When a job seeker worries that AI screened a resume, separate what the application actually shows from what remains unknown.
Some employers use software, including AI, in hiring. The U.S. EEOC’s worker guidance discusses disability-related risks and accommodations; it does not reveal whether a particular employer used a tool or why an applicant was rejected. A portal parser, an eligibility rule, a ranking feature, and a human review are different possibilities. Start with observable evidence instead of trying to optimize for an unseen model.
What you can check
- Read the vacancy and note the required qualifications, location, authorization questions, file type, and deadline.
- Keep the exact resume and answers submitted. Compare them with any fields or preview shown by the portal.
- Record a specific mismatch, error message, or inaccessible step. Do not label a standard rejection as an AI decision unless the employer confirms the tool was involved.
- If the issue concerns an accommodation or a disability-related assessment in the United States, follow the employer’s published accommodation channel and consult EEOC guidance for scope. For other jurisdictions, check the relevant local agency.
- Continue the search while waiting. A recruiter may not be able to share an internal score or individualized decision details.
Worked example: fix the record, not an imagined model
- Task: Submit an application for a customer-operations position and verify the resume and portal record.
- First output/attempt: After reading online claims about AI screening, the candidate rewrites the resume with repeated job-ad terms and removes a clear example of handling billing escalations.
- Mistake or weakness: The candidate is optimizing against an unknown process, making the resume less specific while learning nothing about how the employer evaluates it.
- Human correction: They restore the truthful billing example, check the posting’s actual required tasks, complete each portal field, and inspect the visible imported work history. A date is missing in the preview; they correct that field from their own record.
- Quality check: Confirm the resume and application form agree on dates, titles, contact information, and qualifications. Save the notice and exact submitted file. In this illustrative log, three fields are verified and one visible mismatch is corrected; these counts are not a finding about AI or hiring outcomes.
- Final usable result: The candidate has a role-specific resume, a matching portal record, and a note of the one corrected field. If there is no visible mismatch, the final record says “no observable portal error; screening method unknown” rather than assigning a cause.
A neutral follow-up for a real error
Use the employer’s stated contact route and ask about an observable problem, not a hidden model. Example: “I submitted an application for [role] on [date]. The portal preview shows [field] as [incorrect value], while my submitted resume lists [correct value]. Is there a way to update the application record?” Replace the brackets with accurate details and include only the information needed to identify the application.
If the application is closed, ask whether the record can still be noted or corrected; do not create a second account or submit another application unless the employer directs you to. If there is no contact route or no response, save the original record and follow the employer’s published timeline. Do not invent a response deadline or assume silence confirms an automated rejection.
Keep facts separate from hypotheses
| Observed fact | Possible explanation | Next check |
|---|---|---|
| Portal preview omitted a date | Parser or import issue | Compare source file; correct visible field or contact support. |
| Standard rejection message | Could involve eligibility, timing, screening, or human review | Record it as unexplained unless the employer identifies the reason. |
| Assessment blocks an accommodation | Potential accessibility or process issue | Document the step and use the published accommodation process; check local law. |
EEOC guidance is limited to U.S. federal disability protections and the specific issues it addresses. It does not make every automated decision unlawful or establish that a particular application was screened out. Individual circumstances and state, local, or other-country rules can differ.
Beginner exercise: create an application evidence note
Choose one submitted application. Record the role, date, confirmation number, file name, exact notice wording, and any portal error. Label each item “observed” or “not known.” If a field mismatch exists, identify its source of truth and write one neutral question. The saved output is a one-page note with the record and next step; do not include sensitive personal information unless a legitimate accommodation channel requires it.
Pass check: Your evidence note must have two separate columns: what you observed and what remains unknown. Include the saved file, job ID, portal fields or notice, and date of any follow-up. Pass when no statement says an AI tool was used unless the employer or a reliable record confirms it.
Frequently asked questions
Can I tell whether AI rejected my resume?
Usually not from a standard rejection message. It may not show which tools were used, what stage made the decision, or what the reason was.
Should I buy an AI resume score to get past the screener?
A third-party score is not the employer’s score unless the provider can demonstrate that it uses the same system and criteria—which a job seeker should not assume. Use suggestions only when they improve accuracy or readability.
What if I need an accommodation for an assessment?
Use the employer’s published request channel and describe the adjustment needed for the assessment. In the United States, EEOC material discusses ADA-related accommodation issues involving software and AI; elsewhere, check the local equality or employment agency.
A useful next step
Use this to inspect the document and portal fields you can see. the ATS-friendly file guide. Use this to choose a cautious next step after an unexplained rejection. the suspected-automated-screening guide.
A visible parsing defect is not proof of an AI rejection
Suppose the employer portal shows a blank end date after importing a resume. That is an observable record problem: compare the attachment with the portal field and correct the date if the system allows it. It does not establish that an AI model evaluated the candidate, that the blank field changed a score, or that the application was rejected for this reason. The same visible symptom could come from ordinary parsing, a manual data entry problem, or a portal display issue.
Record the file version, the field shown, the date, and the correction. If the employer offers a contact channel, ask a neutral question about correcting that field. If you have a disability-related concern about an assessment, EEOC guidance discusses possible reasonable accommodations under the ADA; it does not identify the software used in an individual decision or guarantee reconsideration. Keep those questions separate from ordinary file cleanup.
Keep the evidence note private and proportionate
Your application log needs only enough information to identify what happened: employer and role, submission date, confirmation, file name, exact error or notice, and the step you took. Do not store medical details in a general job-search spreadsheet. If an accommodation request is necessary, use the employer’s published channel and keep that correspondence separately with appropriate access protection. The EEOC guide is U.S. ADA worker guidance and is not a substitute for local legal advice.
Be skeptical of online claims that name a particular “AI filter” without documentation. An employer might use software for recordkeeping, extraction, assessment, ranking, or none of these at the stage you encountered. Product features and employer settings differ. Unless you can see a field error or the employer confirms the process, write “screening method not known” in your notes. This keeps a future conversation factual and prevents the search from becoming an endless attempt to reverse-engineer a system you cannot inspect.
If an employer does confirm a tool was involved, ask about the part relevant to your issue rather than demanding a model score that may not be available. For an accessibility problem, describe the step you could not use and the adjustment that would let you complete it. For a missing field, provide the correct value and its source. In either case, use the channel the employer specifies and save the response.
Sources
- U.S. EEOC: Tips for Workers—The ADA and AI — Describes examples of employer technology used to score applications or tests and points workers to disability-rights and accommodation resources; U.S. ADA context, not evidence about an individual rejection. Checked 2026-10-17.
- Greenhouse Support: Unsuccessful resume parse — Documents a parser’s extraction behavior and possible formatting errors in Greenhouse specifically. Checked 2026-10-17.