In simple terms, a plagiarism checker is merely a tool that is used to check the authenticity of the document, to see if the content is plagiarised or not. Most companies tend to hide their secret methods of plagiarism checking but so far, there are few tradition techniques with which the entire concept works. For example, fingerprinting, string matching, citation detection, vector space model, stylometry etc.
Fingerprinting- A fingerprint of a file is actually a unique code that is set by the original document publisher. For example, MD5 fingerprint or hash. It basically converts a huge file into a 32-character fingerprint to define its uniqueness. So, when every content has a unique code, the plagiarism checker has to perform less processing to compare two identical works, which becomes very fast and accurate.
String matching- In this method, the tools take a string of text from one document and try to find in different other documents. It requires more resources and more time, especially when you’re checking a gargantuan document. But it’s very useful when the content is not verified or it's just a blog post.
Citation detection- This method only works with academic documents, especially when there are citations mentioned. Because the tool only compares the given citation to the original documents, judging the citation will be the same in both of them.
Vector space- This model actually analysis the words and compare it to the original documents. Even if you have heavily re-written the sentence to avoid plagiarism, you still write some of the original phrase or words in the sentence, which initially results in plagiarised content.
Stylometry- This method is quite debated, the stylometry refers to comparing the style of the writing to its similarly written content. It tried to detect the person’s unique writing style and compare it to others. Quite unique but also sometimes arguable.