How Plagiarism Detection Actually Works Under the Hood

Plagiarism detection isn't a solved problem. It's a spectrum of techniques, each with different strengths and failure modes. Understanding how these systems work changes how you think about originality, citation, and the difference between inspiration and copying. N-gram fingerprinting The most common technique is n-gram comparison. The system breaks your text into overlapping sequences of n words (typically 3 to 7 words). "The quick brown fox jumps over the lazy dog" with n=4 produces: "the quick brown fox", "quick brown fox jumps", "brown fox jumps over", and so on. Each n-gram is hashed to create a fingerprint. The system compares your fingerprints against a database of fingerprints from indexed sources. Matching fingerprints indicate potentially copied passages. The value of n matters. With n=2, you get enormous numbers of false positives because two-word phrases are common. "The system" appears in millions of documents. With n=10, you miss paraphrased content because any word chan