How to Ask What an Automated Hiring System Did—and What You Can Actually Compel

Last reviewed: August 12, 2026. This guide is informational and distinguishes voluntary requests from rights that may exist under specific laws.

The short answer: seven things to request

If you suspect that automation influenced a hiring rejection, ask the employer or employment agency for:

1. Confirmation of whether an automated or AI-based tool was used.
2. The hiring tasks the tool performed, such as résumé screening, ranking, testing, scoring, or interview analysis.
3. The tool’s influence on the outcome, including whether it recommended, guided, materially influenced, or effectively determined the rejection.
4. The principal reasons and general evaluation criteria behind the result.
5. The personal data considered, where applicable.
6. A way to correct factual inaccuracies that may have affected the evaluation.
7. Meaningful human review and reconsideration, preferably by someone qualified and authorized to change the result.

These are requests you may make—not rights every employer must honor. Legal obligations depend on the jurisdiction, employer type, decision date, technology, degree of automation, human involvement, and applicable exemptions.

What you want	How it is usually obtained	Important limitation
Confirmation that automation was used	Voluntary employer request; potentially required through a covered notice or disclosure process	There is no universal post-rejection duty to answer
Description of the tool’s hiring task and role	Voluntary process request; potentially mandatory jurisdiction-specific disclosure	Screening, ranking, scoring, and decision-making may receive different treatment
Principal reasons and general criteria	Usually a voluntary request; sometimes addressed by applicable privacy or automated-decision rules	A general process description is not necessarily an individualized explanation
Personal data used	Potentially available under a covered privacy or automated-decision law	Coverage, verification, procedures, and exemptions vary
Correction of inaccurate personal data	Voluntary unless an applicable law or policy provides a correction process	Disagreement with professional judgment is not necessarily a factual inaccuracy
Human review or reconsideration	Voluntary in many places; potentially required for certain covered Colorado or qualifying EU decisions	Review does not guarantee reversal
Public bias-audit information	Public compliance material for a covered New York City AEDT	An audit concerns broader outcomes, not whether one rejection was correct
Federal-agency application records	Existing government records may be requested through FOIA	FOIA does not require an agency to create a new explanation
Source code, training data, validation files, or demographic records	Generally available, if at all, through a formal investigation or litigation	An informal applicant email ordinarily cannot compel them

Ask about the particular function instead of using “AI” as a catch-all. Software might have parsed your résumé, searched for keywords, applied a knockout rule, ranked applicants, scored an assessment, analyzed an interview, or supplied a recommendation to a recruiter.

An applicant tracking system, or ATS, may simply store applications and move them through hiring stages. Its presence does not establish that an AI system made the decision. For a practical explanation of the distinction among storage, parsing, keyword searches, knockout questions, and ranking, see this guide to how résumé screening and applicant tracking systems work.

Most importantly, the evidence reviewed for this guide does not establish a universal right to an individualized explanation, numerical score, rank, criteria weighting, appeal, human review, reconsideration, reversal, source code, training data, or model details. You can request any of these, but the ability to compel an answer is much narrower.

An explanation can still be useful. It may reveal a misunderstood qualification, incorrect employment date, failed résumé parse, or knockout response that needs clarification. In a 2025 study involving 921 participants and four hypothetical personnel-selection scenarios, explanations improved several measures of perceived fairness for decisions described as being made by either AI or humans. The study measured perceptions and behavioral intentions, however—not whether actual decisions were accurate, unbiased, lawful, or likely to be reversed (research on applicant fairness perceptions).

Before writing: identify the decision, employer, place, and date

Start with jurisdiction and employer type. Do not assume that a law applies merely because the employer used software, described a product as “AI-powered,” or accepted an application from someone living in a particular city or state.

Use this decision tree.

1. Who was the employer? - A private employer - An employment or staffing agency - A federal executive-branch or independent regulatory agency - A state or local government - Another public, nonprofit, educational, or regulated entity

2. Where were the relevant people and activities located? - Where did you live when you applied? - Where was the job located? - Where was the employer or employment agency operating? - Where did the assessment occur? - Where did the screening or decision-making activity occur, if known?

3. When did each event occur? - Application submission - Assessment or interview - Automated notice, if any - Rejection - Later explanation or correspondence

4. How automated was the process? - Did software make the decision without meaningful human participation? - Did it rank or recommend candidates for human review? - Did a person merely accept the output? - Did a qualified person independently assess the application and have authority to change the result?

5. What function was involved? - Résumé parsing - Keyword search - Knockout question - Machine-learning ranking - Online assessment - Video or audio analysis - Chatbot screening - Scheduling or workflow automation - Another screening or decision-support process

Dates matter because the relevant rules do not operate on the same schedule. New York City’s Department of Consumer and Worker Protection began enforcing Local Law 144 on July 5, 2023, and states that required AEDT notice must be provided at least 10 business days before use (NYC DCWP guidance).

Colorado records SB26-189 as signed on May 14, 2026, with relevant requirements and rulemaking milestones beginning in 2027. California’s finalized regulations generally took effect on January 1, 2026, but businesses subject to the new automated-decisionmaking technology requirements must begin compliance on January 1, 2027.

The distinction between a solely automated decision and an AI-assisted decision can also be decisive. A system that rejects an applicant without meaningful human participation is different from software that organizes records, highlights keywords, or recommends a ranking that a person independently evaluates. A nominal “human in the loop” does not necessarily resolve every legal question, but neither should you assume that any use of software makes a decision solely automated.

Likewise, not every product marketed as AI is necessarily a covered automated decision-making technology under Colorado law or an automated employment decision tool under New York City law. Legal definitions may turn on the tool’s function, influence, use, and surrounding decision process—not its branding.

Preserve the record before postings, portals, notices, or policies change. Save:

• The complete job posting
• The résumé and application you submitted
• The application or requisition number
• Your responses to knockout questions
• Assessment instructions and your responses, where available
• Interview invitations and any recordings you lawfully possess
• AI, privacy, consent, or candidate notices
• Accommodation information
• Screenshots of the application portal
• The rejection message
• Correspondence with recruiters or hiring staff
• The employer’s privacy and appeal policies
• Exact dates and times

If you cannot export a page, take screenshots and record the URL and access date. Preserve original files separately from notes you create afterward.

A very fast rejection—or a résumé that appears to match the posting closely—can justify asking what happened. It does not prove that AI was used, that no person reviewed the application, or that discrimination occurred. A knockout response, incomplete application, changed hiring need, internal candidate, or ordinary recruiter judgment could also explain the outcome.

A neutral request template for the employer or recruiter

Send the request to the employer, recruiter, employment agency, candidate privacy contact, or hiring contact identified in the employer’s own materials. Do not send it to an unrelated publication, commentator, or technology company unless that company is the designated contact for the hiring process.

A concise, factual request is usually more useful than a long accusation:

Subject: Request for information and review—[position title / requisition number]

Hello [name or hiring team],

I applied for the position of [position title] on [application date] and received a rejection on [rejection date]. My application or requisition number is [number, if available].

Was an automated or AI-based system used to screen, rank, test, score, or otherwise evaluate my application, and what role did it play in the decision?

Please provide, where applicable:

• The principal reasons for the result and the general evaluation criteria used
• A description of the hiring task the system performed
• Whether a qualified person meaningfully reviewed the automated output before the rejection
• The personal data used in the evaluation
• A way to correct any factually inaccurate personal data
• Meaningful human review and reconsideration using corrected information
• Links to any relevant candidate notice, privacy notice, public bias-audit summary, accommodation process, or internal appeal policy

I understand that the availability of particular information or review may depend on applicable law and company policy. Please respond in writing and let me know if reasonable identity verification is required.

Thank you, [Name] [Contact information]

Keep the tone factual. You do not have to waive concerns or praise the process, but accusing the employer of intentional discrimination without supporting evidence may make it harder to obtain a useful operational response. Identify the application, describe what happened, ask precise questions, request a written reply, and retain copies of everything you send and receive.

If an assessment condition may have affected you, add a short paragraph:

During the assessment, [describe the accessibility barrier, technical interruption, language issue, incorrect data, or other condition]. Please identify any accommodation or alternative-assessment process available under applicable law or company policy and consider this information during review.

That is a request, not an assertion that every applicant has a universal post-rejection right to retesting or an alternative assessment.

Be clear about the type of information you want:

• A general process description explains how the employer ordinarily uses a tool.
• Individualized rejection reasons explain why your application did not advance.
• A numerical score or rank reports an output assigned to you.
• Criteria weighting describes how inputs contributed to an output.
• A public bias-audit summary reports required aggregate information about a covered tool.
• Model internals may include source code, training data, complete logic, validation materials, or proprietary technical documentation.

Receiving one does not imply access to the others. An employer might describe a screening stage without providing your score. A public audit might exist without supplying individualized reasons. A privacy response might identify categories of personal data without disclosing proprietary model details.

Source code, training data, full model logic, internal bias-testing files, aggregate demographic records, and information about other applicants generally cannot be compelled merely by sending this informal request. Such material may be confidential, may belong to a vendor, or may become relevant only during a regulatory investigation or formal legal proceeding.

Colorado: post-rejection requests for covered decisions

Date note: The Colorado General Assembly records SB26-189 as signed on May 14, 2026. The act has relevant requirements beginning in 2027, and the attorney general must clarify post-adverse-outcome disclosure requirements by rule by January 1, 2027.

Employment is included among “consequential decisions.” The legislative summary defines an automated decision-making technology, or ADMT, as technology that processes personal data and uses computation to generate output—such as a prediction, recommendation, classification, ranking, or score—used to make, guide, or assist a decision about an individual.

The important trigger is not merely that software was present. The supported rights concern a covered ADMT making a consequential decision that results in an adverse outcome. Questions about material influence, covered status, exemptions, and the employer’s actual use of the technology therefore matter.

For a covered adverse employment outcome, the summary supports requests for:

• The personal data used by the covered ADMT
• Correction of factually incorrect personal data used by it
• Meaningful human review
• Reconsideration of the adverse decision

Separately, the deployer must provide a plain-language description of the covered ADMT’s role within 30 days after the adverse outcome. Developers and deployers must retain necessary compliance records for at least three years, but that retention obligation does not automatically give an applicant access to every retained record.

Enforcement runs through the Colorado Consumer Protection Act, and the measure does not create a new private right of action. The signed act and final rules—not a shorthand description—control questions about coverage, procedure, exemptions, enforcement, and remedies (Colorado General Assembly summary and bill materials).

A Colorado-specific request could read:

I am requesting the information and review available for an adverse employment outcome involving a covered automated decision-making technology, if SB26-189 applies to this decision.

Please:

1. Provide the personal data used by the covered ADMT;
2. Allow me to identify and correct factually incorrect personal data used by it;
3. Arrange meaningful human review of the adverse outcome;
4. Reconsider the outcome using accurate information; and
5. Provide the required plain-language description of the covered ADMT’s role.

Please also identify any procedure or reasonable identity-verification step required to process these requests.

This wording avoids inventing an address, verification method, applicant-request deadline, or remedy not established by the supplied summary. Use the contact route in the employer’s candidate notice, privacy notice, or hiring correspondence.

Coverage should not be assumed. An ATS used only for storage or workflow may not qualify. A keyword search, knockout rule, ranking feature, assessment, or decision-support tool must be evaluated against the controlling definition, its material influence on the consequential decision, covered-entity rules, and applicable exemptions.

Before asserting a Colorado right, verify:

• The law was operative on the decision date
• The employer or deployer was covered
• The technology and its use met the covered-ADMT definition
• The tool had the required relationship to a consequential decision
• The decision resulted in a covered adverse outcome
• No exemption applied
• The final rules’ procedures were followed
• The correct enforcement route was used

New York City: seek the audit information and check the advance notice

New York City Administrative Code §§ 20-870 through 20-874 form the subchapter addressing automated employment decision tools, including definitions, requirements, penalties, enforcement, and construction. The cited code structure does not itself establish a post-rejection right to an individualized explanation or appeal (New York City Administrative Code, Subchapter 25).

For a covered AEDT, Local Law 144 focuses on audit, publication, and notice obligations. A covered tool cannot be used unless it received a bias audit within the preceding year, information about that audit is publicly available, and required notices were provided.

After a rejection, ask:

Did the employer or employment agency use an automated employment decision tool covered by New York City Local Law 144?

If so, what tool was used, and where can I find the publicly available summary of the most recent bias-audit results?

Please also identify when and how the required advance notice was provided to me.

Check your email, application portal, assessment invitation, and privacy notices. Do not assume that a generic privacy statement necessarily satisfied the controlling requirements.

It does not determine whether your résumé was parsed correctly, whether your score was accurate, whether a recruiter exercised appropriate judgment, or whether your individual rejection was unbiased, fair, or lawful.

If a covered employer or employment agency appears to have failed to obtain the required audit, publish its summary, or provide required notices, you may report the issue to DCWP. The department permits an online complaint without creating an account, although an account can make tracking and updating the complaint easier.

The supplied New York City materials do not establish an individual right to:

• An AI score or ranking
• Criteria weights
• A detailed individualized explanation
• Human appeal
• Reconsideration
• Reversal of the rejection

You may still request those things voluntarily. Keep that request distinct from a complaint about a missing audit, public summary, or notice.

Do not extend Local Law 144 automatically to every employer, tool, candidate, or hiring decision with some connection to New York. Application depends on the law’s scope and the facts surrounding the tool’s use.

Federal government hiring: request existing records through FOIA

If you applied to a federal executive-branch agency or independent regulatory agency, the Freedom of Information Act may provide a route to existing agency records about your application.

FOIA does not apply to private employers, state or local governments, Congress, or federal courts. A private company does not become subject to FOIA merely because it contracts with the government or uses regulated technology.

Direct the request to the federal agency or component that maintains the records. Include enough information for staff to conduct a reasonable search:

• Your full name
• Position title
• Vacancy announcement number
• Application date
• Relevant date range
• Office or component, if known
• Types of records requested
• Contact information
• Any required identity certification

A bounded request might say:

Under the Freedom of Information Act, I request existing agency records concerning my application for [position title], vacancy number [number], submitted on [date].

The requested date range is [start date] through [end date]. Please search the component or office that maintains records for this hiring action.

I request existing records associated with my application, assessment, referral or eligibility status, processing status, and hiring decision, including any recorded scores, status codes, notices, or decision records associated with me.

This is a first-party request for records about myself. Please tell me what identity certification or verification is required. If any material is withheld, please identify the applicable basis and release reasonably segregable nonexempt portions.

This wording does not assume that a score, report, or other particular record exists. It asks for records the agency already maintains.

FOIA is a records-access law. It does not require an agency to:

• Create a new explanation
• Answer questions that are not requests for records
• Perform a new analysis
• Generate a score that was never recorded
• Explain source code or model logic
• Reconsider the hiring decision
• Conduct an employment appeal

A first-party requester may be required to certify or verify identity. Even when records exist, disclosure can be partial. FOIA contains nine statutory exemptions, including protections for personal privacy, confidential commercial information, privileged communications, national security, and law-enforcement interests. Agencies should take reasonable steps to release segregable, nonexempt portions when full disclosure is unavailable (FOIA.gov guidance on coverage, exemptions, first-party requests, and appeals).

If you are dissatisfied with the initial determination, you may file an administrative FOIA appeal through the process stated in the agency’s response. Keep that appeal separate from any merit-system procedure, discrimination complaint, accommodation matter, or request to reconsider the hiring result. Obtaining records—or winning a records appeal—is not the same as overturning a rejection.

EU and California: important protections, but different timing and scope

For the European Union, the central question is whether the rejection was based solely on automated processing and significantly affected the candidate. If those and other coverage conditions are satisfied, GDPR protections concerning significantly consequential solely automated decisions may apply.

Useful questions include:

• Did AI or automated processing assist the decision?
• What task did it perform?
• What data and general factors informed the result?
• How was the decision made?
• At what point did a person become involved?
• Did that person independently evaluate the case?
• Did the reviewer have authority to change the outcome?
• What process is available to contest the result or request human intervention?

Human intervention or review may be available where GDPR Article 22 applies. This is not a universal right for every use of AI in European hiring. Meaningful human participation, the precise processing activity, legal basis, exceptions, safeguards, and other facts can affect coverage.

The supplied support for this section is secondary law-firm analysis rather than the GDPR text or regulator guidance. Treat it as general orientation and verify the current controlling law and official guidance before asserting a right (analysis of GDPR Article 22 and AI-assisted recruiting).

An EU request should not overreach. Article 22 should not be summarized as a universal entitlement to source code, the complete model, full training data, every scoring detail, or human review for all AI-assisted hiring decisions.

California is on a different timetable and should not be treated as though it adopted Colorado’s specific post-rejection framework. The California Privacy Protection Agency announced that regulations addressing ADMT used for significant decisions were finalized and generally took effect on January 1, 2026. Businesses subject to the ADMT requirements must begin compliance on January 1, 2027 (California Privacy Protection Agency announcement).

That announcement does not, by itself, establish which post-rejection information, correction, explanation, appeal, reconsideration, or human-review requests a particular applicant may make. Before asserting a California right, review the final regulatory text and verify:

• Whether the business is covered
• Whether the applicant and processing activity are within scope
• Whether the technology qualifies
• Whether a significant decision is involved
• The decision date and applicable compliance date
• The required procedure and available exceptions

Do not copy Colorado’s list of post-adverse-outcome rights into a California letter and present it as California law unless California’s controlling text independently supports those requests.

If the employer does not answer—or discrimination is suspected

Silence after a voluntary email is frustrating, but it does not itself prove that the employer broke the law, used AI, or discriminated. The next step depends on the issue.

Issue	Possible next route
Missing NYC AEDT audit, published summary, or notice	DCWP complaint
Inadequate federal FOIA determination	Administrative FOIA appeal
Covered privacy or automated-decision request	Procedure and regulator designated by the applicable law
Colorado compliance concern	Enforcement structure under the operative law and final rules
Suspected employment discrimination	Appropriate discrimination agency or qualified employment lawyer
Accommodation or inaccessible assessment	Employer accommodation channel and any applicable administrative or legal process
Internal error or policy dispute	Recruiter, candidate privacy contact, or internal appeal procedure

Technology supplied by a vendor does not necessarily remove an employer’s responsibility under applicable employment-discrimination law. Relevant U.S. protected-basis concerns can include race, color, religion, sex, national origin, age for people 40 or older, and disability. An unfavorable automated outcome is not automatically discriminatory; evidence must support a connection to an unlawful protected-basis distinction or another covered violation.

The supplied support for these discrimination propositions includes a law-firm article rather than primary agency guidance, so it should be treated as secondary orientation. Before filing, verify the governing requirements with the responsible agency or qualified counsel (secondary employment-law discussion of AI-assisted rejection claims).

Document facts that could matter:

• Incorrect dates, credentials, employment history, or identity data
• A résumé parse that misplaced or omitted qualifications
• Assessment interruptions or inaccessible conditions
• A requested accommodation and the response
• Messages describing how the tool was used
• Inconsistent reasons for the rejection
• Comparable applicants who appear to have been treated differently
• Repeated outcomes suggesting a possible pattern affecting a protected group
• Missing notices or conflicting privacy statements

More technical material may become relevant during a formal investigation or litigation. Depending on the claim, jurisdiction, and procedural rulings, a party might seek system documentation, source code, training data, validation or bias-testing materials, aggregate hiring patterns, or information about other applicants.

Those materials are not ordinarily compelled by a simple candidate email, and access is never automatic. Discovery generally follows the start of a formal proceeding, may face relevance objections or confidentiality protections, and is not guaranteed.

Do not wait indefinitely for an informal response if you may have a time-sensitive discrimination or administrative claim. Filing deadlines can be short and jurisdiction-specific.

No request strategy guarantees a viable claim, damages, class relief, discovery, reconsideration, an invitation to reapply, or a reversed decision. This guide is informational rather than legal, HR, or employment advice. Hiring laws vary by location and change over time, so confirm current requirements with the responsible agency or qualified counsel.

Frequently asked questions

Can I ask whether AI was used to reject my application?

Yes. Ask whether an automated or AI-based system screened, ranked, tested, scored, or otherwise evaluated your application and what role it played.

You can also ask whether the system made the decision, recommended an outcome, prioritized candidates for review, or supplied information that a person independently evaluated. Ask where meaningful human involvement occurred and whether the person had authority to change the result.

The ability to ask does not create a universal duty to answer. A notice requirement, privacy law, automated-decision rule, employer policy, or other applicable obligation may strengthen your position in a particular case.

Am I entitled to my AI score, ranking, or a detailed explanation?

Not universally. You may request your score, rank, individualized reasons, general criteria, and criteria weighting, but the supplied authorities do not establish that every employer must disclose those items after every rejection.

Keep the categories separate. A general description of the process is not your individual score. A public bias-audit summary is not an explanation of your rejection. An existing federal record may be obtainable under FOIA, but FOIA does not require an agency to create a score or explanation that was never recorded.

Can I demand human review or reconsideration?

You can request both, but whether you can compel them depends on the applicable law and facts.

For a covered Colorado consequential decision resulting in an adverse outcome, the enacted-law summary supports requests for meaningful human review and reconsideration. It also supports requests for the personal data used and correction of factually incorrect data (Colorado SB26-189 materials).

In the EU, human intervention may be available when protections for a significantly consequential solely automated decision apply. Meaningful human participation and other coverage conditions can affect that conclusion.

Ask for review by a qualified person who considers the relevant information independently and has authority to change the result. Even where review is available, reversal is not guaranteed.

What can I do if a New York City employer did not provide an AEDT notice or publish a bias-audit summary?

First, confirm that the employer or employment agency used a tool covered by Local Law 144. Ask which AEDT was used, where the public bias-audit summary is posted, and when and how the advance notice was provided.

If there appears to have been a failure to obtain the required audit, publish its summary, or provide the required notices, you may submit a complaint to NYC DCWP. An online complaint can be filed without creating an account, although an account can make tracking and updating the complaint easier (NYC DCWP complaint guidance).

That route concerns Local Law 144 compliance. It does not itself establish a right to your score, a detailed explanation, human appeal, reconsideration, or reversal.

Can I use FOIA to obtain hiring records from a private employer?

No. FOIA applies to covered federal agencies, not private employers. It also does not apply to state or local governments, Congress, or federal courts, although other public-records laws may govern some of those entities (FOIA.gov explanation of institutional coverage).

If you applied to a federal agency, you may request existing records from the agency component that maintains them. If you applied to a private employer, look instead to the employer’s privacy process, an applicable state or local law, an internal appeal policy, or a formal investigation or legal process where appropriate.

The practical sequence is straightforward: preserve the record; identify the employer type and relevant locations; determine the decision date and likely automated function; send a neutral written request; then use the jurisdiction-appropriate route if necessary.

Asking broadly is often worthwhile, but the right to compel an answer remains narrow and fact-specific. If discrimination or a compliance deadline may be involved, verify current requirements promptly with the responsible agency or qualified counsel.