Cognitive Radio is an advanced enabling technology for efficient utilization of vacant spectrum due to its ability to sense the spectrum environment. Various detection methods have been proposed for spectrum sensing, which is the key function in implementing cognitive radio. Most of the existing methods put their interests in detecting TV signal and wireless microphone signals. In this paper, we explore the cyclostationarity of the equally spaced pilot subcarriers in OFDM signal. By averaging over several symbols, the peak property is able to be captured. Simulations in various fading environments show that the proposed cyclostationary based detection method works well for OFDM signal.