What Do You Do When a Candidate Has No Idea They Applied?

What do you say on the call when they don't recognize the role?

Name the role, say where the application arrived from and when, then ask one question: do they still want to be considered. Skip the accusation. The person on the phone is job hunting in earnest and simply did not read this posting. A yes means run the screen as planned. A no, or a long pause, means close it warmly and record the source, because the source is the finding.

Three lines cover it, and none of them ask you to decide what happened:

• This is the [role] position at [company], applied to on [date] through [source].
• I'm calling to check you meant to apply, before this takes half an hour of your day.
• If you'd rather not go forward, that's fine, and nothing here counts against you.

The third line does more work than it looks like it does. Someone who realizes mid-call that a tool applied on their behalf is embarrassed, and embarrassment produces a yes on Monday that becomes an empty room on Thursday. Give them the exit and you get a true answer, which is the only thing on that call worth having.

Then log the outcome against the source, not the person. One blank call is an anecdote; nine of twelve from the same board in a fortnight is a routing decision. Recruiter time is where this lands: Greenhouse's 2026 benchmarks put annual applications per recruiter at 746 in 2025 against 146 in 2022, a 412% rise measured across over 6,000 companies and more than 640 million applications ¹.

Is an agent-filed application actually fraud?

No, in the ordinary case. Fraud is a person misrepresenting who they are or what they have done. An agent filing on a candidate's behalf is usually working from a real resume, for a real person, under a general authorization given once and then stopped being read. What is missing is not honesty. It is the moment where someone looked at this posting and decided it was worth an afternoon.

The distinction is operational rather than philosophical. Identity fraud is answered by identity checks; a missing decision is answered by asking for one, and neither instrument touches the other's problem. File the second under the first and you will buy verification tooling that leaves your calendar exactly as full of blank calls as it was.

It cost the candidate too, which is the part the fraud framing hides. Their agent spent their name on a posting they would have skipped, and the record now shows a no-show at a company they might genuinely have wanted next quarter. In Greenhouse's 2024 State of Job Hunting survey of 2,500 workers across the US, the UK and Germany, 38% of job seekers reported mass applying rather than pursuing targeted roles ³, a strategy people adopt because the volume on the employer side got worse first, not because they enjoy firing off applications.

The volume is not going back down. Greenhouse's benchmark set puts the average job at 244 applications in 2025 against 116 in 2022 ¹. At 244 applications a posting, reading every ad carefully is a losing strategy for the applicant, and the tools that replace the reading are cheap and one click away. That is the behavior you are actually seeing at the other end of the phone.

Where in the funnel does the intent check belong?

Between submission and any human's calendar. Not in the job post, where nothing has been submitted yet and there is nothing to check. Not at the screen, where the half hour is already committed. It belongs in the automatic reply that fires when an application lands: one act, identical for everyone who applied to that posting, with a deadline, and no scheduling link until it is done.

Four lines is the whole message:

• The role in one sentence, carrying the two facts a posting summary usually loses: where the work happens, and the pay band.
• One act, described in a single sentence, with the time it takes stated honestly.
• A window. Three business days is enough; five if the posting went up on a Friday.
• What silence means: the application closes, nothing is recorded against them, and they are welcome to apply again.

That last line is what keeps the check from becoming a trap. Someone who never sees the message (a spam filter, a dead address the agent supplied, a week in hospital) should lose a posting, not a reputation.

Keep it out of the qualification questions. An intent check is a signature, not a screen: it asks whether they meant to be here, not whether they are any good at the job. Bolt it onto a knockout questionnaire and thirty seconds becomes fifteen minutes, which starts costing you the applicants you wanted. Keyword rules already fail at the task they were given, which is why the ATS filter stopped separating anything useful, and a second automated rule pointed at intent fails the same way.

Order matters as much as content. Everyone gets the same act, before anyone reads a resume, so the check never turns into something you apply only to the applications that struck you as suspicious. That version is a different test given to different people, and it does not survive the first person who asks about it.

Which confirmation act reads as intent in your field?

One a person clears in under a minute and an agent cannot clear at scale. The reliable shape is to ask for something the posting contains and a resume never does, so the answer proves the ad was opened. A checkbox fails that test, because a click is precisely what an agent is built to produce. Two sentences of free text, tied to a detail in the ad, holds up far longer.

By the kind of role you are filling:

• Software engineering. The ad names the stack. Ask which two parts of it they have shipped against and which they would be learning here. A confident answer about a framework the ad never mentions tells you the ad was never opened.
• Financial analysis. The ad names two coverage areas. Ask which one they want and why that one. An agent has no preference; an analyst has a strong one and can give it in a sentence.
• Marketing. Ask which of the named channels they have owned end to end, and which they have only briefed an agency on. That line is invisible on a resume and instant to the person.
• Healthcare revenue cycle. Ask which of the payers named in the ad they have worked denials for. Wrong payer mix, wrong posting.
• Legal operations. Ask which contract type in the posting they have redlined without a lawyer on the thread.
• Hourly and shift work. Ask them to confirm the actual pattern in the ad: Tuesday to Saturday, six in the morning. Here a no is worth more than a yes, because availability is the thing that breaks these hires in week two.

Two acts do this without asking a question at all. A scheduling link the candidate has to open and choose a slot in is an intent act that pays for itself, since it also schedules. So is a short work sample sent to everyone who applied: it establishes intent and capability in the same act, which is the argument for running the work sample before the resume screen rather than after. The cost is that it is a bigger ask, so it has to be genuinely short, and you cannot send four hundred people an hour of work and call it a confirmation.

Whichever you pick, use one act per posting and hold it for the whole round. Two different checks on two applicants is two different processes, and the comparison you wanted at the end of the round is gone.

What breaks if the check asks too much?

Your real applicants, first. An intent check filters on available effort, and effort is not evenly held: a parent answering at eleven at night, a candidate on a screen reader, someone applying from a phone on a break all lose to a five-minute recorded video. Keep the act under a minute, accept a plain written reply as valid, and never let a missed window become an adverse note against a person.

Three failure shapes, in the order they usually turn up.

The check becomes a stage. It starts at thirty seconds, someone adds a question, and four months later it is a pre-screen with six fields and a 40% completion rate. Time it against its own clock every quarter and cut anything not asking whether they meant to be here. The general form of that problem, adding a step without lengthening the loop, has the same answer: a new step has to replace something rather than stack on it.

The completion rate gets read as a candidate signal. It is not one. A source where 30% confirm is telling you about the source, not about the people who came through it. Move the budget and leave the individuals alone.

Silence gets filed as ghosting. SHRM's 2025 Talent Trends research found 41% of the organizations surveyed were seeing candidates ghost during the interview process ², and an intent check will pull some of that population out before a recruiter ever meets them, which is the point. An unanswered confirmation email is not the same event as an empty booked interview, though, and recording it as one corrupts the number you were trying to repair. Watch how the rest of the funnel metrics move once a real share of applications stops converting on purpose.

The ceiling on this is low. Confirming intent proves attention and nothing more: this person read the ad and meant to be considered. It says nothing about whether they can do the work, and there is a version of this that goes quietly wrong: a funnel that becomes excellent at measuring who answers email within three days. The evidence about judgment still has to come from work, and the earlier that work sits, the fewer blank half hours anyone spends. See how Olive measures this.

See a sample report >