Measuring phrase-level overlap between two texts
Paste two pieces of text - for example, your draft and a typical AI-generated answer on the same topic - and this tool measures what percentage of 3-word phrases appear in both, along with the specific shared phrases. It is useful for checking whether your content reads as distinctly your own rather than closely mirroring generic AI-style phrasing, or for comparing any two texts for similarity in general.
Why 3-word phrases specifically, rather than single words or full sentences
Single shared words are far too common between any two pieces of text on the same topic to mean anything, while requiring a full sentence match would miss meaningful overlap in wording that doesn't happen to form an identical sentence - a 3-word phrase is long enough to represent a genuine shared turn of phrase, while still short enough to catch overlap that falls short of a word-for-word identical sentence.
Why the overlap score uses a ratio rather than a raw shared-phrase count
A raw count of shared phrases would make a longer text pair look more similar purely because it generates more 3-word windows to compare, regardless of actual similarity - dividing shared phrases by the total number of distinct phrases across both texts corrects for length, so the percentage reflects genuine overlap rather than just how much text was pasted in.
Why both texts are shown with the overlapping words highlighted
A percentage score and a list of shared phrases tell you how much overlap exists, but not where it sits inside each piece of text. Below the score, both texts are rendered in full with every word that belongs to a shared 3-word phrase highlighted in place - so you can see directly whether the overlap is scattered across a few common transitions or concentrated in one block worth rewriting.
Why this is a different tool from a plagiarism checker
A plagiarism checker searches the web for matches against a piece of text you didn't necessarily have a second copy of. This tool instead compares two pieces of text you already have side by side - useful specifically for checking your own draft against a reference text (like a typical AI-generated answer) you want to differentiate from, rather than for scanning the web for unauthorized copies.