AI is Becoming the Recruiter’s Lie Detector

For years, the hiring conversation around artificial intelligence has centred on speed: faster shortlisting, faster scheduling, faster offers. However, according to Mridul Kumar, Senior Vice President, Products, HirePro, that framing misses the more consequential shift now under way in Indian recruitment: using AI not merely to accelerate hiring, but to make it more honest.

“AI is primarily looked at as an automation tool, a tool that makes things faster and takes fewer people,” Kumar says. “But at a higher level, how we hire better is where the true opportunity lies.” 

That distinction, he argues, has become urgent for a simple reason: candidates now have the same generative AI tools that recruiters do, and are using them to manufacture the appearance of competence.

The Problem of the ‘Perfect’ Candidate

Kumar frames the challenge as twofold. First, technical skills now carry an expiry date—what a candidate learnt even six months ago may already be outdated, making it unreliable to judge people purely on past credentials. 

Second, and more pointedly, candidates can use AI to produce flawless resumes, flawless test answers, and rehearsed interview responses. “A polished profile, a polished answer, no longer proves the real ability of the candidate,” he says.

The concern is reflected in HirePro’s newly released report, The State of College Hiring in India 2026, compiled from responses by more than 10,000 students, over 80 corporates, more than 100 colleges and over 100 campus recruiters.

The report finds that 66% of corporates believe AI-proctored assessments improve fairness, yet only 29% feel these tools adequately capture soft skills. Meanwhile, 56% worry that automated screening risks filtering out unconventional talent. In other words, employers want AI-enabled fraud detection, but remain wary of collateral damage.

Building the ‘Unfair Advantage’

Kumar’s argues that genuine differentiation will not come from AI alone, since “everyone is renting the same AI”. Instead, he points to what Careernet-HirePro describes internally as an “unfair advantage”: more than 25 years of recruitment data, access to over 1,500 recruiters, and a processed history of more than 20 crore candidates across assessments and interviews spanning every industry and seniority level. 

This accumulated memory, he says, is what allows the platform’s AI to distinguish a candidate’s authentic capability from an AI-assisted performance by comparing patterns across millions of prior interactions, rather than judging a single response in isolation.

In practice, this shows up across three stages of hiring that Kumar calls attract, screen and evaluate. At the screening stage, deep contextual resume-matching goes beyond keyword parsing to weigh a company’s culture, a role’s real requirements, and the way a specific team operates, allowing the system to extract genuinely meaningful signals even from an AI-polished resume.

Catching Fraud in Real Time

It is at the evaluation stage, however, where fraud detection becomes most explicit. Kumar described a layered model combining structured assessments, AI-led interviews, and human expert interviews, each calibrated differently depending on the seniority and sensitivity of the role.

Within AI-led interviews, the system is designed to move beyond static multiple-choice testing towards scenario-based evaluation. Candidates are presented with a business case, a policy dilemma, or an ethical technology problem, and the system judges the depth and originality of their reasoning rather than a single scored answer. 

Follow-up questioning is generated dynamically to establish whether a candidate is “actually answering on their own” or reciting memorised or AI-generated content, Kumar said. The platform also assesses interview delivery itself, including pacing and structure, to flag anomalies consistent with off-screen assistance.

For senior roles, where candidates may be reluctant to be interviewed solely by an AI agent, Careernet’s approach pairs a human interviewer with an AI assistant working alongside them in real time. The AI guides the interviewer on which questions to ask, and gives them the option to permit AI as a ‘hint tool’ or to test coding ability without any AI support at all, depending on what the role actually demands once the candidate is hired.

Talking about the how recruitment has undergone a massive shift in the past decade, S Pasupathi, Chief Operating Officer, HirePro says, "I believe the changes have been remarkable, especially in the last five years. Before COVID-19, the recruitment was done almost entirely physically. For example, companies would visit campuses for entry-level hiring, and lateral candidates would walk into offices for multiple interview rounds. In simpler terms, candidates saw the workplace, met people, and absorbed the organisation’s culture. The process had the element of physical connection."

Bias: The quieter Half of the Problem

Fraud detection dominates the discussion, but the HirePro report suggests bias correction is just as significant a design goal—because human bias in hiring often hides behind seemingly objective, credential-based decisions. Despite an industry narrative of a ‘skills-first’ hiring revolution, the report found that only 21% of corporates have fully adopted skills-based hiring, while 51% still allow CGPA and college pedigree to influence decisions. 

Notably, 96% of colleges believed most corporates had already gone skills-first—revealing a considerable perception gap that AI-led assessment is intended to close by standardising evaluation criteria across candidates, rather than leaving decisions to individual, potentially biased, human judgement.

The report also uncovered what it calls “the cultural-fit blind spot”: About 58% of corporates cite cultural fit and personality as top hiring priorities. Yet, only 27 % of students believe these factors matter to employers—the single largest perception gap recorded in the survey. 

Left unaddressed, such gaps can entrench bias, since recruiters may unconsciously weight unstated criteria that candidates have no opportunity to prepare for or contest. Structured, AI-assisted evaluation is positioned as a way to make such criteria explicit and consistently applied.

Enduring Skills Over Expiring Ones

Kumar argues that the more durable answer to both fraud and bias lies in recalibrating what is measured altogether. Rather than weighting current technical syntax or tool familiarity, skills he says can expire within months, organisations should give greater emphasis on enduring capabilities such as reasoning, problem-solving and, above all, learning agility. 

“How fast can somebody learn what's next? How fast can they adapt?” he says, describing this as the most future-proof marker of talent as AI reshapes job requirements.

The report’s own data lends some support to this shift. While 85 % of corporates say they prefer candidates to be GenAI-ready, only 12% currently apply this as a hard filter, suggesting most organisations are still choosing to train new hires in AI tools rather than screen for them outright.

A Platform, Not a Single Tool

Kumar was careful to frame all of this as configurable rather than prescriptive, describing an “intelligent, trusted, effortless” platform in which agentic and do-it-yourself approaches coexist, and where the mix of assessment, AI interview and human interview varies by role and urgency. 

Whether such systems can keep pace with increasingly sophisticated AI-assisted candidates remains an open question. But as this year’s data shows, the perception gaps between recruiters, colleges and students are wide enough that even incremental progress on fraud detection and bias correction may reshape how Indian companies hire from campus.