Aptitude tests were originally pen and paper assessments where cheating was almost impossible. Now they are administered online where candidates have access to very powerful AI tools. The question isn't whether candidates can cheat them - it's how easily, and how many are.
The methods range from low-tech tactics that have existed for years to AI-assisted approaches that require almost no skill or effort. Both are worth understanding, because they expose different weaknesses in how standard assessments are delivered.
Before AI candidates were finding ways around aptitude tests
Answer sharing sites
The most widespread method was question sharing. Candidates post test questions to Reddit, Discord, or dedicated groups after sitting an assessment. For high-volume graduate schemes where hundreds of candidates sit the same test across an open window of days or weeks, a circulating question bank can exist within hours of launch.
This was a huge problem with fixed question banks, which surprisingly some legacy test publishers still use or were slow to react to this threat.
Any serious test publisher now uses adaptive item banks where questions are drawn from large pools and have question difficulty ratings. This ensures candidates don't see the same set of questions.
Getting someone else to complete it
For years candidates have been seeking assistance from a more capable friend, family member, or a paid service.
Without verification tests or cheat measures in place, a candidate with poor cognitive ability could quite easily clear the assessment screen and make it to the interview stage.
Using fake details to access the test twice
For quite some time, candidates were able to apply for roles with fake details, sit the assessment and take note of all the questions, then reapply with their actual details and ace the test.
Again, adaptive item banks prevents this from being an issue. But for any test publisher using fixed item banks, all of which you should be avoiding using, this is still a major vulnerability.
Adaptive item banks have fixed most of the traditional cheating methods used by candidates”
The real threat is the advancement of AI capabilities at cheating tests
AI has removed the need for the traditional cheating methods, by providing a free tool that is able to score in the very top percentiles. The barrier to entry is now very low for cheating assessments, but the good news is this can be prevented, as we will discuss at the end.
Copy and paste
This requires no technical skill. A candidate screenshots each question, pastes it into ChatGPT or an equivalent tool, and gets the correct answer in seconds. It works reliably on numerical and verbal reasoning. The whole process takes a few seconds per question and is extremely accurate.
Using a second device
Instead of using the same device, the candidate photographs their screen with a phone and submits the question to an AI chatbot on that separate device. Without active webcam monitoring (and sometimes even with it) this is essentially undetectable.
More recently, AI apps with live camera features mean the candidate doesn't even need to type the question. They hold their phone up to the screen and talk through the assessment in real time with an AI.
Numerical and verbal test types are more vulnerable than others
ChatGPT, and other AI tools, are extremely effective at answering numerical and verbal reasoning questions. A simple screenshot of the question and the AI will be able to quickly and reliably score highly across a set of questions.
Situational judgement tests (SJTs) are widely assumed to be cheat-resistant because there's no single correct answer, however that's not correct. Candidates can feed an AI the job description, the organisation's stated competencies, the role's seniority level, and any publicly available information about the company's values - then ask it to answer the SJT on that basis.
Most SJT scoring keys are reverse-engineerable from information candidates already have access to. A well-prompted AI can identify the preferred response to a competency-based scenario with reasonable accuracy.
Inductive reasoning (also known as logical reasoning, abstract reasoning) has stronger resistance to AI. Due to the questions being based on visual patterns, sequences, and shape manipulations, AI struggles with these image-based items. Someone using AI to answer their inductive reasoning questions will not be able to instantly score in the higher percentiles like they would with numerical or verbal formats.
Conclusion and next steps
Cheating on aptitude tests is real and can be fairly easy, if using the wrong test publisher, especially now with AI.
The answer isn't to drop aptitude testing. Cognitive ability is one of the strongest predictors of job performance, and removing it from your sift would cost you more than cheating does.
Instead it's important to know how to prevent candidates from cheating on aptitude tests, as there are many methods which provide very strong AI-resistance, allowing you to assess your candidates with the assurance that you are testing their genuine ability - not whether they chose to use AI or not.
Ben Schwencke
Chief psychologist at Test Partnership. MSc in Organisational Psychology with over ten years experience in psychometric testing.
