A polished résumé has never been proof that someone can do the job. AI has simply made that weakness harder to ignore. Candidates can now rewrite, tailor, and submit applications faster than most hiring teams can review them. The answer is not to guess who used AI. The answer is to build a hiring process that verifies the evidence.
The résumé signal is getting weaker
Greenhouse analyzed more than 640 million applications across more than 6,000 companies for its 2026 hiring benchmarks. It reported that application volume increased 111 percent between 2022 and 2025. More applications can expand the pool. They also create more noise when candidates can tailor the same résumé to dozens of openings in a day.
A Robert Half survey published March 10, 2026 found that 67 percent of U.S. HR leaders said reviewing AI-generated applications had slowed hiring. Twenty percent reported delays longer than two weeks. Sixty-five percent of hiring managers said the application surge made candidate skills harder to verify.
The issue became more serious this summer. Researchers from Duke University, the University of North Carolina at Chapel Hill, Arizona State University, UC Berkeley, and hireEZ examined 196,682 real-world résumés. They found hidden prompt injection in about 1 percent. Some résumés contained text that a person could not see but an automated screening system could read. More than 90 percent of the injected content involved hidden data, such as skills or experience, rather than direct instructions to the system.
One percent does not mean most applicants are dishonest. It does mean employers should stop treating an automated résumé score as verified evidence. A candidate document is untrusted input until the important claims are confirmed.
Hiring demand makes weak evidence expensive
The U.S. Bureau of Labor Statistics reported 7.3 million job openings in July 2026. Openings in durable goods manufacturing increased by 76,000. In August, manufacturing employment increased by 16,000, including gains in machinery and fabricated metal products.
Those numbers do not prove that every employer faces a talent shortage. They do show that important hiring continues while application review is becoming more complicated. For a plant, construction operation, health care organization, or service business, a delayed or poorly supported decision can affect overtime, customer commitments, safety exposure, manager time, and early turnover.
More résumés are not the same as more qualified candidates. The hiring team needs a faster way to separate presentation from evidence without turning one score into a pass-or-fail decision.
What a résumé can and cannot tell you
Career claims
A résumé can organize work history, education, credentials, stated accomplishments, and relevant keywords. Each important claim still needs appropriate verification.
Whole-person job-fit evidence
PXT Select measures thinking style, behavioral traits, and interests in relation to a position-specific Performance Model. It does not verify the résumé or test technical skill.
Specific examples and judgment
A structured interview asks every finalist relevant questions and probes how the person handled real situations, decisions, pressure, conflict, learning, and results.
Demonstrated capability
A job-relevant work sample, technical demonstration, credential check, or reference can verify requirements that neither the résumé nor a job-fit assessment can prove.
A six-step response to AI-polished applications
1. Define success before reviewing candidates
Start with the work. Identify the results, decisions, pace, complexity, people demands, problem-solving requirements, and interests that matter in the role. If the hiring team cannot agree on success, it will not agree on the best candidate.
A PXT Select Performance Model creates a role-specific reference point. The model may come from job analysis, research-backed models, top-performer data, or a carefully reviewed combination. The job remains the standard.
2. Separate minimum requirements from preferred signals
Decide which requirements are essential and which are merely convenient screening shortcuts. Confirm licenses, certifications, schedule, location, legally permissible work status, and required experience at a consistent stage. Do not let an AI-friendly keyword become a substitute for evidence.
3. Use PXT Select for the question it can answer
PXT Select brings three areas into the decision: thinking style, behavioral traits, and interests. Results are compared with what the position requires. The report can surface areas of alignment, possible gaps, and focused interview questions.
PXT Select provides clear, actionable results showing how closely a candidate matches the position’s Performance Model across thinking style, behavioral traits, and interests. The higher the fit score, the greater the alignment with the role and the likelihood of success. The report gives hiring managers specific interview questions to explore potential gaps and make a better-informed hiring decision.
4. Structure the interview around evidence
Ask every candidate the same core job-related questions. Then use the assessment report to probe areas that need clarification. Ask for a specific situation, the person's action, the result, and what they would do differently. Vague answers should lead to follow-up questions, not assumptions.
5. Verify the requirements the assessment does not measure
If the role requires welding, financial analysis, equipment operation, software proficiency, writing, or another technical capability, verify it directly. Use a lawful, job-relevant work sample or demonstration. Confirm material credentials and experience through the organization's approved process.
6. Set a clear AI-use policy and apply it consistently
Employers should tell candidates what AI assistance is permitted in the application and interview process. Editing truthful content is different from fabricating experience, receiving hidden help during a skills test, or attempting to manipulate a screening system. Apply the same rule to comparable candidates and document exceptions.
Put simple economics around the problem
Consider an employer filling 20 positions, with 150 applications per opening. At three minutes per résumé, initial review consumes 150 hours. At an illustrative loaded labor cost of $50 per hour, that is $7,500 before interviews, verification, or hiring-manager time.
If a consistent qualification process limits deeper 15-minute review to eight credible candidates per opening, that stage requires 40 hours, or $2,000 at the same rate. The illustrative difference is $5,500 in review capacity. That is not a promised saving. It shows why the order of the process matters. The employer still must review candidates fairly, provide required accommodations, and retain enough human oversight to catch context an automated screen may miss.
Compliance does not disappear when the applicant pool grows
The Equal Employment Opportunity Commission states that employment tests and selection procedures should be job-related and appropriate for their purpose. Employers remain responsible for how a vendor's tool is used. They should administer procedures consistently, consider less discriminatory alternatives when required, keep the process current as jobs change, and understand the tool's limitations.
The same discipline applies to résumé screening. An AI score, keyword rank, assessment result, interview rating, or work sample is one piece of evidence. None should become an unexplained shortcut that nobody can defend.
Start with one role where the cost is visible
Do not redesign every hiring process at once. Choose a role with high application volume, recurring vacancies, inconsistent interviews, early turnover, or a costly learning curve. Define the job. Map each selection step to the question it answers. Train the people using the evidence. Then measure time, quality, retention, and manager confidence.
The Assessment Company® can help you build a position-specific hiring process using Occupational DNA®, PXT Select, structured interview guidance, and practical implementation support. Assessments support the decision. Your leaders remain responsible for it.
Bring us one difficult role. We will help you determine what evidence is missing and whether a focused pilot makes business sense.