2026-11-18 08:26
Lead-in: Save 87% in labor costs, but embed hidden bias. 600 resumes, 15 AI interviews, 7 human interviews, 0 job offers.
600 resumes, 15 AI interviews, 7 human interviews, 0 job offers.This is the job-hunting record of a recent graduate in autumn 2026.
This year, AI-driven interviews have become mainstream in recruitment markets. Job seekers are now studying AI’s preferences to pass algorithmic evaluations. But these shifts go far beyond the interview room.
Companies use AI to screen candidates, schools use AI to assess teachers and students, hospitals use AI to grade surgeons’ procedures… Humanity is lining up, one after another, under the scrutiny of artificial intelligence.
Previously, humans evaluated AI. Now, AI has taken the examiner’s seat and begins assessing us. Every ordinary person faces a fundamental question: How do you achieve a passing score within an AI evaluation framework?
Living individuals cannot be reduced to mere numbers. Yet, the AI systems holding the scoring rubrics must ultimately be governed by human oversight.
I can’t remember how many times I’ve seen this woman. Xu Qing opens her computer and encounters the familiar interviewer once again.
She has short, neat hair, wears a light-colored suit, and maintains a rigid smile. As Xu responds, the interviewer nods thoughtfully or shakes her head slightly.
In the 2025 autumn recruitment season, Xu Qing, a second-year postgraduate at Shanghai University of Finance and Economics, noticed that nearly every foreign company responding to her application sent this AI interviewer. From September onward, she participated in nearly 20 AI interviews.
Xu Qing is naturally outgoing and articulate. In the past, she viewed interviews as a battle. The high-pressure atmosphere and tricky questions only fueled her competitive spirit—she loved seeing interviewers stunned by her answers.
But when she clicked into the French electrical firm’s interview link, that drive hit a wall of cotton.
No matter how passionately she spoke, the interviewer on screen never altered her expression—no slight upward curve of the lips. She would simply nod politely when Xu paused, then respond in a mechanical tone: “Alright, let’s move to the next question.”
This year, such interviews have become increasingly widespread. One recent graduate said over a third of their 20+ interviews were AI-led. Facing the screen and speaking to no one, they felt “like a bunch of greens picked at will.”
AI interviews follow strict protocols. Some candidates haven’t even finished their five-minute self-introduction when the countdown hits zero—the system cuts them off without hesitation and proceeds to the next stage.
Meanwhile, software instability remains a major issue. One candidate experienced four crashes during the interview, each requiring re-login via facial recognition—and having to restart all previously answered questions.
No sacrifice too great for education; no hardship too severe for children. The most advanced technology has already entered classrooms.
A few years ago, Hangzhou No. 11 Middle School installed multi-camera arrays in classrooms to capture six key student behaviors—reading, raising hands, slouching—and seven emotional states including joy, aversion, and fear—to alert teachers.
This year, Jiangnan University deployed AI teaching analytics across all 462 public classrooms, monitoring attendance rates, front-row seating, and head-up posture in real time. Nanjing University sparked controversy with its “smart classroom” reporting a 95% head-up rate.
Students openly admit feeling suffocated in such environments: “It feels like eyes are always watching.”
Using AI to grade student essays has long been routine.
One student wrote about her grandmother’s vegetable garden, moved by genuine emotion and childlike language—earning praise from her teacher. But the AI assigned a low score, citing “flat language” and lack of rhetorical devices.
Teachers and physicians are also under AI surveillance. In Shanghai Normal University’s mock classroom, lecturers face nine AI-generated virtual students, each with distinct personalities.
At Shandong University Qilu Hospital’s annual performance review, 87 resident physicians sat across from an AI-standardized patient developed through industry-academia collaboration.
After completing a round of CPR training, a pediatric attending physician at Maoxian County People’s Hospital found a large screen immediately replaying the procedure frame by frame: hand placement deviation, compression depth, rhythm speed—all marked and assessed by the system.
Over twenty doctors underwent testing on simulation tools, while cameras, pressure sensors, and audio recorders captured movements, force, and verbal commands in real time. Any pause triggered an AI comparison against expert benchmarks, delivering scores and improvement suggestions.
Job seekers answer AI questions, students learn under camera surveillance, teachers generate data reports from lectures, and doctors’ every action is machine-monitored.
Humans who once evaluated AI are increasingly becoming the observed, tested, and judged subjects.
Last autumn, Xie Yi, a third-year postgraduate at Xiamen University, was asked in an AI interview:“When encountering conflicting information sources at work, what would you do?”
Xie Yi felt a surge of excitement—she had just memorized the ideal response the night before. She immediately launched into a story about mediating conflicts during her internship.
But the AI pressed further: What exactly caused the disagreement? Can you detail your process for balancing both parties?
Xie Yi panicked instantly. In reality, there had been no internal conflict during that internship—those two sides were fabricated solely to fit the interview narrative.
When AI becomes the gatekeeper, people must begin studying its inclinations. To decode this unpredictable examiner, young job seekers turn their entire preparation into a high-stakes exam drill.
Dai Lin, immersed in endless practice questions, gradually noticed that assessment questions across companies were converging. Master the pattern, and high scores come effortlessly.
On Xiaohongshu (Little Red Book), there are over 40,000 posts about AI interviews. Users have uncovered high-scoring strategies: fill time completely, structure responses point by point. Most importantly, hit keywords—these are the AI’s scoring triggers.
Chen Qian, a third-year postgraduate at Xiangtan University, tried every trick. To appease the AI, she revised her speech based on experience posts, even applying subtle makeup to project alertness and energy.
Yet after eight AI interviews, she received nothing. Yang Yang, who endured over ten AI interviews, went from deliberate practice to complete numbness—“I stopped bothering with makeup altogether.”
Those driven to desperation by AI decide to fight fire with fire.
On Xiaohongshu, a ¥68.8 AI interview assistant sold 1,068 units. Purchasers receive two hours of premium access—enough to handle four interviews. It claims real-time question monitoring, millisecond answer generation, and even eye-contact correction features.
Some job seekers simply taped their smartphone running a large language model app directly onto their computer screen. When the AI interviewer posed a question, the phone generated a response, which the candidate read verbatim.
China Media University, following the trend, launched a career guidance course for third-years. Students can freely trial the “AI Mock Interview & Soft Skills Analytics Platform,” experiencing the full AI interview flow and receiving analytical reports.
However, Professor Yan Yugang from the School of Economics and Management admitted that the course failed to fill even half its 500 spots—“Clearly, most undergraduates still don’t grasp its importance.”
Across the ocean, 26-year-old graduate Meriem Karasi not only uses AI to optimize her cover letters but also purchased an automated application tool.
“It lets me apply for jobs even while dreaming,” Karasi said. “A carefully crafted resume gets scanned for only 30 seconds in the system. If so, why not let AI do the work?”
In classrooms, children are already compromising to algorithms. In 2026, a primary school Chinese language researcher in Foshan, Guangdong, noticed a shift in student essays.
Previously, writing “My Mom,” children described her cooking meals or riding her bike in the rain to pick them up.
But since introducing AI grading, many learned shortcuts. They began every essay with “Time flies, years pass swiftly,” stuffed the middle with ornate parallel sentences, and concluded by elevating the theme to life philosophy.
Yet no matter how much humans conform, outcomes remain unpredictable—you never know where you’ll fail.
In an AI interview, job seeker Chen Chen was asked a psychological test question: “Given a blank sheet of paper, what do you think of?” He answered honestly: “An A4 sheet.” He was then eliminated—with no explanation provided.
In October 2026, the topic “AI Interview Uncanny Valley” trended on Weibo. One applicant commented: “This interview is one-way—it can scrutinize you, but you can’t challenge it.”
Lin Yi, responsible for developing AI interview tools at a human resources platform, oversees the debugging of so-called AI interviewers.
But the results have long exceeded his expectations. Even he cannot explain how the AI interprets or evaluates candidates: “The algorithm’s inner workings are a black box.”
Still, the market rushed to put it on the table.
The founder of an AI recruitment company calculated: a full-time HR can conduct about 10 interviews per day, costing roughly ¥1,000.
By contrast, their AI recruiter can interview over 1,000 candidates simultaneously, with daily cost under ¥200—achieving 66% time savings and 87% reduction in labor expenses.
Amid massive volumes of resumes, efficiency trumps everything. Zhang Zhu, an HR at a Beijing-based foreign enterprise, knows this well.
Before AI integration, her department processed thousands of resumes during peak seasons, requiring her to wear headphones continuously, listening to candidates’ English recordings—“It was simply overwhelming.”
Xu Ying, a hiring manager at a Shenzhen tech firm, receives 30,000 applications annually. State-owned banks receive over a million campus recruitment resumes yearly, yet hire only 10,000.
Efficient and cheap, AI naturally became the first choice for many companies.
Operating in a black box, bias is inevitable. AI relies only on rigid textual knowledge, lacks clear evaluation criteria, and may even reflect the subjective biases of its creators.
In 2026, after six months of trial, a university recruitment platform discontinued its AI interview tool. One graduate discovered the AI forcibly merged two sections of her resume into a fabricated story—one that never existed—then questioned her about it.
In July 2026, a People’s Daily reporter tested 18 primary and secondary school essays using six popular AI grading tools. The same essay scored up to 20 points apart across platforms.
Even the same essay received different ratings when submitted in the morning versus afternoon—differing by as much as 10 points.
Researchers at the University of Washington found that identical responses received varying scores based on accent. Amazon’s AI recruiting tool automatically downgraded resumes containing the word “female.”
Remote proctoring software Proctorio and ExamSoft faced backlash from over a thousand educators and students after being accused of failing to recognize darker-skinned test-takers—forcing some to take chemistry exams in Starbucks parking lots.
In Zhang Zhu’s company, all AI-passed interview videos undergo manual re-review. But those rejected by AI have no appeal pathway.
To achieve true fairness in technology, human intervention must occur before systems gain total control.
Since last year, Jiangxi Province has adopted AI surveillance in its Gaokao exams. However, the Education Examination Authority stipulates that AI serves only as a monitor alert—not as direct evidence, nor can it independently determine cheating. All findings must be verified by human examiners.
In classrooms, teachers are explicitly prohibited from using AI’s automatic grading of open-ended assignments—essays, art projects—as final student evaluations.
In medical settings, AI usage is even more cautious.
Dr. Yang Liqun, an anesthesiologist at Renji Hospital in Shanghai, encountered a suspected case of malignant liver tumor during rounds.
He input the patient’s records into the AI system. The machine detected the patient’s habit of eating raw freshwater fish and shrimp, along with abnormally elevated eosinophils. The AI flagged: “Watch out for Clonorchis sinensis.” Following the hint, doctors added antibody tests—ultimately confirming a parasitic infection.
“AI provides us with insights and prompts,” Yang said. “We make the judgment. So far, AI remains merely a tool. Medical decisions must always be made by doctors—this has never changed.”
In the recruitment market, growing professional voices are calling for a return to human judgment. Experts advocate that AI screening should not carry veto power.
Many large enterprises that adopted AI screening still retain final decision-making authority over second-round interviews and job offers in the hands of HR.
For ordinary people facing screens, the long adaptation continues.
In the British series Black Mirror, one episode is titled “White Christmas.”
In that fictional world, everyone has a score. People rate each other. One’s score determines the quality of life.
To earn a high score, protagonist Lexi practiced smiling daily. She spoke cautiously, flattered everyone around her, striving to become someone universally liked.
But the ending was not happy. Her score plummeted repeatedly—until, after being expelled by the system, she finally shed her mask and screamed aloud in relief.
Yet scores are not people. One interview cannot measure job capability. One essay cannot represent overall writing skill. One operation report cannot replace a doctor’s clinical judgment.
AI may calculate standard answers—but living, breathing humans are never part of the algorithm’s evaluation criteria.
Source: NEXT Trends
Disclaimer: Contains third-party opinions, does not constitute financial advice
The AlphaGo Moment in Cybersecurity, or Starting from These 466 Agent Matchups
12 days ago
The Trump administration plans to offer Constellation Nuclear a $5 billion loan for capacity expansion, with the additional power mostly pre-allocated to AI giants such as Google
20 days ago
Silicon Valley Executives Spark Controversy Over Using AI for Household Tasks; Survey Reveals Half of American Adults Worry About Impact on Relationships
20 days ago
Tensorlake npm package suffers supply chain attack, malicious version 0.5.144 injected with Shai-Hulud data-exfiltration worm
10-08
Tesla and Musk Have Yet to Define "Amazing Abundance" — Visionary Slogan Sparks Skepticism
10-01
Instinct Launches Curated Recommendation Feature, Sparking Frustration Among Some Users Receiving Passive Product Suggestions
09-30
Security Weekly Report: Placeholder Domain Takeover Impacts Approximately 1,700 Repositories, with Major Cyberattacks Worth $387M and AI Agent Escalation Risks Highlighted
09-28






