You only save new copies of the files that change from version to version. Dat saves entire copies of the files and not just the changes (which is how e.g. git does it.) Changing the copyright footer on a website would generate new copies of all the pages and result in the Dat archive doubling in size.
There isn’t currently a solution to this in Dat.
There’s no compression or data deduplication to mitigate the effects either.
I’ve tried maintaining a Dat version of my own websites, but stopped because the archives consumed an inordinate amount of disk space. More on this.
Users normally only download the parts of a Dat archive that their browser has requested (sparse mode) and don’t need to fetch all the previous revisions of files. However there’s currently a bug with the Windows implementation that cause Dat archives to take up the full size of the entire archive on disk even though it only contains a few kilobytes of data.