The hiring manager who was rejected by their own job ad
Screening software rejected a manager's own CV in minutes over one word. The real problem wasn't the filter. It was that nobody looked at what it threw away.
Arku team
A story keeps doing the rounds online. A hiring manager's role has sat open for months, and HR keeps reporting that the applicant pool is thin. So the manager applies for the job themselves, under a different name and email, with their own CV.
They're rejected automatically within minutes.
The screening software was set to look for one exact phrase, "React.js". The manager's CV said "React", like most developers' do. Everyone who wrote it that way was filtered out before a person saw their application.
In the version that went viral in 2024 the framework was AngularJS rather than Angular, and the details shift with every retelling. We can't verify any version of it. It keeps spreading anyway, because anyone who has hired through screening software recognises it.
The filter worked perfectly
Nothing broke. The software did exactly what it was told. Someone typed a rule, the rule was slightly wrong, and nobody checked what it was rejecting. A rejection nobody sees is a rejection nobody can question.
That is the real failure. Not the keyword, but a process where the first reader of every application is a rule, and no person ever reads the ones the rule turns away.
The usual advice fixes the wrong side
The post that shared this story ended the way these stories usually do. Its advice was for candidates: copy the job ad's exact wording, and perhaps use a tool that rewrites your CV to match it.
That works, in the way a workaround works. It asks every candidate to guess how your filter is spelled. The people best at that guessing aren't necessarily the people best at the job. And the employer is still missing everyone who didn't play along.
AI ranking doesn't fix it either
The newer pitch is to replace keyword rules with AI that scores or ranks applicants. That makes the problem harder to see, not smaller. A keyword rule can at least be read and corrected. A score from a model gives no reason you can check, so the "React" candidate doesn't get rejected. They quietly end up at the bottom of a pile nobody gets to.
Some places now regulate this. New York City, for example, requires a bias audit before employers use automated tools to screen candidates for jobs in the city.
| Topic | Automated screening | How Arku works |
|---|---|---|
| Who reads each application | Software first. A person only if it passes. | A person on your team, every time. |
| "React" instead of "React.js" | Rejected, if that's what the rule says. | Read by someone who knows they're the same thing. |
| Screening questions | A wrong answer can reject someone automatically. | Answers are shown to the reviewer. Nobody is rejected automatically. |
| Rejections | Can happen without anyone noticing. | Someone on the team rejects the application and records a reason. |
| AI | Ranks, scores or summarises candidates. | Never reads, ranks or scores candidates. |
Reading applications is the job
Reading every application takes time. For a small team hiring a handful of people a year, it is time well spent, and it is usually less time than the months a role sits empty waiting for a filter to let someone through.
That is why Arku is built to make reading faster rather than to skip it. Every application, CV and answer is in one place, searchable, with scorecards your team fills in. People make the decisions. The software keeps them organised.