The rapid growth of non-fungible token marketplaces (NFTMs) has led to a significant rise in NFT plagiarism, especially for digital assets stored on public blockchains like Ethereum. However, detecting and curbing NFT plagiarism presents three major challenges. First, NFTMs typically operate on different blockchains, each with different underlying technologies and standards, leading to fragmentation and data silos. Second, current techniques for detecting plagiarized NFTs, such as similarity detection and digital watermarking, are largely ineffective across multiple NFTMs. Third, the non-transparent processes for NFTMs to certify the originality of digital assets create an environment in which malicious ones can conceal plagiarism. To address these challenges, we propose CAS, a Cross-Market Anti-plagiarism Scheme. First, CAS constructs consortium blockchains to create a shared infrastructure for transaction records and NFTs metadata across multiple NFTMs, effectively tackling the problem of data silos. Moreover, a reward and punishment mechanism is proposed to incentivize the maintenance of consortium blockchains. Second, CAS incorporates an improved consensus protocol that effectively detects the NFTs’ originality before minting and improves transaction transparency, addressing the last two challenges. Third, to further curb plagiarism, an elaborate reputation mechanism is proposed to increase the cost of plagiarism. Finally, our theoretical analysis and comprehensive experimental simulation confirm CAS is effective, scalable, and robust.