In the growth of Transforms such as Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), Graph Based Transform (GBT) has its steps in signal processing applications. In this paper, the weights of the edges in GBT are adjusted to form an improved GBT. The improved GBT is implemented in audio watermarking. The proposed method is tested on NOIZEUS speech database and MIR1k music database and proved that the improved GBT outperforms its base GBT method against various attacks. Also, the quality of the audio signal is measured after watermarking using Peak Signal to Noise Ratio (PSNR), Short-Time Objective Intelligibility (STOI) and Perceptual Evaluation of Speech Quality (PESQ). The proposed method obtains appreciable results than its base GBT method.