AI interview cheating is a live help problem. A model, an overlay, or a second device feeds answers while you watch. Identity fraud is a different problem: deepfakes, proxy sitters, and bait and switch day one hires. Engineering and IT hiring managers need both on the radar, but they are not the same claim, and they need different fixes.
LinkedIn News on August 30, 2026 summed Wall Street Journal reporting by Katherine Bindley: stricter screening so candidates verify qualifications, cheating detection for tab toggling, camera pans around the room, and live assignments without chatbots. That LinkedIn News recap is the hook for this page. We cite Bindley’s own LinkedIn notes and the WSJ lede that is public. We do not claim we read the full paywalled WSJ body.
This is employer advice for engineering and IT hiring. It is not a cheat app guide. Apollo Technical is a specialized engineering staffing firm. We do not sell proctoring software, and we do not invent a detection rate for our own desk.
What is AI interview cheating?
AI interview cheating means unauthorized help during a live or recorded screen. ChatGPT interview cheating, voice mode on a phone, a second screen, or an invisible overlay that feeds lines while the camera stays on are in that bucket. Candidates using AI in interviews can look calm and fluent. Live coding interview AI can produce clean code with thin debugging history.
Identity fraud is separate. Deepfakes, glasses that do not move with the head, a different person day one, and schemes that put a remote IT worker who is not who they claimed are identity and security problems. Keep those labels apart so you do not “solve” one with a tactic built for the other.
SHRM Labs calls skillfishing the pattern of claiming skill you only borrowed. In a SHRM Labs brief, Fabric’s own platform data on nearly 20,000 interviews from July 2025 to January 2026 found 38.5 percent of candidates on that platform showed signs of cheating, with technical roles highest at 48 percent. That is not a census of every job seeker. It is vendor interview data cited by SHRM Labs.
How big is the problem for hiring managers?
Use named surveys. Do not mash them into one “the number.”
Gartner, in a Gartner newsroom release, predicts that by 2028 one in four candidate profiles worldwide will be fake. That is a forecast, not today’s rate. The same release says a 2Q25 survey of 3,000 job candidates found 6 percent admitted to interview fraud, either posing as someone else or having someone else pose as them.
Greenhouse’s 2025 AI hiring release surveyed 4,136 people, including U.S. recruiters and hiring managers. Exact claims from that release: 91 percent of recruiters have spotted candidate deception; 65 percent of hiring managers have caught applicants using AI deceptively, including reading from AI generated scripts (32 percent), prompt injections in resumes (22 percent), or showing up as deepfakes (18 percent); 39 percent of U.S. hiring managers are conducting more in person interviews to verify authenticity.
Checkr’s Hiring Hoax survey of 3,000 U.S. managers says 59 percent suspected a candidate of using AI tools to misrepresent themselves; 31 percent interviewed someone later revealed to be using a fake identity; 35 percent say someone other than the listed applicant participated in a virtual interview; 23 percent report losses of more than $50,000 in the past year due to hiring or identity fraud, and 10 percent say losses exceeded $100,000. Only 19 percent are extremely confident their process would catch a fraudulent applicant.
Those figures are manager self reports and survey samples. They are not Apollo Technical’s placement data.
How to prevent AI cheating in interviews?
How to prevent AI cheating in interviews starts with design and clear rules, not with a panic buy of spyware.
Say which prompts ban AI and which allow it. Bindley wrote on the LinkedIn News page that companies want to see AI use on some questions and pure thinking on others. Put that split in the invite. If the rule is silent, every candidate invents their own.
Use live work. LinkedIn News says recruiters are pivoting to live assignments so they can see problem solving without a chatbot. For engineering, that means a ticket style task, a bug you already closed, or a design you can change mid call. Ask them to change the solution after they “finish.” A pasted answer breaks. A real engineer adapts.
Ask better follow ups. An Editors’ Pick on that same LinkedIn story put it plainly: LLMs are fluent on abstract leadership prompts and weak on lived detail. “So how did your boss react?” “What was going through your mind when you heard that objection?” “Going back to that earlier supply chain problem, what did you learn from that?” Walk the story. Jump back to an earlier claim.
Bring high signal work earlier in person when the seat is scarce or remote. Stanford’s Nick Bloom noted on the LinkedIn News thread that two large tech firms are pulling in person interviews and on site coding tests earlier for AI concern reasons. An earlier WSJ story, summarized by Axios and Computerworld, named Google, Cisco, and McKinsey among firms adding or restoring in person stages. Greenhouse says 39 percent of U.S. hiring managers are doing more in person work. That is “more,” not “most employers.”
Allow AI on a second round on purpose when the job will use it. Have them use a model, then defend what they kept and what they threw away. That separates people who can think with tools from people who can only read tools.
How to detect AI in interviews?
How to detect AI in interviews is a stack of signals, not one red light.
Watch for timing that does not match difficulty. Watch for perfect code with no intermediate runs. Watch for answers that collapse when you ask why they chose that trade off.
Bindley describes cheating detection software that monitors browser activity and eye movements, plus low tech moves like showing hands while a question is read, or covering eyes with a hand. LinkedIn News attributes tab toggle detection and camera pans to WSJ reporting on “some companies” and “hiring managers.” Treat those as reported examples, not a measured standard. Tab locks alone do not catch invisible overlays that never switch tabs. SHRM Labs says the same: group behavioral signals; a single glance or tab switch is a false positive risk.
Kamdar, quoted in the SHRM Labs piece, points to three practical tells: strong assessment scores that fall apart under AI aware follow up, assessments finished far faster than the designed window, and a drop when the work shifts from multiple choice to scenario application. A genuine high performer can explain. A skillfisher cannot.
Do not auto fail. Honest engineers look at notes, think, and use a second monitor for work. Document who reviewed a flag and why.
How do we prove the person on the call is the person we will badge?
That is the identity lane. Do not fold it into “ChatGPT interview cheating.”
Public reporting on biometrics, government ID at the start of video interviews, and outlet checks on Zoom comes from Bloomberg coverage reprinted by the Los Angeles Times. Gartner’s forecast about fake profiles by 2028 sits in that same conversation. Prudence Pitter, commenting on the LinkedIn News story, described a conference example where glasses did not move with the head, and warned employers to protect integrity without treating every legitimate candidate like a suspect.
North Korean remote IT worker schemes appear in WSJ social framing and in DOJ related reporting summarized by Bloomberg and LAT. That is a sanctions and security issue with its own documentation. It is not evidence that a typical plant engineer applicant is a cheat. Keep it out of your everyday “AI slop” talking points unless the role is remote IT with production access and your security team owns the process.
Put identity verification early for high risk remote seats. Checkr found only 19 percent of managers are extremely confident they would catch a fraudulent applicant. Gartner’s Jamie Kohn warns that candidate fraud can create cybersecurity risks beyond a bad hire.
What should engineering and IT hiring managers change this quarter?
Pick three roles that hurt when the hire is fake: a remote engineer with repo access, a controls lead, a senior IT admin. Rewrite those loops first.
Write a one page interview policy per role: which steps ban AI, which allow it, what identity proof you need, and what happens when a flag appears. Share it with every interviewer.
Replace one static coding gate with a live mutation. Change a constraint. Ask for a test. Ask what they would open first if the service page is down at 3 a.m.
Add a short supervised or on site coding block for remote critical seats. Keep lighter screens where the downside of a weak hire is smaller.
Train interviewers for five minutes on follow up. “What specifically did you do versus the team?” “What happened next?” “Walk me through the first production incident on that system.”
Point application screening elsewhere. We already covered integrity tooling and application red flags on AI integrity tools and AI application checkers. This page stays on the live interview.
What should you put in the candidate invite?
Will AI be allowed?
Yes on section A. No on section B. Say it in one sentence. Gartner also recommends defining acceptable use up front.
What will we ask you to show?
Camera, workspace pan if requested, hands visible for unaided prompts, ID for identity checks on high risk seats.
What happens if we see a flag?
Human review. Chance to explain. Not a silent ghost.
Where will the coding happen?
Your machine, our machine, or on site. If it is on site, say so early.
Short answers hiring managers keep asking
What is AI interview cheating?
Unauthorized live help during a screen: ChatGPT interview cheating, overlays, second devices, or a person feeding lines. Not the same as a polished resume.
How do candidates using AI in interviews usually show up?
Fluent generic answers, thin personal detail, code that does not survive a mid call change, and timing that ignores difficulty.
How to prevent AI cheating in interviews without a circus?
Clear rules, live mutable tasks, better follow ups, identity checks on high risk seats, and early in person work when access is critical.
How to detect AI in interviews without auto rejecting good people?
Layer timing, explanation, and tool signals. Review evidence. Let a human decide. Do not treat one tab switch as proof.
Should every engineering seat go on site?
No. Use supervised or on site coding where remote fraud would hurt production. Keep lighter screens for roles that do not need that cost.
Is this the same as using AI to prep?
No. Prep with a model is common. Undisclosed real time outsourcing of the thinking is what you are trying to catch. Identity fraud is a third lane.
If the last three “perfect” remote remotes fell apart after start, fix the interview design before you buy another monitoring logo. Apollo Technical can help you staff engineering and IT seats. The integrity rules still belong to your hiring managers.
