Talent SummonerHow AI Candidate Ranking Works, and What to Demand from a Vendor
How AI candidate ranking scores people against a role, the four mechanisms sold under that name, and six things to demand before you trust a shortlist.
AI candidate ranking means scoring every candidate against a role's criteria and putting them in order. The phrase sits on tools doing four quite different things underneath, and the difference decides whether you can check the result.
We build Talent Summoner: our ranking scores candidates against your must-haves and tradeoffs, then explains each in plain language so you can audit the decision. Ranking CVs you already have needs no account; sourcing new candidates is $39 a role or $79/month if you hire every month, free to start with no credit card.
It is our product; every claim below is sourced and stands on its own.
Screening hundreds of inbound applications a week? You want ranking inside your applicant tracking system — the last section says what to buy instead.
A few dozen candidates across a handful of roles a year? This page is for you.
The four mechanisms sold as AI candidate ranking
| Mechanism | How it decides the order | What you can audit afterwards |
|---|---|---|
| Keyword and Boolean filters | Presence of terms, titles, and filter values | Only the query you wrote — nothing about the person |
| Embedding similarity | Mathematical closeness between profile text and role text | A similarity score, with no reason attached |
| Structured criteria scoring | Explicit must-haves and nice-to-haves, weighted | The criteria list, and each candidate's hits and misses |
| Model-judged evidence | A language model reads the evidence and writes the reasoning | The written reason for each candidate |
| Talent Summoner (our product) | Must-haves and tradeoffs, scored, then explained | Plain-language reasoning per candidate, free to try |
The top two rank without telling you why; the bottom two produce a reason you can argue with. That difference, not the letters "AI", is what you are choosing between.
What makes a ranking trustworthy
A ranking is trustworthy when you can reconstruct it. That means seeing the criteria it used, the evidence it found per candidate, and what changes when you disagree.
Software applies the same criteria to every CV, which a tired human reviewer does not. Consistency is a real gain — but it is not the same as unbiased, because a consistent tool applies a bad criterion consistently too.
What to demand from any vendor
| Demand | The question to ask | A good answer |
|---|---|---|
| Per-candidate explanation | "Show me, in words, why number four outranked number nine" | A written reason citing evidence from that candidate |
| Auditable criteria | "Where is the criteria list, and can I edit it?" | Visible, editable, attached to the report |
| Feedback loop | "If I reject three of these, what changes next round?" | A named mechanism, not "the model learns" |
| Bias-audit posture | "Is this tool in scope for New York City's Local Law 144, and what audit can you show me?" | A straight yes or no, plus the published audit if yes |
| Data provenance | "Where did this candidate data come from?" | Named sources and a stated position on private data |
| Shareable record | "Can my hiring manager review this without a seat?" | A link that works without an account |
The bias-audit question is not hypothetical, and the duty lands on the employer, not the vendor. New York City's Local Law 144 bars employers and employment agencies from using an automated employment decision tool without a bias audit in the year before that use, public information about the audit, and notices to employees or job candidates.
How Talent Summoner ranks
Two entry points, both free:
- Rank the pile you already have — upload up to 50 CVs and a job description, no account required; every CV scored against the same criteria, with plain-English reasoning and a shareable report.
- Find the people you do not have — describe the role and the agent searches LinkedIn, GitHub, and other public sources, over 200 million profiles, then ranks them on must-haves and nice-to-haves.
In the sourcing flow, your yes/no feedback and preference edits directly adjust the next round's search and weighting — criteria you disagree with reshape the next round rather than re-order the report in your hand. The free ranking tool requires no account, so there you edit the brief and re-run.
Our published data position: LinkedIn profiles and other publicly available professional information.
Pay to match your hiring cadence:
- Occasional: $39 for one role, or $99 for three, which saves $18. Unused roles stay valid 24 months.
- Monthly: $79/month for 4 roles a month; unused roles roll over 3 months, and extra roles are $25.
Trade-offs, stated plainly: no ATS integration today, no always-on pipeline, and no inbound-applicant screening — we rank the candidates you bring us or the ones we source, not a continuous application stream. Hold us to the same six questions.
When a different approach is right
Buy a platform instead of per-role ranking if any of these is you:
- High-volume inbound — SeekOut's published feature matrix gates inbound evaluation and AI screening to its Full Recruiting Funnel tier; that class of product belongs inside your application flow, and our guide to automated resume screening software covers what to look for in it.
- Re-ranking past applicants — ATS rediscovery and silver-medalist search starts at SeekOut's Sourcing plus Integration tier, and per-role ranking has no equivalent.
- A rubric your compliance team owns — job-specific workflows and rubrics sit in SeekOut's Full Recruiting Funnel tier, which is what one enforced scoring standard needs.
If that is you, budget for a sales call. Everyone else: the cheapest way to judge a ranking is to run one real role and read the reasons.
FAQ
What is AI candidate ranking?
AI candidate ranking is software scoring each candidate against a role's stated criteria and returning them in order of fit. The useful versions also return the reason for each candidate's position, so a human can check the decision.
How does AI rank candidates against a job description?
The tool turns the job description into criteria, then evaluates each CV against them — by keyword match, by mathematical similarity, by weighted must-haves, or by a language model reading the evidence. Which of the four decides whether you get a reason or just a number.
Is AI candidate ranking accurate?
Treat any single accuracy number as unsourced until you see the method behind it, because a ranking is only as good as the criteria it was given and the evidence in the documents. Judge it instead by whether the stated reason for each candidate survives you reading that CV yourself.
Is AI candidate ranking legal?
Rules differ by jurisdiction, and some regulate it directly: New York City's Local Law 144 bars employers and employment agencies from using an automated employment decision tool unless it has had a bias audit within one year of that use, information about the audit is public, and notices have gone to employees or job candidates. Check whether the tool you are buying is in scope wherever you hire.
Can I rank CVs I already have without paying?
Yes — Talent Summoner's ranking tool takes a job description and up to 50 CVs with no account required, and returns a ranked report with reasoning you can share by link.
Still deciding? Rank the CVs on your desk with no account, or start free on a live role and let the shortlist argue for itself.
Product facts read from our live /candidate-ranking, /candidate-sourcing and /pricing pages, and SeekOut tier entitlements from seekout.com/pricing, all verified August 2026; Local Law 144 requirements from New York City's Department of Consumer and Worker Protection. We re-verify this page quarterly — tell us if something changed.
