A comment I saw this week that stuck with me:
“Apparently it’s even more intense in North America: you use AI to mass-apply, they use AI to mass-screen.”
That’s a description of an arms race. And neither side is sure they’re winning.
What’s actually happening
AI-assisted job applications are now common in North America. A candidate can use AI to quickly customize a resume and cover letter for each specific job description — something that used to take 30 minutes now takes 5. Application costs have dropped, so application volume has gone up.
On the HR side, ATS (applicant tracking systems) have been filtering applications for years before AI. With AI assistance, screening is faster and more automated. A role might get hundreds of applications; the average human review time could be seconds.
Both sides are using AI to accelerate. Both have become more efficient. But the collective outcome — better matching between jobs and people — isn’t obviously improving. Candidates spend more time applying to more jobs; HR spends more time filtering more applications; the signal-to-noise ratio may not have improved for either party.
This is a classic prisoner’s dilemma. Each individual using AI is a rational choice. But when everyone uses AI, the collective result isn’t better.
The Taiwan version
Taiwan hasn’t reached North American scale, but the pattern is emerging. AI-generated resumes are becoming recognizable. HR teams report being able to identify “this cover letter was written by AI.” Some companies have shifted to opening interviews with specific, detailed questions to verify whether candidates actually did what their applications describe.
Another arms race: candidates use AI to make applications look stronger; HR uses questions and intuition to filter out the AI layer.
The counterintuitive result: the person who spent the most time carefully crafting their resume may have no advantage over someone who mass-applied, simply because volume beats precision in the early filter.
Who’s actually benefiting?
Short-term, two groups come out ahead in this bilateral AI race:
On the candidate side: People with genuinely distinctive backgrounds and real accomplishments who weren’t good at expressing them on paper. AI helps them say true things clearly. That’s the legitimate use case.
On the HR side: Companies with well-designed, specific JDs and clear screening criteria. The more concrete your criteria, the more AI tools can accurately execute against them. Vague JDs and vague screening standards, hit with a wave of AI applications, become more chaotic, not less.
Who benefits least: anyone treating AI as a low-cost spray-and-pray tool without actually improving their underlying match with the roles they’re targeting. More volume, same (or worse) signal quality.
A question without a clean answer
Arms races reach new equilibria eventually. The current one is still early. Possible endpoints: AI applications become the baseline and no longer differentiate anyone, and attention shifts to referral networks and real interactions. Or some new verification layer emerges. Or something else entirely.
Before that equilibrium arrives, what’s actually worth doing?
My read: use AI to express yourself clearly, but make sure you’re actually the person your application describes. AI can accelerate packaging. It can’t fill a substantive gap.
At the end of the process, the person screening applications is still looking for a person — not a well-designed document.