How do I check a resume after an AI rewrite?

Run two checks. Compare the rewrite against your original and your own record, flagging any claim or number that is new — new claims are the ones a model invented. Then re-extract the text from the new file and confirm the structure survived, because rewrites usually arrive in a new template.

A rewrite changes two things at once

This is why checking a rewrite is different from checking a resume you wrote yourself. The content changed — new phrasing, new emphasis, sometimes new claims — and the container usually changed too, because rewrites tend to arrive inside a fresh template.

Most tools only look at one of those. A keyword scorer reads the text and says nothing about whether the file parses; a formatting checker reads the file and says nothing about whether the claims are yours. You need both, and the claims check is the one no tool can finish for you.

The claims check: diff against your own record

Put the rewrite beside the version you started from and read for additions. Anything present in the new version and absent from the old one was supplied by the model rather than by you, and each one needs a source.

New numbers. The most common invention, and the most damaging in an interview. Either source it or cut it.
Upgraded scope words — "led", "owned", "drove" where the original said "worked on". Ask whether the stronger word is true.
Tools and technologies you did not list. A model that sees them in the job posting will sometimes place them in your history.
Whole bullets with no counterpart in the original. Read these hardest.

The parsing check: does the new file still read back?

Our own study is the clearest argument for doing this second check rather than trusting a score. The same resume rendered as a two-column PDF extracted to 19 lines instead of 25 — the sidebar interleaved through the work history — and still scored 98. The number looked fine while the stored text did not.

The check that actually tells you something is reading the extracted text. Select all in your PDF, copy, paste into a plain editor; or use the checker below, which does the extraction in your browser and shows you what it recovered.

What this checker does and does not do

It extracts the text from your file locally — the PDF and DOCX parsing both run in your browser — scores structure and keyword alignment against a posting if you paste one, and shows the recovered text. Nothing is uploaded and no account exists. The single exception is opt-in: if you ask for a sample AI rewrite of one bullet, that bullet is sent for the rewrite and nothing is stored.

It cannot tell you whether a claim is true. Nothing can, except you and your record — which is the argument for keeping one.

Frequently asked questions

Does the checker upload my resume?

No. Text extraction from PDF and DOCX runs in your browser, and there is no account and no storage. The one network call is opt-in: requesting a sample rewrite of a single bullet sends that bullet and a short excerpt of the job description, and nothing is persisted.

Can a checker tell me if the AI made something up?

Not on its own — a tool has no access to what you actually did. What it can do is surface what changed between two versions so you know which claims to verify. The verification itself is yours, against your own record.

My rewrite scored well. Is it fine to send?

A score is a way to rank your own drafts, not a gate anyone is enforcing, and it can be high on a file that parsed badly — we measured exactly that. Read the extracted text and the claims before you trust the number.

Should I go back to my original resume instead?

Usually not. Keep the rewrite's structure and phrasing where it is genuinely better, and pull the substance back to what you can defend. The goal is a document that is both well-written and true, and you rarely have to choose.