67% of Indian HR Leaders Say AI Resumes Are Slowing Hiring. Voice Screening Is the Only Layer That Makes the Resume Stack Honest Again.

The same AI that’s writing the resumes is the AI you have to detect them with. The only screening layer that actually catches a ChatGPT-built profile is one that talks back — and won’t accept a memorised answer.

People Matters India published the stat this week that every TA leader in the country already knows from experience: 67% of HR leaders say AI-generated resumes are slowing the hiring process. Not marginally. Not as a theoretical future concern. Right now, in the April–June campus season, your screening queue is full of profiles that were written, formatted, and keyword-optimised by GPT-4 or its successors.

The resume is no longer a signal. It’s a performance.

What the AI Resume Problem Actually Looks Like

Here’s what’s changed in the last 18 months: generating a convincing, ATS-optimised resume for any role now takes a candidate approximately seven minutes and zero domain knowledge.

Paste the job description into ChatGPT. Ask it to generate a profile that matches the requirements. Add your name. Submit.

The resume will contain the right keywords. It will describe “experience” with tools the candidate has never used. It will quantify achievements with numbers that are either invented or inflated. It will look, to a resume parser, like a strong match.

This is the “AI vs AI” crisis: the AI that’s optimising candidate resumes is the same category of tool that HR teams are using to parse and score them. Resume-parsing AI sees an AI-generated resume and says “strong match.” Neither system is assessing reality.

The result: your shortlist is full of candidates who look great on paper and have no idea what they claimed.

The Specific Numbers from India

The People Matters data isn’t abstract. At HireQwik, our own pilot data tells the same story from the other direction:

• 1,500–3,000 applications per role is now the baseline for entry-level and fresher positions
• TCS alone is targeting 42,000 fresher hires — meaning the market is generating millions of applications across the sector
• After our voice screening pilot was re-evaluated against GPT-4.1 output in April 2026, the tier breakdown was: 67% No Go / 31% On Hold / 1.5% Strong Proceed

That 67% No Go figure isn’t just candidates being filtered out. It’s candidates whose resumes cleared initial screening but whose live voice conversations revealed they couldn’t speak to what they’d claimed. The resume said “proficient in data analysis.” The voice conversation revealed they’d never opened a pivot table.

Why Voice Is the Only Anti-Scripting Layer That Works

Resume parsing can’t fix this problem. Parsing is exactly what AI-generated resumes are optimised for — they’re designed to pass parsers.

Keyword matching can’t fix it either. If anything, keyword matching rewards AI-generated resumes more than human-written ones, because the AI knows which keywords the ATS is looking for.

Video submissions don’t fix it. Candidates can script and memorise a 90-second video response just as easily as they can generate a resume.

The only screening layer that’s genuinely hard to pre-script is a real-time voice conversation that probes beyond the first answer.

This is how HireQwik’s conversational AI works:

• The first question is about the role or the candidate’s stated experience
• If the answer sounds scripted or generic, a follow-up question probes the specific claim: “You mentioned you managed end-to-end recruitment at your internship. Walk me through what the process looked like for a specific hire.”
• If the follow-up also sounds scripted, the conversation goes deeper: “What did you do when a candidate withdrew late in the process?”

ChatGPT can generate a plausible first answer. It struggles to generate a plausible, contextually coherent response to an unexpected follow-up that references what the candidate just said. Real-time dynamic probing is the gap that scripted responses can’t cross.

What HR Teams Should Change Right Now

If you’re running campus drives this season and relying on resume screening as a first filter, you’re selecting for the candidates who are best at using AI to write resumes — not the candidates who are best at the job.

Three changes to make before your next drive:

1. Stop using resume quality as a first-round filter. Open the gate wide. Let everyone apply. The resume is no longer predictive.

2. Move the first substantive filter to a voice conversation. 15–20 minutes of structured, dynamic dialogue will tell you more about a candidate’s actual capability than any document they’ve submitted.

3. Specifically evaluate follow-up question performance, not opening answers. The opening answer is where preparation helps most. The follow-up is where genuine understanding shows — or doesn’t.

The Competitive Moat This Creates

For HireQwik, the AI resume crisis is directly our positioning. Our product was built communication-first from day one — because our founder’s view was that resume screening was already a weak signal before AI made it weaker.

The current moment is when “communication-first” goes from a philosophical differentiator to an urgent operational necessity. When 67% of your country’s HR leaders are telling People Matters that AI resumes are actively slowing their hiring process, the conversation has changed.

The moat isn’t the voice AI itself. It’s the real-time follow-up questioning that makes scripted answers visible. That’s the layer that makes the resume stack honest again.

Want to see the 5 questions our voice agent would ask for your specific job description — the ones ChatGPT can’t pre-script for? Send us your JD and we’ll show you the probing layer in action.