In the early days of google... back in the 90's... I was listening in on a discussion between Google developers and some folks at NCSA at the U of I...
the outcome of that discussion was a system could be used where the first (or last) occurrence of up to 13 consecutive words after a secret offset could be used to determine if the snippet is duplicated. There are 250000 words in the English language and 25000 are used very regularly. So the odds of someone else writing the exact same 13 digit base 25000 number are pretty astronomical... grains of sand on a beach....
My take away, which has served me flawlessly for the entire life of google, is to make sure at least every sixth word is different. This covers the possibility of a boundary condition between two 13 word blocks from the unknown offset.
I admit that I do not know for certain that this is the exact system they used then or now, but I would bet it is similar.
If I were Google I would not consider stop words like "the,and,a,..." I think the easiest way to defeat such a system is to create uniqueness by only adding words... example: "the blue car" becomes "the very blue car"... saves context and readability and is easier to do than thesaurus lookups in code or in your head. Also increases your word count over the original if that matters at all.