Re-extracting your CV
On the CV section, once a profile exists, a second button appears next to the file picker: Re-extract from CV. It runs the extraction again — and it is worth knowing exactly what it runs on.
It re-reads the stored text, not a new file
When you import a CV, the text pulled out of the file is kept. Re-extract sends that stored text back to the model and rebuilds your profile from the answer. It does not go back to the document on your computer.
So if you have edited your CV, re-extracting will not pick the changes up. Upload the new file instead — that is a fresh import, and it replaces the stored text as well as the profile.
What it overwrites
Everything the extraction produces, including anything you have typed in by hand: headline, summary, seniority, years of experience, target titles, skills, sectors, must-haves, dealbreakers, your contact block, your work history, your education, your certifications and the Other field.
It asks first — Re-extract from your stored CV? This overwrites the current profile fields (including manual edits). — and there is no undo afterwards.
What it leaves alone
The stored CV text itself, and the name of the file it came from. Your search titles, salary floor, home postcode and maximum commute on the Search tab. The jobs already found for you, your applications, and any documents already drafted. Nothing outside the profile moves.
When it is the right button
- The first extraction came back wrong or thin, and you would rather give it another attempt than type the lot in yourself.
- You have made no edits worth keeping.
And when it is not:
- You have corrected the profile by hand. Re-extracting discards that work.
- Your CV document has changed. Upload it — re-extract cannot see it.
- Only one or two fields are wrong. Fix them on the section that owns them. That is quicker, more precise, and costs nothing.
Afterwards
Re-extraction is an AI call like any other, so it draws on your credit, or on your own provider key if you have connected one. When it finishes you get Re-extracted from your stored CV.
Check the sections before moving on. A second answer is not automatically better than the first, only different, and this is the same reading of the same text. If it comes back wrong the same way twice, the CV itself is probably the problem — a heavily designed two-column layout, or text that is really an image, gives the extraction very little to work with. A plain export usually reads far better.
Because the profile has changed, your existing fit scores go stale and are worked out again on the next run.