Job Seekers Use Hidden Resume Prompts To Outsmart AI Screeners

The tactic, known as prompt injection, can involve placing commands in extremely small white text that is effectively invisible to a human recruiter.

Job applicants are embedding hidden instructions in resumes to manipulate artificial intelligence screening tools, a practice that is raising new questions about employers’ growing reliance on automated hiring systems, Inc.com reports. reports.

The tactic, known as prompt injection, can involve placing commands in extremely small white text that is effectively invisible to a human recruiter but readable by an AI system. Some instructions tell the technology to disregard other information and advance an applicant in the hiring process, according to an Aug. 21 Inc. column by recruiting executive Deepali Vyas.

Vyas, founder and CEO of Vyas Media and The Elite Recruiter, argued that applicants’ attempts to circumvent AI screeners expose shortcomings in hiring systems designed to process large numbers of applications.

ahhhhhhh so I'm hiring a tech to help me out in lab and screening applications now and THE KIDS ARE USING PROMPT INJECTION!!!!!! 2.25 pt white text, here's what I've found so far pic.twitter.com/s8TtKxqUgE
— Ya'el Courtney, PhD (@ScienceYael) July 28, 2026

“When you automate the front end of your hiring process without any human judgment in the loop, you don’t just screen out bad candidates. You screen out good ones, too,” Vyas wrote.

Research from Duke University suggests the practice extends beyond isolated cases. Researchers working with industry partners examined 200,000 real resumes submitted to recruiting platform HireEZ and found that at least 1% contained concealed instructions intended to manipulate AI systems. Duke announced the findings on July 22 and said tutorials, templates, and online videos were contributing to the tactic’s spread.

Prompt injection can include hidden commands or keywords designed to influence how an AI tool evaluates an applicant. Researchers said early large language models were particularly susceptible to following such instructions, creating vulnerabilities as employers increasingly incorporate AI into recruiting.

Vyas, who said she has spent more than 25 years placing senior executives, cautioned that automated screening can help employers narrow thousands of applications but does not ensure the remaining candidates are the most qualified.

She argued that unconventional candidates can be disadvantaged when their experience, career changes, or accomplishments do not align neatly with the criteria an automated system is designed to recognize.

The issue, Vyas wrote, is ultimately less about applicants finding ways around AI than about whether employers have become too dependent on automated filters for decisions that require human judgment.

“If your hiring process can’t tell the difference between a great candidate and a well-prompted one, the problem is not the candidates,” Vyas wrote. “It’s the process.”