What Asking After an AI Rejection Can Actually Get You

Should You Bother Asking at All?

Yes, if the whole message costs you three sentences. Feedback after an automated round is discretionary almost everywhere in the United States, so the realistic expectation is silence. What makes three sentences worth typing is that a reply, when it comes, is sometimes specific enough to change the next application. Nothing about asking obliges anyone to answer.

The person you are writing to may be working at a volume that explains some of the silence. Ashby, reporting on its own customer base of venture-backed technology companies rather than on employers generally, says the average recruiter now processes 291 applications per hire against roughly 100 in early 2021 ². That is a ratio of two company-level totals and never your own odds. Duplicate and automated submissions inflate it too, so read it as a sketch of one kind of inbox, not yours.

Silence is the default, not a verdict on the message you sent.

The case for an explanation has some measurement behind it, though it stays a case and not a right. In a four-scenario vignette experiment with 921 working-age Austrians, a rejection from an AI with no explanation scored lowest on all four outcomes measured ¹. An AI rejection that did carry an explanation drew the same ratings as a human rejection without one ¹.

Those were people on an online panel imagining a rejection, not applicants who had actually been turned down, and the vignette described a poor experience on purpose.

Every condition scored below the midpoint of that five-point scale. This ranks unhappy outcomes against each other, and the gaps are fractions of a point.

Ask Something a Recruiter Can Answer in One Line

Ask a question with a one-line answer, and ask it once. An open request for feedback asks a stranger to compose a paragraph about you and to clear it with somebody first, which is work a stranger has no reason to do. A narrow question about the process asks for a fact the recruiter already has in front of them, and a fact is cheap to send.

• Was this round assessed automatically, or did a person review my answers?
• Which requirement did my application not meet?
• Is this a decision on this role only, or should I read it as a decision on future openings too?

Reading what employers are told about this shows what they are advised to send at minimum. The employer-side guidance on what an employer owes a candidate the AI screened out lands on three things: notice that a tool was used, an accurate account of what it evaluated, and a person who can correct wrong data. A candidate-specific reason is not on that list, because no US rule in force requires one.

Three sentences, one question, and no follow-up.

Send it to whoever signed the rejection, or to the address the invitation came from, within a week or so while the requisition is still open.

Keep it to what happened, not to what you think you deserved.

Where Information Is a Right, Not a Favor

In a short list of places, a written request for information is something the employer has to answer, and that is a different thing from feedback. Feedback is an opinion somebody chooses to write. Information is a record that already exists: what data a tool collected, where it came from, what part a person played. Which of these reaches you depends on the law that covers you and the job, so check each one's conditions before you rely on it.

• EU and EEA. Where a decision about you was based solely on automated processing within Article 22, Article 15(1)(h) of the GDPR entitles you to be told that it exists and, at least in those cases, to meaningful information about the logic involved, along with the significance and the envisaged consequences of that processing ³. A round a recruiter genuinely decided may sit outside it. The Court of Justice, in a February 2025 credit-scoring case rather than a hiring one, read the logic duty as the procedure and principles actually applied to your data to reach that specific result, and held that a trade-secret claim does not end the request ⁴.
• New York City. Where the employer has not published it on its website, the type of data collected for an automated employment decision tool, the source of that data and the employer's data retention policy have to be provided within 30 days of a written request ⁵. Three items, and none of them is your own result.
• California, from 1 January 2027. A business covered by the CCPA that uses automated decisionmaking technology, meaning technology that replaces or substantially replaces a person's decision, to make a hiring decision must answer an access request with plain-language explanations of the specific purpose, the logic, and the outcome, including whether the output was the sole factor and what part a human played ⁶.
• Colorado, from 2027. SB 26-189 requires a plain-language description of a covered tool's role within 30 days after it makes a decision with an adverse outcome for you, and lets you request the personal data used, correct factually incorrect data, and ask for meaningful human review ⁷. The attorney general enforces it, and it creates no new private right of action.

Two of those four do not start until 2027.

A vendor's own notice sometimes offers more than the law where you live requires. HackerRank's Candidate AI Notice lists four things a candidate may request ⁸. They are an alternative selection process or a reasonable accommodation, and human review of a decision made or substantially assisted by AI. The other two are correction of inaccurate personal information and, if an adverse decision is made, an explanation of how AI was used ⁸.

The notice directs those requests to the employer rather than to HackerRank, and it does not say the employer must grant them. Outside a place with a statutory right, the request rests on the employer's own policy.

If you are unsure whether any of this reaches you, where you're entitled to know AI screened your application works through an overlapping list from the notice side. The separate ask, for a person to look at the decision again, is covered in how to ask for a human review after an automated rejection.

What to Do When Nothing Comes Back

Treat silence as the default result and stop after one message. Outside a covered jurisdiction, nothing obliges a US employer to answer a request for feedback or for information, and a second email does not change that. The hour you would spend on a follow-up is worth more spent on the requirement you can now make a decent guess at.

The worry that stops some people from asking is that the request will mark them as difficult. Nothing measured shows how one courteous question about process lands with a recruiter, but the rejection has already happened. What the message will not do is reopen the decision.

Send it for the information, not for the reversal.

One reply is worth acting on the same day. If the answer names a requirement you actually meet, say so in one line and attach the evidence, because a wrong data point is the thing Colorado's 2027 law and HackerRank's notice both name a route to correct. Anything vaguer than that was a courtesy, and a courtesy is what the message asked for.

See how it works >