist them while completing experimental tasks. However, excessive use of cognitive aids in virtual laboratories can lead to a cognitive load on students which affects their performance. In this paper, we proposed the concept of adaptive aids virtual reality chemistry laboratory. User expertise, quantitatively measured in terms of errors and time spent, is used as an adaptation criterion to dynamically update adaptive aids (arrows, text or animations). The system modifies the contents of such aids for good, average, and weak students to give them the opportunity to improve their performance. In this manner, adaptive aids assist the users to perform an experiment in the proposed system correctly according to the correct procedure with high performance. For evaluation, we conducted a comprehensive user study with 59 participants from various institutions. These students were selected randomly by their teachers and contained different levels of students (i.e., weak, average, and good students). We also conducted a written quiz and based on their scores we divided them into three groups (i.e., good, average, and weak). The participants (students) were classified into three groups G1, G2, and G3 based on their expertise levels (i.e., good, average and low). Then each group conducted an experiment four times. Evaluations revealed that the proposed system consistently improved students' performance in four trials. In particular, weak students' performance greatly improved as compared to that of good and average students in terms of a variety of factors, including time and errors during the performance of experiments.