In the autumn recruitment interviews for the Class of 2026, such "confrontations" are becoming more and more common: on one side of the screen, the AI interviewer asks, times, and scores; on the other side of the screen, the job seeker uses AI-generated content to assist in answering in real time; the recruitment unit has subsequently upgraded anti-cheating detection systems such as eye tracking, speech speed analysis, template recognition, and dynamic questioning, and an offensive and defensive battle between AI interviewers and AI job seekers has begun. AI versus AI, who is the winner?
Survey data confirms the scale of this game: 74% of job seekers admitted to using generative AI during the application process, while 91% of recruiters found that job seekers used AI to give deceptive answers; 70% of recruiters trust AI's recruitment decisions, but only 8% of job seekers recognized the fairness of AI interviewers. Job seekers want to use AI to seek competitive advantages, and companies want to rely on AI to complete talent evaluations, but there is a lack of trust between them.
Are AI interviews fairer?
From the perspective of recruiting units, the use of AI interviews can help complete batch preliminary screening under the pressure of massive autumn recruitment submissions. It can compress the interview cycle and reduce labor costs by relying on unified question setting and standardized scoring. But its disadvantages are also prominent: on the one hand, it is difficult for the algorithm to understand the true ability, and it is easy to misjudge job applicants who are not good at expressing on camera; on the other hand, the biological information such as videos, faces, and voices collected by the camera is sensitive information, and if it is not properly kept, it will bring the risk of privacy leakage.
Additionally, AI interviewers use a consistent assessment framework in an attempt to reduce subjective bias. The results of a meta-analysis (a statistical method that quantitatively synthesizes the results of multiple independent studies - reporter's note) show that traditional interviews have limited ability to predict job performance. Interviews can only explain 9% of the differences in job performance, and the remaining 91% of the true ability is difficult to fully demonstrate through a short conversation. Live interviewers are extremely susceptible to the halo effect and unconscious bias, and will be confused by the applicant's eloquence and on-the-spot appeal. The birth of the AI interviewer also carries the mission of weakening human subjective bias and achieving fairer talent selection. The reality is the opposite: AI may make unfairness more hidden and difficult to hold accountable. The upper limit of the capabilities of an AI recruitment tool is determined by the human data used to train it. The bias inherent in the training data will be further amplified by the algorithm. If the company uses existing internal excellent employees as the training data source, and the old employee group itself lacks diversity, AI screening will solidify the original homogeneity problem of the team. What’s even more frightening is that AI interviews have not eliminated the unfairness of interviews, but have instead spawned new technological confrontations.
AI, originally used to balance interview bias, failed to eradicate injustice and instead opened up a new technological game. The old problems remain, and a new game arises. To understand why this confrontation cannot stop, we must first answer another question: Why are some people willing to take the risk and use AI to answer interviews?
Why do you still choose AI to answer questions when you know there are risks?
From the perspective of job seekers, AI can sort out the logic of answering questions, fill in shortcomings in expression, and help it output interview answers that are more in line with job expectations. However, under fierce competition, some people began to cheat by using AI to output answers in real time. Some psychological research points out that when people rely on AI tools for interpersonal interactions, their own moral constraints will weaken and they will easily engage in deceptive behavior. More importantly, individuals will also complete the psychological construction of self-rationalization, find excuses for their own behavior, and generate thoughts such as "Everyone does it, I'm just coping with the system rules", "It's not that I want to cheat, it's the AI interview itself that is full of routines, I just use technology to fight against technology" and other thoughts, in order to eliminate the inner guilt.
Researchers have revealed the psychological mechanism behind this phenomenon through a series of experimental studies: entrusting tasks to AI will create a sense of psychological distance and promote moral disengagement. Job seekers do not need to directly issue explicit cheating instructions such as "help me make up interview answers." Instead, they only need to make vague demands, and then they can use large models to generate complete interview responses and complete self-interested deception with the help of machines. In this process, it is easy for people to transfer responsibility for their behavior outwards and attribute problems to "the routines of the interview system itself", thereby eliminating their inner guilt.
In the autumn recruitment, the use of real-time AI question and answer prompting tools will transfer the interview response pressure of job seekers to AI, reducing the perception of unethical consequences, and their own sense of guilt will also be weakened. However, the ability disguised by AI is difficult to withstand the test of real scenes. There are many real-life cases: some fresh graduates relied on AI to pass the preliminary screening, but after entering the final interview with a real person, the interviewer followed the questions and dug into the details of the project. The candidate's answers were full of loopholes, and he was ultimately unsuccessful.
A war of attrition for trust with no winners
The foundation of recruitment is two-way trust: the recruiting unit believes that the interview reflects the true level of the job seeker, and the job seeker believes that the evaluation criteria are fair and objective. This AI offensive and defensive war has eroded two-way trust, and no one is the real winner.
If units want to identify talents from templated answers, they can only continue to increase anti-cheating measures, design anti-routine test questions, add written tests and multiple rounds of real-person retests, and the time and labor costs of recruitment continue to rise.
For job seekers, when tools that use AI to cheat are rampant, the job market falls into the dilemma of "bad money drives out good money": those job seekers who stick to the bottom line and refuse to use tools, but their simple and sincere answers are at a disadvantage in the algorithmic scoring system compared to the watertight AI rhetoric.
Reposition AI and return to the essence of recruitment
AI itself is not a scourge, but recruiting parties need to redefine its positioning.
Job seekers can use AI to prepare, but do not use AI to replace answers. Before the interview, job seekers can ask the AI to simulate different styles of interviewers and conduct a full mock interview; after the exercise, ask the AI to comment on the answers to identify logical loopholes and expression shortcomings; and then modify the reference words output by the AI based on their own real experience. During the interview process, job seekers should not use cheating tools such as real-time monitoring or floating prompts. Once flagged by the anti-cheating system, it will leave a stain on their job search. Even if they are lucky enough to get hired, they will be easily exposed during subsequent in-depth questioning and face the risk of terminating the labor contract.
Recruiting units require human-machine collaboration and return the final decision to real people. AI is suitable for completing preliminary screening, recording interviews, transcribing minutes, counting behavioral signals, and marking suspicious risks, but the final judgment of talents must be left to real people.
There is no winner in this AI battle, but a continued loss of trust. AI cannot replace a person's real experience and thinking. What determines career development is not the perfect answer during the interview, but the real ability accumulated behind the job seeker.
