The Internet of Vehicle (IoV) aims to provide efficient vehicular communication by enhancing connectivity and user safety. IEEE 802.11p, a medium access control protocol, plays a vital role in the IoV protocol stack, which uses the enhanced distributed channel access (EDCA) method for channel contention. To deal with safety and nonsafety vehicular applications, EDCA adopts four access categories (AC), and each AC manages the channel access based on the priority of packets along with the contention window size and arbitration interframe spacing number (AIFSN). However, these contention parameters are fixed for each AC, and it does not account the real‐time network traffic. Hence, the fine‐tuning of AIFSN values and maintaining strict priority are challenging tasks for the highly dynamic vehicular network. To address these issues, this article proposes a novel probabilistic approach, namely, the randomized AIFSN tuning (RAT) algorithm for EDCA. It updates AIFSN value for every retransmission based on the vehicle density and contention intensity by preserving the strict priority behavior of the AC queues. A three‐dimensional Markov model is developed to analyze the proposed algorithm in terms of collision probability and throughput. The proposed RAT algorithm is simulated using the NS3 network simulator along with simulations for urban mobility (SUMO) for the real‐time traffic, and results are compared with other related algorithms in terms of packet loss, goodput, packet delivery ratio, and delay of each AC queue. These results reveal that the proposed RAT algorithm outperforms related algorithms under varying vehicle density.