“"The practice exams made AI testing concepts much easier to understand."”
Evelyn Hart
AI Test Engineer
The ISTQB Certified Tester AI Testing (CT-AI) v2.0 exam checks whether you can test systems that learn from data, including machine learning models, neural networks, generative AI and large language models. It is a specialist certification from the International Software Testing Qualifications Board, built for testers, test analysts, test managers, developers and data professionals who work on AI products. You must already hold the Foundation Level certificate (CTFL) to sit it. This page covers the format, syllabus, registration and validity.
CT-AI v2.0 focuses on testing AI systems, covering machine learning and generative AI such as large language models, and it deals with traits like probabilistic behavior, nondeterminism and dependence on data. The point is not to teach you to build models. It is to show that you can plan and run tests when the same input may not always produce the same output, and when quality depends heavily on the data behind a model.
ISTQB lists several outcomes for certified holders. They include understanding the AI quality characteristics defined in ISO/IEC 25059, calculating and interpreting performance metrics for machine learning models, recognizing the two test levels specific to machine learning systems, and designing and executing test cases for those systems.
The Foundation Level certificate (CTFL) is a prerequisite. Beyond that, ISTQB says no prior AI or machine learning expertise is required, since the syllabus introduces the key concepts before moving to testing approaches, although some familiarity with testing or development is recommended. The official audience includes testers, test managers, test consultants, data analysts, data scientists, developers of AI systems and user acceptance testers. Project managers, quality managers and business analysts can also take it for a structured overview.
The exam has 40 questions and 44 total points, with 29 points needed to pass and 60 minutes allowed, plus 25% extra time for a non native language. The total points exceed the question count because some questions carry more than one point. That makes a point total, not a simple count of correct answers, the figure you track while practicing. For candidates in the extra time category, the allowance brings the exam to 75 minutes.
ISTQB offers a sample exam (questions and answers) alongside the syllabus on its CT-AI page. Doing it under timed conditions is a sensible way to learn the question style before booking.
Version 2.0 reduced the syllabus from 11 chapters to 7 and reorganized it around the machine learning lifecycle. The seven chapters, as listed by ISTQB, are:
Introduction to Artificial Intelligence: the current state of AI, including generative AI.
Quality Characteristics for AI Based Systems: the AI quality characteristics and how to set acceptance criteria.
Machine Learning: core concepts, data for machine learning, performance metrics for classification, and neural networks.
Testing AI Based Systems: testing approaches, testing generative AI and large language models, and the test levels for machine learning systems.
Input Data Testing for Machine Learning Systems: checking the data that trains and feeds a model.
Model Testing for Machine Learning Systems: evaluating the model itself.
Machine Learning Development Testing: testing across the development process.
The syllabus also includes practical exercises covering machine learning workflows, data preparation, model evaluation and testing approaches. Official chapter weightings should be taken from the syllabus PDF on the ISTQB site rather than from third party summaries.
Version 2.0 removed material on using AI to support testing and added dedicated coverage of generative AI and large language models, including exploratory testing and red teaming. People interested in using AI as a testing tool are pointed to the separate CT-GenAI certification. Version 1.0 certificates remain valid, there is no bridging exam, and version 2.0 is taken as a separate exam. Version 1.0 exams and training stay available until April 21, 2027 in English and October 21, 2027 in other languages.
ISTQB does not sell exams directly. Exams are delivered by ISTQB exam providers, and ISTQB’s site includes a finder for providers offering the CT-AI v2.0 exam. Providers differ by country and language, and they set their own fees, scheduling options and retake terms. Compare these before you book, and follow the provider’s own instructions for identification and exam delivery. ISTQB recommends accredited training but confirms that studying on your own with the syllabus and sample exam is also an option Start a free Implementing Cisco SD-WAN Solutions (ENSDWI) practice test.
CT-AI sits in a narrow niche: it is aimed at people who already test software and now face products with machine learning or language model components. A tester on such a team needs to talk about training data quality, classification metrics and nondeterministic behavior in the same terms as the data scientists. The certification gives that shared vocabulary a recognized, standards aligned form. It also builds on CTFL, so it extends a credential many testers already hold. For test managers and consultants, it signals the ability to shape a test strategy for a machine learning system, which is a different job from planning tests for conventional applications.
Reading the syllabus tells you what is covered, but timed questions show where your understanding is thin, especially on calculating metrics and applying techniques to scenarios. AllExamQuestions.com’s free CT-AI practice tests can help you get used to the question style and spot weak chapters. They work best alongside the official syllabus and the ISTQB sample exam, not as a replacement for them.
Last updated on Oct, 31 2026