When Candidates Can Fake Skills in Real Time, Assessment Design Has to Catch Up

Candidates are leveraging artificial intelligence in various capacities, and in 2026, that practice has become an expected part of the hiring funnel. Unfortunately, some hiring managers might be unaware of just how deep this challenge persists and the hidden impact this has on new-hire evaluation. Applicants are employing agentic AI, generative AI (GenAI), chatbots, coding assistants, and in-call response generators to circumvent multiple phases of the traditional hiring process. The industry term for this is “skillfishing,” and the problem is not specific to any one industry, company size, or geographical location. 

In an analysis of nearly 20,000 interviews conducted between July 2025 and January 2026, AI interview platform Fabric found that 38.5% of candidates showed signs of cheating, with technical positions showing the highest number of detected cheaters at 48%.

Every hire based on falsified skill signals creates invisible liabilities for organizations: money spent, teams reshaped, impacts on project velocity, and the time required to get the new hire ramped. All resting on skills that the hiring process never actually verified. 

Namrata Kamdar, co-founder and COO of Testlify and a member of the 2026 SHRM Labs WorkplaceTech Accelerator cohort, argues that the real fix goes beyond stricter surveillance —  it requires rebuilding the assessment itself.

What Skillfishing Actually Looks Like

The simplest way to understand the problem, Kamdar says, is this: "Skillfishing is like someone saying they're a great swimmer because they've watched tutorials."

Previously, overt signs like a second monitor, a split screen, or an occasional glance down or away from the camera provided telltale indicators of cheating. However, these techniques could be easily detected. New invisible overlay tools let candidates view AI-generated answers directly on their screen (these are undetectable in shared-screen interviews without special tooling). To the interviewer, it looks like the candidate is working; in reality, they’re sending queries, reading responses, or otherwise artificially tailoring their interview. A 2026 Fabric report found that 59% of hiring managers now suspect candidates of using AI tools during live assessments.

Testlify keeps a close eye on the results in its own platform data. Kamdar identifies three common indicators. First, a performance discrepancy occurs when candidates excel on assessments, but then perform poorly on AI-assisted follow-up questions about the answers provided in the assessments. Second, there's the issue of shortened response times; for example, assessments that are typically 25- to 30-minute exercises are completed in under 10 minutes. Lastly, there tends to be a drop in accuracy for scenario-based assessments compared to multiple-choice questions due to scenarios involving the application of knowledge, which is more difficult to paste from a chatbot.

"A genuine high performer can explain their thinking," Kamdar said. "A skillfisher can't."

Why Filling Fast Makes It Worse

The skillfishing problem is landing just when hiring teams are feeling the most pressure to fill roles quickly, and as the use of AI in applications is skyrocketing. According to the 2025 SHRM Benchmarking Survey, the average cost-per-hire and average time-to-hire have increased over the last three years, a time period during which GenAI has become widely adopted across the funnel. 

Speed creates its own bias. A CareerBuilder survey found that 30% of employers who made bad hires said they felt pressured to fill the role quickly, and another 33% said they thought the candidate could pick up missing skills on the job. In both cases, the hiring team accepted clean-looking data at face value.

"Speed without signal is just expensive guesswork," Kamdar said. "When you rush a hire without validated skills data, you're not solving a talent problem; you're postponing it."

The more useful metric, she argues, isn't day-to-fill. It's whether the person is still in the role and performing, a year later. SHRM research puts the total replacement cost of a bad hire at 50% to 200% of the employee's salary, depending on seniority, and Leadership IQ's long-running tracking data finds that 46% of new hires fail within 18 months. A process optimized for speed routinely spends that money twice.

The Post-Hire Cost Is Compounded

The impact of a skillfisher on a business is not immediate; it takes months before the damage becomes apparent. For the first two to three weeks, Kamdar says, managers tend to notice that a new hire cannot independently tackle an unfamiliar problem, deflects with questions that should be clarification questions, or takes much longer than necessary to complete work for the job they were hired to perform. By then, the issue escalates into a formal performance discussion, which typically takes place 60 to 90 days into the role — well past the point where the manager, team, or project timeline can easily recover.

One Testlify customer described this pattern: a candidate who had passed a previous testing process at their organization began to drift on the job. The employee required about a month to be onboarded and another two months of coaching before the company ultimately let them go.

"It wasn't a typical bad hire," Kamdar said. "The signals they relied on were never real, which made it much harder to catch early or fix later."

Design, Not Surveillance

When tests are leaking, the natural response is to add more monitoring. Kamdar pushes back on that. Monitoring can deter cheating, but it’s reactive in the fact that it controls behavior without fixing the root problem: Standardized tests are easy to game. It also comes at a cost to candidate experience and increases false positives that penalize honest applicants.

Assessment design remains the real fix. Because AI tools rely on standardization, tests that are static and predictable present vulnerabilities; conversely, assessments that are adaptive, scenario-based, and linked to actual job tasks are extremely difficult for these tools to cheat on because they necessitate judgment in context.

In practice, this looks like three shifts:

• Moving to multiple layers of signals, rather than relying on a single layer. Each result should provide interview questions tailored to the answers the candidate has given. This means recruiters are probing for reasoning, rather than simply accepting a score.
• Replacing “off-the-shelf” tests with role-specific tests designed to emulate tasks candidates would have to perform. This brings the assessment closer to the reality of the work.
• Grouping behavioral signals (e.g., tab switching, completion time, answer changes, and mouse movements) rather than treating each as an indicator of cheating. This reduces false positives.

"The best assessments are the ones where it's actually easier to just know the answer than to try and game the system," Kamdar said.

What HR Can Do Now

There's a paradox at the center of skillfishing: The same AI tools that are corroding old hiring signals are also what most companies are now hiring people to use.

"Real skills have texture, gaps, and a learning curve," Kamdar said.

The line between AI-assisted prep and AI-assisted cheating is genuinely blurring, Kamdar acknowledges. Using AI to learn a concept or build a skill is reasonable, and trying to stop it is a losing battle. Outsourcing the cognitive work itself in real time is something else.

The real question isn't whether candidates should be using AI — they will, and they'll use it on the job, too. The question is whether your process can tell the difference between someone who knows how to think with these tools and someone who's relying on them to fake the thinking entirely.

You can see this difference in three shifts: 1) Focus on quality-of-hire metrics (such as 12-month retention, manager ratings, and time-to-productivity), rather than time-to-fill; 2) adjust your assessments so they assess applied reasoning rather than pure recall; and 3) incorporate more interview rounds so that candidates need to explain how they arrived at a solution, not just that they found the correct answer.

"You're not hiring a score," Kamdar said. "You're hiring how someone approaches a problem they've never seen before."

To learn more about Testlify, join us at the SHRM 26 Annual Conference & Expo for the session: Right Hire, Right Growth: A Practitioner's Unfiltered Look at AI Across Hiring and Retention.