A few weeks into a search you usually understand the role better than when you wrote the post. This recipe turns that new clarity into a re-ranked shortlist: edit the description, and Chosen re-scores every candidate against the sharper version. It's worth knowing because most recruiters edit the post and never realize the rankings moved underneath them — or that their saved search didn't.

The whole recipe hinges on two behaviors: match ratings recompute when you edit a description, and a saved search is a snapshot. Learn both as one story and you'll never be misled by a stale shortlist.

What you'll need

An open job with candidates already scored against it.
A clearer idea of what the role actually needs — the thing that makes the edit worth doing.
A saved search over those candidates, if you want the re-rank step. Optional, but it's the point.

The chain

Edit the job description. Open the job and sharpen the description — swap vague lines for specific requirements, the things that genuinely matter for this role. A strong description produces strong, testable claims; a vague one produces a soft score. You can also ask HQ to sharpen the description for you.
Match ratings recompute automatically. You don't trigger anything. Because the match rating is derived from the job description, editing the description makes Chosen rebuild the claims and recompute every candidate's score for that job. The numbers update on their own.
Re-run the saved search. Here's the catch: a saved search is a snapshot from when it last ran — it does not reflect the new scores until you run it again. Re-run the query from the bar in AI mode. The fresh run re-scores against the recomputed ratings.
Review the re-ranked shortlist. Open the results. Candidates rise or fall against the sharper description — someone who matched the old vague post may slip, someone the post used to overlook may climb. This re-ordered list is the one to act on.

Where it can break

You edited the post and trusted the old shortlist. This is the whole trap. The scores recomputed; the saved search didn't. Acting on a search you haven't re-run means acting on pre-edit rankings. Re-run it after any meaningful description edit.
The re-rank barely moved. A cosmetic edit — reworded for tone, no new requirements — produces nearly the same claims and nearly the same scores. If you want the ranking to shift, change what the role actually asks for, not just how it reads.
A good candidate dropped. A sharper post can add a requirement a strong person's resume never mentioned, and an unmentioned requirement counts as a gap. The match rating is a sorting aid, not a verdict — read the candidates who fell, not just the ones who rose.
The description got narrower than the role. Over-specify and you'll score down candidates you'd actually hire. Describe what genuinely matters, not every wish.

Variations

No saved search yet. If you never saved a search, just sort the candidates table by the match rating column after the edit — the column reflects the recomputed scores directly. Save the search this time so the next edit is a one-click re-run.
The edit changed who the role is for. A big enough pivot is a different search. Run a fresh smart or deep search with new wording rather than re-running the old query.
Re-rank, then advance. Once the shortlist is re-ordered, bulk-move the new top band to phone screen — see turning applicants into a shortlist.
The post is also your careers page. The same description feeds the public careers page candidates read. Sharpening it improves the score and what applicants see — effort that pays off twice.