AI is screening your application: how to get read by a human in 2026

If you have applied for a hospital, county or federal health job recently, there is a good chance software handled your application before any person did. It may have split your résumé into fields, asked you knockout questions, ranked you against other applicants or scored a recorded video interview. And in late August 2026 the federal government’s own personnel office told agencies to start using AI across the hiring process too.

This guide is for nurses, allied health staff and anyone else applying into public-sector health care who wants to know what these tools actually do, what the law now lets you ask for, and how to make sure a human being reads your application. It separates what is documented from what is folklore, because a lot of the advice on this subject is the second kind.

What does “AI screening” actually mean?

The phrase covers several different jobs that software does, and they fail in different ways. Knowing which one you are dealing with tells you what to fix.

• Parsing. An applicant tracking system reads your uploaded résumé and copies it into fields: employer, job title, dates, license. This is not judgment, but it is where a lot of applications quietly go wrong. A two-column layout, a table, or contact details typed into a page header can come out scrambled or blank.
• Knockout questions. Yes-or-no questions the employer sets: do you hold an active RN license in this state, can you work nights, do you have two years of acute care experience. Answer no to a required one and the system may remove you automatically. A person chose that rule; the software applies it.
• Matching and ranking. Some systems compare your record with the job’s requirements and sort or score applicants, so a recruiter sees the strongest apparent matches first. If your experience is described in different words from the posting, you can sink down a list that nobody reads to the end.
• Assessments and recorded interviews. Online tests, and one-way video interviews where you answer questions on camera with nobody on the other end. Some tools analyze these automatically; others simply queue them for a recruiter to watch.
• Summaries and chat assistants. Newer tools summarize an application for the reviewer, answer candidates’ questions, or schedule interviews.

The practical point is that most of what you can control sits in two places: whether the software can read what you wrote, and whether you have answered the employer’s screening rules accurately. Both are fixable, and the rest of this guide shows how.

Does USAJOBS scan your résumé for keywords?

Not in the way private-sector advice assumes, and the picture is changing. USAJOBS describes a process in which the hiring agency starts reviewing applications after the announcement closes, checks each applicant’s eligibility and qualifications, and sends the highest qualified applications to the hiring official, who decides whom to interview. Nothing in that account involves a vendor ranking your résumé on keyword density. Your application is rated against the qualification requirements and assessments stated in the announcement.

Two things have shifted underneath that description. First, under the Merit Hiring Plan, OPM told agencies to stop using self-assessments for rating and ranking by September 30, 2025, except for positions at GS-4 and below, seasonal positions and some Federal Wage System jobs. The occupational questionnaire where you once rated yourself “expert” on thirty tasks can still screen you on minimum qualifications and eligibilities such as veterans’ preference, but it is no longer supposed to decide your place in line. Agencies must now use at least one technical assessment before they issue a certificate of eligibles.

Second, on August 27, 2026, OPM Director Scott Kupor issued a memorandum titled “Use of Artificial Intelligence in the Federal Hiring Process.” It says agencies may be compromising efficiency and quality by failing to adopt AI, and advises them to use it across the hiring lifecycle, in support of the 80-day time-to-hire goal. The uses it works through include screening résumés against minimum qualifications, reviewing documents for veterans’ preference and other eligibilities, and rating, ranking or scoring candidates.

The memo does not hand the decision to software. It treats résumé screening and AI scoring as “presumed high-impact” under OMB memorandum M-25-21, and says such a use escapes that label only when an authorized official reviews the underlying documents and makes the determination on that record rather than on the AI’s summary. A token sign-off is not enough: where officials ratify AI outputs without independent review of the underlying record, OPM says the use must be treated as high-impact, with the minimum risk-management practices that brings.

What that means for you: in federal hiring, software is increasingly likely to locate, extract and summarize your evidence for a human reviewer, and OPM says more AI features are coming to USA Staffing, USAJOBS and USA Hire. The best response is to make that evidence easy to find. Month-and-year dates, hours per week, grades, license details and the specialized experience the announcement asks for should all be stated plainly inside your two pages.

One caveat for clinicians. Many VA clinical posts are appointed under title 38 rather than title 5, and OPM’s Merit Hiring Plan FAQ says systems outside title 5 should align with the plan to the greatest extent practicable rather than follow every element of it. Read the specific announcement to see how you will be assessed.

What hospitals, counties and state employers are likely to use

Outside federal service there is no single system. A county health department may run its own civil service portal, a public hospital system usually buys a commercial applicant tracking system, and a state agency may do either. The tools described above turn up in all of them, in different combinations, and it is the employer that decides which features are switched on.

Health care adds its own twist. Many knockout questions are about licenses, certifications and shifts, which are genuine legal or operational requirements. That is where an automated rejection is most likely to be applying a rule someone set deliberately, which makes reading each question carefully and answering it accurately the single most useful thing you can do.

How do you get your application read by a human? A checklist

• Use a plain, single-column résumé. Put your name and contact details in the body of the page, not in a header or footer, and use ordinary section headings such as Experience, Education, and Licenses and certifications.
• Submit the file type the employer asks for. A text-based PDF or Word file can be parsed; a scanned photo of a printed résumé usually cannot.
• Write licenses and certifications the way the posting writes them, with the full name once and the abbreviation after it, for example Basic Life Support (BLS), plus the issuing state and expiry date.
• Give month and year for every role, and hours per week for federal applications.
• Describe your experience in the posting’s own terms where they are true. If it asks for telemetry and you worked a step-down unit with continuous cardiac monitoring, say both.
• Complete every field in the online form, even when it repeats your résumé. Some systems search or sort on the form fields rather than the attachment.
• Answer knockout questions exactly. If one is ambiguous, such as two years of acute care when you have 22 months plus clinical rotations, answer truthfully and use any text box to explain.
• Check the parse. Many systems show the fields they filled in from your upload; if job titles or dates come out wrong there, correct them before you submit.
• Find a person as well. A recruiter’s name on the posting, a hiring event or a former colleague on the unit are all legitimate routes. A short note saying you have applied, with the requisition number, costs nothing.
• For federal jobs, use the agency contact listed on the announcement for questions about how the job will be assessed. That contact cannot move you up a list, but can tell you what the assessment is.

Should you use AI to write your application?

For federal jobs, be careful. The Merit Hiring Plan’s essay questions come with a certification that you used your own words and did not use a consultant or AI, and that rule is set out in our guide to the essays. One of the four questions, asking applicants to name an executive order, is on hold after a federal judge in Massachusetts stayed it on September 11, 2026, Government Executive reported. Everything on the résumé itself must be accurate too, because it will be checked against your records and references.

For other employers, using a tool to tidy your wording is common and legal. Two uses cause real harm. One is letting it invent: a generated résumé that claims a certification you do not hold, or a number you cannot support, will fall apart at credentialing, which in health care is thorough. The other is hidden text, such as pasting the job description in white or tiny type. Hidden text is still text. Whatever a parser extracts can be shown to the recruiter reviewing your record, and there it simply looks like an attempt to game the system.

A better use of an AI assistant is as a checker. Give it the posting and your résumé and ask which stated requirements your résumé does not clearly show. Then close those gaps yourself, in your own words, with facts only you can supply. Before you paste anything into a consumer chatbot, strip out your Social Security number, license numbers, date of birth and home address.

What rights do you have when AI screens you?

As of September 2026 there is no federal law written specifically for AI in private-sector or local-government hiring. What applies is the law that already covers every hiring decision, however it is made: Title VII of the Civil Rights Act, the Americans with Disabilities Act and the Age Discrimination in Employment Act. On top of those sit a handful of state and city rules, which have moved a great deal in the past year.

At the federal level, the Equal Employment Opportunity Commission had published technical assistance on AI under Title VII in 2023 and under the ADA in 2022, and both were taken off its website in early 2025, as the law firm Cooley reported at the time. In April 2025, Executive Order 14281 directed federal agencies to deprioritize enforcement of disparate-impact liability, which is the legal theory most AI-bias claims rely on, since a tool rarely discriminates on purpose. The statutes themselves did not change, and private lawsuits continue. In May 2025 a federal court in California granted preliminary certification of a nationwide age-discrimination collective in Mobley v. Workday, over allegations that the company’s AI recommendation system disadvantaged applicants over 40. Those are allegations, not findings.

New York City’s Local Law 144 applies when a job is located in the city, at least part time, and the employer uses an automated employment decision tool to substantially assist screening or assessment. The tool must have had an independent bias audit within the past year, a summary of the results must be posted, and candidates who live in the city must be told at least 10 business days before it is used, including what qualifications or characteristics it will assess and how to request a reasonable accommodation. Complaints go to the Department of Consumer and Worker Protection through 311; discrimination claims go to the city’s Commission on Human Rights. Enforcement has been thin: a December 2025 audit by the New York State Comptroller found the department received only two complaints about these tools over two years, identified one compliance issue where the auditors found at least 17 in the same set of companies, and called its complaint process ineffective. Penalties run from $500 to $1,500 per day.

Illinois has two relevant laws. Under the Artificial Intelligence Video Interview Act, an employer that uses AI to analyze recorded video interviews must tell you beforehand, explain how the AI works and the general types of characteristics it uses to evaluate applicants, and get your consent; it may not use AI to evaluate you without that consent. It may share your video only with people whose expertise or technology is needed to evaluate it, and must delete it within 30 days if you ask. Separately, since January 1, 2026, Public Act 103-0804 has made it a civil-rights violation under the Illinois Human Rights Act for an employer to use AI that has the effect of discriminating on protected characteristics, to use zip codes as a proxy for them, or to fail to give employees notice that it uses AI for employment decisions. The Illinois Department of Human Rights is responsible for the rules on when and how that notice is given.

Colorado’s 2024 AI Act, SB 24-205, was pushed back to June 30, 2026 by a 2025 special-session bill and then never took effect. According to a client alert from the law firm McDermott, under a stipulated order a federal magistrate judge entered on April 27, 2026, Colorado agreed not to enforce it while a challenge brought by xAI, which the Justice Department joined, was pending. On May 14, 2026, the governor signed SB 26-189, which repeals and replaces it from January 1, 2027. For consequential decisions, including employment, the new law calls for notice when automated decision-making technology is used, a plain-language description of its role within 30 days of an adverse outcome, the ability to obtain and correct the personal data it used, and the right to request meaningful human review and reconsideration.

California’s Civil Rights Council adopted regulations, in effect since October 1, 2025, that make clear an automated-decision system can violate state anti-discrimination law if it harms applicants on the basis of protected characteristics. They require employers to keep automated-decision data for at least four years, and warn that assessments, including tests, questions or puzzle games that draw out information about a disability, may be an unlawful medical inquiry.

Expect this map to keep moving. Executive Order 14365, signed December 11, 2025, directed the Attorney General to set up an AI Litigation Task Force to challenge state AI laws the administration considers inconsistent with national policy, and named Colorado’s algorithmic discrimination law as an example. The order does not itself cancel any state law.

What to do if you think software rejected you unfairly

• Write down what happened while it is fresh: the job, the requisition or announcement number, when you applied, when the rejection arrived, and any notice you saw about automated tools. A rejection within minutes is not proof of anything, but it is worth recording.
• Reread the knockout questions if you can still see them. A wrong answer to a license or shift question is a fixable cause, and some employers let you withdraw and reapply.
• Ask. Email the recruiter or HR contact with the requisition number and ask whether an automated tool was used and whether a person reviewed your application. In New York City, ask where the required bias audit summary is posted.
• For federal jobs, contact the HR office named on the announcement if you believe you were wrongly found ineligible or unqualified, and ask how to request reconsideration. If veterans’ preference was not applied, the redress route is covered in our veterans’ preference guide.
• If you believe a tool discriminated, the clock matters. A charge with the EEOC generally has to be filed within 180 calendar days, extended to 300 days where a state or local agency enforces a similar law. For age discrimination, only a state law, not a local one, extends the deadline to 300 days. Federal applicants must contact the agency’s EEO counselor within 45 days. State agencies such as the Illinois Department of Human Rights and California’s Civil Rights Department take complaints as well.
• Keep applying while you wait. Formal routes are slow; the next application is not.

Does USAJOBS reject résumés that lack keywords?

USAJOBS’s own account of the process describes no keyword score. Applications are reviewed for eligibility and qualifications against the announcement. What is new is that OPM has told agencies they may use AI to locate and summarize your evidence for the human reviewer, so clear language that matches the announcement still helps.

Is it legal for an employer to reject me using AI?

Generally yes, provided the outcome is not discriminatory and, where a local law requires it, the employer audited the tool and gave notice. The law regulates how a tool is used and what it does, not whether it is used.

Do I have to agree to an AI-scored video interview?

In Illinois the employer needs your consent before AI analyzes the video. Elsewhere you can decline, but the employer may then decline to consider you. Asking whether a person will watch the recording, or asking for an adjustment you need, is reasonable anywhere.

Should I copy keywords from the posting into my résumé?

Use the posting’s words where they truthfully describe your experience, once, in context. Repeating them or hiding them does not help a human reader, and a human reader is the goal.

Last reviewed Sep 13, 2026. First published Sep 21, 2026. Federal hiring rules change — confirm anything you act on against the agency's own announcement or OPM guidance.