In this paper, we focus on the chipless radio-frequency identification (RFID), where the tag information bits are encoded by the peak/notch pattern appeared in the frequency spectrum of the radar cross section (RCS) of the tag. In particular, we restrict our attention to a simple yet prevalent “binary” coding method, where a bit 0 or bit 1 is encoded by the absence or presence of the peak/notch, respectively. We provide an information-theoretic framework for the tag identification based on such a binary coding method. Our aim is to accommodate more bits in the limited bandwidth without degrading the identification performance. To this end, we first formulate the detection of each bit as a binary asymmetric channel (where a signal processing approach is integrated into each interrogation to enhance the underlying channel quality). Moreover, it is proposed to perform multiple interrogations with majority rule-based detection (in correspondence to the signal processing approach in each interrogation). Furthermore, we introduce some error-detecting codes to further improve the performance of tag identification. For instance, motivated by the asymmetric property of the channel model, we propose to apply the constant weight code and the Berger–Freiman code (as two representatives of non-systematic and systematic codes, respectively) to the problem to be addressed in this paper. In addition, an investigation is also conducted into the cyclic redundancy check (CRC) codes (as a representative of those codes that are not dedicated to the binary asymmetric channel but could be potentially competitive for error detection). The system’s performance is analyzed through the key parameters, namely the successful transmission rate, the false identification rate (i.e., the probability of undetected errors), and the expected number of retransmissions/interrogations. The effectiveness of the proposed methods is demonstrated by the numerical results.