Learn with It, Not from It: Using AI in IT Certification Prep

Generative AI has changed how people study for IT certifications. A learner who gets stuck on subnetting at 11pm can get a clear explanation in seconds. Practice questions, concept breakdowns, terminology comparisons — all of it is available on demand, free, and patient. Five years ago, none of that was true.

At ACI Learning, we train thousands of IT and cybersecurity professionals every year, and we pay close attention to how preparation methods affect outcomes. AI tools can have a legit role in that preparation. But they also have specific, well-documented limits, and understanding those limits is essential for serious learners.

What AI Gets Right and Who It Gets It Right For

AI tools work differently depending on what a learner already knows, and most discussions of AI in certification prep ignore that distinction entirely.

For a network admin studying for CCNP, or a security analyst adding a cloud certification, AI is a capable study companion. They can evaluate its explanations critically, push back when something seems off, and use it to fill specific gaps in an existing knowledge base. The tool produces useful output because the person using it knows enough to assess it.

For a beginner, the situation is different. Without an existing framework, there's no reliable way to tell a correct explanation from a plausible-sounding one, a current exam objective from a stale one, or a complete answer from one missing a nuance that determines whether you pass. The AI presents all of these with equal confidence.

That distinction has to be called out before anything else in this article, because the problems below are not evenly distributed. Experienced professionals using AI as a supplement will get a massively different outcome than beginners using it as a foundation.

The Accuracy Problem Beginners Don't Know They Have

IT and cybersecurity certifications test precise, vendor-aligned knowledge. CompTIA exam objectives change with each version release. Cisco implementation details matter at a granular level. AWS service behaviors are specific and frequently updated.

AI tools are trained on broad internet data, not on the specific objectives of your certification exam. While you can ask them to tailor their answers to a particular exam, they generally have no reliable way to determine whether a topic is actually covered. As a result, they’ll likely frame information as exam-relevant simply because you've asked them to do so, not because they possess authoritative knowledge of the exam itself. An explanation can be technically accurate in general while missing the specific nuance that separates a right answer from a plausible distractor on the current version of the exam.

For a beginner, this AI failing is completely invisible until the exam score reveals it. There's no feedback mechanism telling them the explanation they just read was correct for 2021 but doesn't reflect how the question is tested now.

ACI Learning's curriculum is built directly against current exam objectives and updated when certification bodies revise their requirements. For someone building their understanding from the ground up, that alignment is the best foundation to lay in the journey to becoming a successful IT professional.

Getting Answers Is Not the Same as Learning

A June 2026 study from researchers at the University of California, Irvine and McGraw Hill examined more than 3.2 million student interactions with ALEKS, an adaptive math learning platform used by more than four million students annually. The dataset covered more than a decade of student behavior, spanning the period before and after ChatGPT's public release in late 2022. (Rismanchian et al., "Faster Completion, Less Learning," arXiv preprint, June 2026.)

How the study worked

The researchers compared two types of problems based on how easily a student could outsource them to a chatbot:

• Word problems — text that can be copied directly into an AI tool for an instant answer
• Graphing problems — tasks requiring the student to construct a graph inside the platform, which can't be delegated to AI

After ChatGPT's release, the two problem types diverged sharply.

What the data showed

By the end of the study period, students were spending 31 percent less time on word problems among high school students, and 27 percent less among college students. Time spent on graphing problems was flat.

On proctored placement tests, performance on word problems dropped from approximately 80 percent correct to approximately 60 percent correct , a roughly 25 percent reduction in the odds of a correct answer. Graphing problem performance was unchanged.

The researchers call this "cognitive surrender": when getting an answer is easy, learners take the shortcut, and the result is an answer obtained without the cognitive work that produces retention. As lead researcher Sina Rismanchian put it: "If ChatGPT does it for you, then you haven't learned it."

A note on scope:

The study is a preprint and has not yet undergone peer review. It focuses on math education, not IT certification. The connection we are making here is inferential, but the underlying mechanism applies directly. IT certification exams test applied reasoning under timed conditions. Getting an answer from an AI is a different cognitive event than working through the problem yourself, and that difference shows up on the actual exam.

This is what ACI Skill Labs were built for:

ACI Skill Labs use virtual machines and sandbox environments, not simulations or click-through walkthroughs. When a learner configures a firewall rule, troubleshoots a VPN misconfiguration, or deploys a cloud resource in a Skill Lab, they are doing the cognitive reasoning work in a live environment. That's where retention comes from.

An AI can explain what happens when you misconfigure an ACL. A Skill Lab puts you in front of one and asks you to find the problem. The ALEKS research gives us a precise way to understand why one of those experiences produces durable understanding and the other doesn't.

What Certification Exams Are Actually Testing

CompTIA redesigned its performance-based question formats specifically to close the gap between knowing and doing. Cisco's practical exams require candidates to configure and troubleshoot live environments. AWS scenario questions present architecture decisions with the kinds of constraints that come up in production environments.

Scenario reasoning (CompTIA Security+) A performance-based item places candidates inside a simulated network environment and asks them to identify misconfigured firewall rules, explain the problem, and suggest a remediation. There is no multiple-choice scaffold. The question tests whether the candidate can work through a diagnostic process, which has to be built through repeated practice, and cannot be absorbed from a reading.

Hands-on configuration (Cisco CCNA) A candidate who understands OSPF conceptually but has never actually configured it, watched it fail, and traced a misconfiguration is not prepared for the exam's hands-on component, and they’re not prepared for the job that follows.

Applied judgment (AWS) Scenario questions present architecture tradeoffs under defined constraints. They test how candidates reason through a problem, not what they have memorized.

AI can explain any of these topics thoroughly. What it cannot do is give a learner the experience of having worked through the problem themselves. ACI Skill Labs exist specifically for that gap with industry-leading environments built at the difficulty and specificity that certification exams and employers require.

Free Learning Has a Completion Problem AI Doesn't Solve

A meta-analysis of 221 open online courses found a median completion rate of 12.6 percent — fewer than one in seven learners who enroll actually finishes. (Jordan, International Review of Research in Open and Distributed Learning, 2015.) A study published in Science by MIT and Harvard researchers examined 12.67 million course registrations across edX and found completion rates between 3 and 6 percent for free courses, with no meaningful improvement over six years of platform development. (Reich & Ruipérez-Valiente, Science, 2019.)

Free, open-enrollment courses carry no stakes worth protecting. There's no financial investment to protect, no credential meaningfully on the line, and no reliable signal that the content is mapped to what the exam tests. When a learner steps away for a week, there's nothing pulling them back.

Payment and credential stakes change those conditions in measurable ways. Coursera's paid certificate programs reach approximately 55 percent completion, four to five times the rate of free courses on the same platform. edX's enterprise learners, enrolled through employer-sponsored programs, complete at around 54 percent.

At ACI Learning, on-demand video courses mapped to current exam objectives, paired with practice exams and ACI Skill Labs, drive an 80-plus percent course completion rate, well above even the paid-platform baseline. Learners finish because the content is credible, the path is clear, and the labs make the work engaging enough to sustain momentum.

What AI Can't Replace

The difference between AI-only preparation and structured training can be a little different depending on your perspective.

For individual learners:

For businesses investing in workforce development:

For academic institutions and program directors:

Where AI Makes Itself Useful

AI tools have a productive role in certification prep. That role is specific.

Where AI helps:

• Checking understanding after working through a module
• Getting a second explanation of something that didn't click in the primary course
• Generating additional practice questions in an area that needs more repetition
• Exploring a related topic the structured curriculum doesn't cover in depth

Where it falls short:

• Providing answers to problems a learner hasn't attempted themselves
• Replacing hands-on lab work with explanations of what the lab would have taught
• Validating whether understanding is accurate enough to pass a current exam
• Maintaining momentum when motivation drops

At ACI Learning, we build programs so learners can use AI as a complement to credible, objective-aligned training. The video courses, labs, and practice exams cover what AI can't. AI handles the parts of studying where it works.

What to Look For in a Training Program

Individual learners, L&D managers, and academic program directors evaluating training options tend to find that the same elements separate programs that produce certified professionals from those that don't.

Content aligned to current exam objectives. Specifically mapped to what's tested on the current version of the exam, and maintained as certification bodies update their requirements. Generic explanations, whether from AI or a poorly maintained course, don't meet this bar.

Industry-leading hands-on labs. For networking, security, cloud, and systems certifications, lab experience is how candidates build the diagnostic reasoning that scenario-based exam items test. ACI Skill Labs use virtual machines and sandbox environments, because the only way to learn how to do something is to do it.

Practice exams under realistic conditions. Timed and structured to surface gaps before exam day, not question banks worked through casually.

A credible program with a credential at the end. Learners enrolled in a paid, objective-aligned program with a credential at the end have a concrete reason to finish. The research on completion rates consistently reflects that.