The AI-resume flood: how to screen when every application looks perfect

The average tech job posting in India now receives 340 applications in the first 72 hours, up from 82 in early 2023. The share of those applications that are AI-generated, AI-optimised, or in some cases AI-fabricated is somewhere between 60% and 80%. The traditional signal — keyword match, formatting quality, cover letter effort — is now noise.

The signals that still work

• Specificity of impact. AI still writes in generalities. A resume that names the specific system, the specific metric, and the specific timeframe is almost always human-written and worth reading.
• Coherence across the arc. AI-generated resumes optimise each role individually. Human-written resumes tell a story where each role builds on the last.
• The public trail. GitHub, blog posts, conference talks, open-source PRs — inputs that require multi-year effort and cannot be generated overnight.
• The reference call. Still the highest-signal, lowest-adopted tool in screening. If you skip it, you deserve what you get.

The screening question that cuts through

One question, asked in a 15-minute call, resolves most of the AI-noise problem: 'Walk me through the last non-trivial technical decision you made and why.' Candidates who wrote their own resume can answer this in specifics. Candidates whose resume was generated struggle to produce the underlying detail, because the detail was never theirs to begin with.

The uncomfortable implication

The volume of applications a job posting receives is now nearly meaningless as a signal of the quality of the pipeline. Talent teams that still report on it as a KPI are measuring the wrong thing. The right metrics are qualified candidates per week, offer accept rate, and 12-month retention of the resulting hires. Everything else is theatre.

"In a market where every resume looks perfect, the recruiter's job is not to filter. It is to talk to enough humans to find the ones who are."