Why comparing versions is harder than it looks
After an hour of editing, nobody can remember what they changed. You know it feels better. You do not know whether you removed a bullet that was carrying real evidence, or whether the section you tightened lost the one number a screener was going to notice.
So this does two separate things. It diffs the two documents line by line - added, removed, reworded, unchanged - and it scores both versions on the same quality dimensions and shows the movement. The diff tells you what you did. The dimensions tell you what it cost or bought.
The free version makes no AI calls at all
Not "usually does not" - the free comparison never contacts a language model or an embedding service. It is a diff and a set of deterministic checks running in our own code, which is why it can be free, why it is fast, and why the result is identical every time you run it on the same two files.
It is also why it needs no account. Upload two files, get the comparison. Your uploads are deleted within 24 hours, and names, emails and phone numbers are removed before anything is stored.
The free version reports no total score, on purpose
One of the quality dimensions - skill coverage - needs a layer the free tier does not run. Rather than quietly re-weight the rest and print a total, we show every dimension we could measure and omit the headline, because a number built from a partial rubric would look exactly like a full Readiness Score and would not be comparable to one.
You still see which dimensions improved and which regressed, which is the part that answers the actual question. The signed-in version adds skill coverage, a real score for both documents, and a written read on whether the rewrite was worth it.
What this tool will not do
It will not tell you which version is better. It shows what moved and in which direction; deciding whether a tighter summary is worth a lost metric depends on the specific role you are applying to, and a diff cannot know that.
It will not merge them for you. There is no "best of both" output, because choosing between two of your own sentences is the editorial judgement that makes it your resume.
It will not compare more than two versions at once, and it is not version control. If you are tracking a resume across ten drafts, compare consecutive pairs.