To check for plagiarism, SafeAssign uses a proprietary algorithm to match the submitted text to text from several other sources. These sources include archives of previously submitted SafeAssign work, publications, articles and the Internet in general.
The algorithm used by SafeAssign analyzes the submitted document and attempts to match words and phrases to its database of sources. The sources include all work from a particular institution that has been submitted previously as well as student-volunteered works. In addition, SafeAssign searches both its own publication database, called ProQuest ABI/Inform, and the Internet using the Yahoo! BOSS search service. SafeAssign updates the publication and article database weekly, while the Internet database is updated every one to three days, as of 2015.
The results from these searches are given a weight. The SafeAssign algorithm modifies this weight and produces an originality report, which tells what percentage of the submitted text matches other sources. In addition, the originality report highlights which sources the text may have come from.
SafeAssign does not guarantee that a low originality score means that the student plagiarized, nor does it guarantee that a high originality score means the student has not plagiarized. Instead, the score serves as a warning that plagiarism may or may not have occurred.