You have probably already talked to your next hire. They applied for a role eight months ago, interviewed well, and lost out to someone marginally better. Or they were a strong fit for a job that turned out to be wrong for them. Either way they are sitting in your ATS right now, and almost nobody ever looks at them again.
Candidate rediscovery is the practice of fixing that: re-surfacing past applicants and contacts already in your database when a new role opens, instead of sourcing from scratch every time. Some tools call it talent rediscovery. The idea is the same. The people who raised their hand before are warmer, cheaper to reach, and already known to you, and a growing set of tools now finds them automatically. Greenhouse and Ashby both shipped dedicated rediscovery features in 2026; Gem and SeekOut have run it for longer.
This guide walks how far you can automate it, with a setup for each rung, the point where a human has to step back in, and one compliance line that matters.
The job to be done
Rediscovery is sourcing pointed inward. You have a new req, and before you go find strangers, you check the people you already have: past applicants, silver medalists, prospects you talked to once, candidates who went cold. The task is to match that existing pool against the new role and pull out the handful worth a fresh conversation.
It is worth being clear about how this differs from ordinary sourcing, because the two get blurred. Sourcing looks outward, at passive candidates who never applied to you. Rediscovery looks inward, at people who did. They are complementary, and the strongest pipelines use both, but they are different motions with different tools. (For the wider split, see candidate sourcing vs recruiting.) Rediscovery happens inside your ATS, which is also why the ATS you run shapes how well it works.
The automation ladder
Every recruiting task automates to a different degree. Three rungs.
| Rung | What it means | Effort |
|---|---|---|
| L1: one-shot prompt | You export a slice of your ATS, paste it into a chat tool with the job description, and ask which past applicants are worth a second look. | Minutes |
| L2: saved playbook | A reusable rubric and prompt you run against an ATS export for every new role, so the review is consistent instead of ad hoc. | An hour to set up |
| L3: full agent | A rediscovery tool that sits on your ATS, matches the whole database against each new req automatically, and surfaces the people worth contacting. | Tool config |
Unlike resume screening, rediscovery reaches L3 without the same legal wall, because surfacing a past applicant for outreach to a new role is sourcing, not a decision that rejects anyone. The line where that changes is below.
L1: the one-shot prompt
The free way in. Export a relevant slice of your ATS, say everyone who applied to a similar role in the last two years, and paste it into Claude or ChatGPT with the new job description. Ask it to list the past applicants whose experience maps to the must-haves, and to quote the line it based each call on. You get a short list to revisit in a couple of minutes instead of scrolling the ATS by hand.
Where it breaks: the model only sees what you exported, and that data is often stale. Someone who applied two years ago may have changed roles, comp, or location, and the export will not tell you. L1 surfaces names worth checking. It cannot tell you whether they are still a fit, or still interested.
L2: the saved playbook
Build the prompt once. Save your must-have rubric and output format as a Claude Project or custom GPT, then run every new role's export through it. The benefit is consistency: every search starts from the same questions, so you stop missing good people just because you reviewed the list when you were tired. For a recruiter who reworks similar roles often, this turns rediscovery into a five-minute habit at the start of each search rather than a thing you mean to do and never get to.
Where it breaks: L2 still depends on you pulling the export and running it. It also still works off a snapshot, so the staleness problem from L1 remains. Removing both of those is what L3 does.
L3: the full agent
This is where rediscovery becomes automatic. A dedicated tool connects to your ATS, keeps matching the whole history against open and incoming reqs, and surfaces the past applicants and prospects who fit, with the supporting context attached. Greenhouse and Ashby added this natively in 2026; Gem and SeekOut offer it as part of their platforms. Instead of remembering to mine your database, the database mines itself and brings candidates to you. It is the same background pattern as the rest of recruiting workflow automation, pointed at the people you already have.
This rung works well for rediscovery, and here is the important distinction from screening. Surfacing a past applicant to consider for a new role, and reaching out to them, is proactive sourcing. No one is being rejected, so it sits outside the bias-audit and adverse-action rules that govern screening applicants in a live funnel.
That changes the moment the same scoring is pointed at people who are currently applying. If a tool automatically ranks or screens out active applicants to a role they applied for, you are back in the territory covered by hiring laws in places like New York City, Illinois, and California, and our guide to AI resume screening covers that line in full. Keep rediscovery aimed at re-engaging past candidates for new roles, and keep a human on any decision that affects a live application.
Where it breaks, and what stays human
Two things.
The data and the read. A rediscovery tool surfaces a name and a reason, not a current truth. People move on, get hired elsewhere, change what they want. You still have to confirm the person is relevant and available before you treat them as a lead, and the tool cannot do that for you.
The re-approach. Reaching back out to someone you passed over is a delicate message, not a mail merge. "You applied a year ago and we said no, but" needs a human touch to land well, especially with a strong candidate who remembers the rejection. The tool can tell you who to contact. What you say is yours.
One compliance note beyond the screening line above: rediscovery only works because you kept candidate data, so make sure your retention and notice practices actually allow you to hold and re-contact it. That matters more if you hire across regions with strict data rules, and it is worth a quick check with whoever owns your data policy.
Where Glozo fits, and where it doesn't
Disclosure: I work on Glozo. For rediscovery that lives inside your ATS, the native tools above are the right place to look.
Where Glozo helps is joining the two motions that rediscovery usually keeps separate. You can add candidates from your own pipeline into Glozo, source new ones from 30+ external sources alongside them, and then run the whole combined list, the people you already had plus the ones you just found, through a single outreach sequence. Those sequences are recruiter-built and recruiter-sent, not written by AI on your behalf. Before you spend that outreach, Glozo's Open to Offers signal flags who among them is actually likely to respond, which matters most with rediscovered candidates who may have moved on since you last spoke. Rediscovery finds the people you already have; Glozo lets you act on them and your new finds in one motion.