Knockout Questions in Recruiting: Why Ours Flag a Candidate Instead of Rejecting Them

A knockout question is the fastest filter a recruiter has. One yes/no answer, and a candidate who cannot do the job leaves the pile before anyone reads a CV. Most applicant tracking systems wire that answer straight to a rejection.

We built PlacementFlow's screening to ask the same kind of question and then do something different with a failed answer: flag it for the recruiter, never reject on it. This post covers what knockout questions are, why an agency should be more careful with them than an in-house team, and how to write ones that hold up.

Written in September 2026. This is how we read the rules and what we built, not legal advice. Check the rules that apply to your desks with your own counsel.

What a knockout question is

The definition most hiring software uses is short. In Truffle's words, "A knockout question is a mandatory screening question with a single correct answer. Get it wrong and you're automatically disqualified."

The typical ones:

• Are you authorised to work in the country where the role is based?
• Do you hold the licence or certification the role legally requires?
• Can you work on-site in the city where the role is based?

What happens next is the part that matters. In most ATS workflows, "candidates who give a disqualifying answer are automatically routed to a rejected status, often with a templated email." Nobody reads the answer. Nobody reads the CV.

Why recruiters use them

Time. One vendor claims that "auto-rejection of candidates who fail knock-out filters can eliminate 30-60% of the application pool before a human reviews a single CV." That is a vendor's estimate for high-volume roles, but the direction is right: a good knockout question removes the candidates who were never going to work.

The problem is the second half of that sentence. Before a human reviews a single CV.

When a client asks me to just auto-reject anyone without X, my answer is short: "We don't auto-reject on it. The mismatch gets flagged, I look at it myself, and if it really matters I call you about it." The rest of this post is why.

Why an agency carries this risk directly

An in-house team rejecting its own applicants is one thing. An agency filtering candidates on a client's behalf is named in the law in its own right.

• Title VII names employment agencies. The statute says "It shall be an unlawful employment practice for an employment agency to fail or refuse to refer for employment, or otherwise to discriminate against, any individual because of his race, color, religion, sex, or national origin" (42 U.S.C. § 2000e-2).
• A neutral-sounding filter can still be the problem. Under the same section, once a practice is shown to cause a disparate impact, the question becomes whether it is "job related for the position in question and consistent with business necessity." A knockout question that screens out one group more than another has to earn its place.
• There is a number for "more than another". The federal guidelines treat a selection rate for one group "which is less than four-fifths ( 4/5 ) (or eighty percent) of the rate for the group with the highest rate" as evidence of adverse impact (29 CFR 1607.4). An automated knockout applies the same cut to everyone at scale, so if it skews, it skews on every candidate.
• Citizenship questions reach agencies too. The immigration-related discrimination rule covers conduct "with respect to the hiring, or recruitment or referral for a fee, of the individual for employment" (8 U.S.C. § 1324b). Recruitment or referral for a fee is what an agency does. "US citizens only" as a knockout, where the role does not legally require it, is exactly the kind of filter this rule is about.
• Health is off limits before an offer. The ADA says a covered entity "shall not conduct a medical examination or make inquiries of a job applicant as to whether such applicant is an individual with a disability or as to the nature or severity of such disability." It also says "A covered entity may make preemployment inquiries into the ability of an applicant to perform job-related functions." (42 U.S.C. § 12112). Ask whether someone can do the essential task. Never ask why they might not.
• The software vendor is not a shield, for you or for us. In Mobley v Workday, the court let the claims proceed on the basis that the complaint "adequately alleged that Workday is an agent of its client-employers" (Maynard Nexsen). We take that personally: it is the reason our screening is built the way it is.

If you place candidates in Europe, two more rules apply. GDPR Article 22 says "The data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling" (GDPR Art. 22). And the EU AI Act lists "AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyse and filter job applications, and to evaluate candidates" as high-risk (Annex III). An automatic rejection on a single answer is the plainest example of both.

What we built instead: flag, never reject

PlacementFlow calls them must-knows, because that is what they are to a recruiter: the few facts you must know before a call is worth anyone's time. The candidate answers them first, one at a time, in a short web chat, before any scored questions.

What happens to a failed answer is the whole design:

1. A failed or missing must-know answer becomes a flag, not a rejection. The candidate's screening carries on and completes normally. On the recruiter's side, the response shows a must-know flag with the candidate's own answer.
2. The candidate never gets an automated rejection. They get the same neutral close as everyone else. Nothing is sent that says "you did not qualify".
3. Only a recruiter can decline. A decline is a human decision, recorded against the person who made it. The AI never sets it.
4. Flagged candidates are never auto-advanced. A flag stops automation for that candidate until a person has looked.
5. A yes/no must-know needs its expected answer. The recruiter records which answer they are hoping for. A must-know without one cannot raise a flag, so it is not treated as one.

What stays out of must-knows on purpose:

• Relocation. Whether someone will move is a preference we score, never a gate, and the screen never asks why not. The "why not" is where family and health information arrives, and that is not something to collect before an offer.
• Security clearance. It stays a soft filter a recruiter can search on, not a knockout, because a clearance knockout easily turns into the citizenship filter described above.
• Salary history. Never asked, anywhere in the flow.

The trade-off is honest: a recruiter reads a short list of flagged candidates instead of never seeing them. In return, nobody is rejected by a single answer nobody read, and every decline has a person's name on it.

There is no auto-reject path in the code, so no candidate can be lost to one.

What our must-knows look like

These are the questions our own desk asks first, word for word:

1. "Where are you currently based? City and country is perfect."
2. "Are you legally authorized to work in your country of residence, without needing visa sponsorship now or in the future?"
3. "What's your notice period, or how soon could you start something new?"
4. "What working setup suits you best?" (Remote, Hybrid, On-site, or Flexible on this)
5. "Are you open to other cities, or to relocating, for the right move?" (Open to relocating, Open to specific cities, Staying where I am, or Remote only)

Only the second one can raise a flag. It is a yes/no question and the answer we hope for is yes. A no, or no answer, does not end anything: it puts a flag on the response, and a recruiter decides whether it matters for the roles in front of them. The other four are facts we need before a call, recorded as the candidate gives them. None of them is a gate, and the relocation question offers "Staying where I am" as a normal answer, with no follow-up asking why.

A checklist for writing knockout questions that hold up

Use it whether or not your tool auto-rejects.

• Tie it to the job, not the ideal candidate. If you cannot say which duty the answer protects, it is a preference. Score it or drop it.
• Ask about the ability, not the reason. "This role is on-site in Austin, does that work for you?" is a question about the job. "Is there anything that would stop you coming in?" invites exactly the answers you must not use.
• Keep work authorisation to what the role actually requires. Match the wording to the legal requirement of the job, and no further.
• Ask every candidate the same words. Consistency is part of what makes a screening step defensible. Our screen delivers each question word for word from the recruiter's approved set.
• Record the answer you expect. A yes/no question without an expected answer is not a knockout, it is a survey.
• Keep a human between the answer and the decision. Flag, then read. If your tool can only reject, at least review a sample of the rejections every month.
• Never ask for salary history, date of birth or health information.

If you run screening on PlacementFlow, the setup is covered in Set Must-Knows in the Question Builder, the full flow in How the Two-Stage Screening Gate Works, and the boundaries of the AI in What the AI Does and Never Does.