In passive multiple-input multiple-output (MIMO) radar, the transmit signals of the noncooperative illuminators of opportunity are usually not completely known. They are usually standard communications signals following specific protocols which stipulate the preamble format. In this article, we show that the preamble information can be exploited to improve the performance of passive MIMO radar. We have derived two generalized likelihood ratio tests (GLRTs) for passive MIMO radar detection, one for the case where the noise variance is known, and the other for the case where the noise variance is unknown. Our analysis shows that the derived two GLRTs have constant false alarm rate. Simulation results show that under the same condition the derived GLRTs outperform the GLRTs without using the preamble information. Moreover, the derived GLRTs are also compared with the GLRTs for active MIMO radar that totally knows transmit signals. These comparisons show that the performance of the derived GLRTs vary between the GLRTs for active MIMO radar and those for the passive MIMO radar with the transmit signals completely unknown.