The ISTQB Certified Tester AI Testing (CT-AI) v2.0 certification is the updated specialist certification for professionals involved in testing AI-based systems. Released on April 17, 2026, this version replaces CT-AI v1.0 with a sharper focus on testing machine learning systems and generative AI, while removing the “using AI for testing” content that now lives in the separate CT-GenAI certification. If you work with AI-based systems in any testing, quality, or development role, this is the certification that validates your ability to assess their quality, reliability, and safety.
What is the ISTQB CT-AI v2.0 Certification?
The CT-AI v2.0 certification provides structured, practical knowledge for testing AI-based systems, with particular emphasis on machine learning systems (MLS) and generative AI systems such as large language models (LLMs). It covers the full lifecycle of AI testing, from evaluating input data quality and testing ML models to validating system-level behavior in production environments.
The v2.0 syllabus is a major rewrite, not a minor update. It is built around how AI systems are actually developed and tested in practice: data goes in, models get trained and evaluated, and systems get deployed and monitored. The syllabus follows this lifecycle, teaching testers what to validate at each stage and which techniques to apply.
Key areas covered include AI-specific quality characteristics defined by ISO/IEC 25059, ML functional performance metrics such as accuracy, precision, recall, and F1-score, testing approaches for generative AI including red teaming, and specialized test levels for input data and ML models. The certification also addresses regulatory frameworks like the EU AI Act and their impact on AI system testing.
What Changed from v1.0 to v2.0?
This is not a revision you can afford to ignore. The structural and content changes are significant:
- Generative AI coverage added. The v2.0 syllabus includes dedicated sections on testing GenAI and LLMs, covering topics like red teaming, prompt-based evaluation, and the unique challenges of non-deterministic text, image, video, and audio outputs.
- “Using AI for testing” removed entirely. The v1.0 chapter on using AI to improve testing activities (test case generation, defect prediction, regression optimization) has been dropped from this syllabus. That content is now covered by the separate ISTQB CT-GenAI certification.
- Two ML-specific test levels defined. The syllabus introduces input data testing (Chapter 5) and ML model testing (Chapter 6) as distinct test levels, each with its own risks, techniques, and learning objectives.
- ML development testing added. A new Chapter 7 covers testing risks during MLS development and deployment, including shadow testing and canary testing.
- Pretrained models, fine-tuning, and RAG covered. Reflecting modern AI practice, the syllabus now addresses pretrained models, fine-tuning workflows, and retrieval-augmented generation.
- Training time reduced. Minimum required instruction time dropped from four days (26 hours) to three days (19.5 hours), but the material is denser and more practice-oriented.
- Test environments section removed. Test environment considerations for AI systems are no longer part of the examinable content.
Who Should Consider the CT-AI v2.0 Certification?
The CT-AI v2.0 certification is designed for a broad range of professionals working with AI-based systems:
- Testers, Test Analysts, and Test Engineers who need to validate the quality and reliability of AI-based systems, especially those built on machine learning or generative AI.
- Data Analysts and Data Scientists who want to understand how the data they prepare and the models they build will be tested and validated.
- Test Managers and Test Consultants responsible for defining test strategies for projects that include AI components.
- Software Developers building AI-based systems who need to understand testability requirements and quality characteristics.
- Quality Managers, Project Managers, and Business Analysts who need a foundational understanding of what it takes to assure the quality of AI-based systems.
- Operations Team Members involved in deploying and monitoring ML systems in production.
This certification is also suitable for professionals who already hold other ISTQB certifications, such as CTFL, CTAL, or CT-GenAI, and want to deepen their expertise in the testing side of AI.
CT-AI v2.0 Syllabus Structure
The CT-AI v2.0 syllabus is organized into seven chapters with a total of 19.5 hours of instruction time:
- Chapter 1: Introduction to Artificial Intelligence (120 minutes) – Covers the fundamentals of AI, including the differences between AI-based and conventional systems, the spectrum from narrow AI to general AI to super AI, types of AI technologies, generative AI concepts, hardware and hosting options for ML systems, ML development frameworks, and the impact of regulations and standards (including the EU AI Act) on AI development and testing.
- Chapter 2: Quality Characteristics for AI-Based Systems (45 minutes) – Focuses on the AI-specific quality characteristics defined in ISO/IEC 25059, such as AI robustness, functional adaptability, user controllability, and intervenability. Also covers safety considerations for AI in critical systems and how to define appropriate acceptance criteria for AI-based solutions.
- Chapter 3: Machine Learning (375 minutes) – The largest chapter, covering supervised, unsupervised, and reinforcement learning; the ML development workflow; pretrained models, fine-tuning, and retrieval-augmented generation (RAG); data preparation activities; training, validation, and test dataset roles; ML functional performance metrics including the confusion matrix, accuracy, precision, recall, and F1-score; and neural network structure, working, and coverage measures. This chapter includes multiple hands-on exercises.
- Chapter 4: Testing AI-Based Systems (195 minutes) – Covers the unique testing challenges of AI, including locked vs. adaptive system testability, the need for statistical testing approaches, test oracle problems, testing generative AI and LLMs, red teaming for GenAI systems, test levels for MLS (input data testing, model testing, component testing, system testing, and acceptance testing), and risk-based testing of machine learning systems.
- Chapter 5: Input Data Testing for Machine Learning Systems (180 minutes) – Dedicated to the first ML-specific test level. Covers input data risks and mitigations, testing for bias (including disparate impact analysis), data pipeline testing, testing for data representativeness, dataset constraint testing, and label correctness testing (including multiple annotation techniques).
- Chapter 6: Model Testing for Machine Learning Systems (225 minutes) – Dedicated to the second ML-specific test level. Covers ML model risks and mitigations, ML model documentation review, ML functional performance testing for probabilistic systems, adversarial testing, metamorphic testing (including deriving and applying metamorphic relations), drift testing (data drift and concept drift), testing for overfitting and underfitting, A/B testing, and back-to-back testing.
- Chapter 7: Machine Learning Development Testing (30 minutes) – Covers ML development risks and mitigations, and ML system deployment testing including shadow testing and canary testing.
Business Outcomes of CT-AI v2.0
A certified professional holding the CT-AI v2.0 credential should be able to:
- BO1: Understand the current state of AI, including generative AI.
- BO2: Experience the implementation and testing of machine learning models.
- BO3: Understand the working and testing of simple neural networks.
- BO4: Understand the specific AI quality characteristics defined by ISO/IEC 25059.
- BO5: Calculate and interpret ML functional performance metrics for machine learning models.
- BO6: Recognize the scope and importance of the two test levels specific to the testing of machine learning systems.
- BO7: Contribute to the development of an effective test strategy for a machine learning system.
- BO8: Design and execute test cases for machine learning systems.
Exam Structure and Preparation
The CT-AI v2.0 exam is a multiple-choice exam with the following structure:
- Number of questions: 40
- Exam duration: 120 minutes (150 minutes for non-native language speakers, with the standard 25% extension)
- Pass score: At least 65% of total points
- Prerequisite: ISTQB Certified Tester Foundation Level (CTFL) certification
- No negative marking: Wrong answers do not subtract points, so candidates should answer every question
The exam includes a mix of 1-point and 2-point questions. Questions worth 2 points typically correspond to K3 (Apply) level learning objectives that require candidates to apply techniques to given scenarios, such as calculating ML performance metrics, implementing red teaming, applying dataset constraint testing, or using metamorphic testing to derive test cases.
Questions are assessed at four cognitive levels:
- K1 (Remember): Recall of key terms and definitions related to AI, ML, and testing.
- K2 (Understand): Understanding and explaining AI concepts, testing techniques, quality characteristics, and their relationships.
- K3 (Apply): Applying learned techniques to testing scenarios, such as evaluating confusion matrices, performing red teaming on GenAI systems, applying dataset constraints, or deriving metamorphic test cases.
- K4 (Analyze): Analyzing testing scenarios to identify suitable approaches, risks, and strategies (note: K4 is defined in the syllabus but the v2.0 sample exam primarily uses K2 and K3 level questions).
ISTQB strongly recommends that candidates have at least six months of practical experience in software testing, data science, or software development before attempting the exam. Candidates can prepare through accredited training courses or self-study using the official syllabus.
CT-AI v2.0 vs CT-GenAI: Which Should You Take?
One of the most common questions from professionals planning their ISTQB certification path is whether to take CT-AI v2.0 or CT-GenAI first. Here is the key distinction:
- CT-AI v2.0 focuses on testing AI-based systems. It teaches you how to validate the quality, reliability, and safety of systems that use AI and ML, including generative AI systems. The focus is on the system under test.
- CT-GenAI focuses on using generative AI for testing. It teaches you how to apply tools like ChatGPT, Claude, and Copilot effectively in your testing work, including prompt engineering, test case generation with AI, and managing the risks of AI-assisted testing.
If your role involves testing AI-powered products or features, CT-AI v2.0 is the priority. If your role is about using GenAI tools to make your testing more efficient, CT-GenAI is the priority. Many professionals will benefit from both certifications over time.
Frequently Asked Questions (FAQ)
1. Does CT-AI v2.0 replace v1.0?
Yes. The CT-AI v2.0 certification fully replaces v1.0. Candidates should study the v2.0 syllabus for current and future exams. Existing v1.0 study materials are not sufficient for the v2.0 exam due to the extensive structural and content changes. See more difference between v1.0 and v2.0 here.
2. What is the prerequisite for the CT-AI v2.0 exam?
The main prerequisite is holding a valid ISTQB Certified Tester Foundation Level (CTFL) certification. ISTQB also strongly recommends at least six months of experience in software testing, data science, or software development.
3. How can I prepare for the CT-AI v2.0 exam?
Candidates can prepare by attending accredited training courses (minimum 19.5 hours of instruction) or through self-study using the official syllabus and sample exam. Hands-on practice with ML tools and frameworks is highly beneficial, as the syllabus includes multiple hands-on objectives.
4. Can I take the CT-AI v2.0 exam online?
Yes. The CT-AI exam is available through ISTQB-accredited exam providers, and many offer online proctored options alongside in-person testing at Pearson VUE centers and other locations.
5. How many questions are on the exam, and what is the pass score?
The exam has 40 multiple-choice questions. The pass score is at least 65% of the total available points. There is no negative marking.
6. Do I need programming or ML experience to pass the exam?
Programming expertise is not required, but a basic understanding of software development concepts is helpful. The syllabus introduces ML concepts from a testing perspective, so deep ML expertise is not assumed. However, the hands-on exercises in accredited training courses involve working with ML models and data, so some comfort with technical tools will help.
7. If I already hold CT-AI v1.0, do I need to recertify under v2.0?
Your existing CT-AI v1.0 certification remains valid. However, the v2.0 syllabus reflects the current state of AI testing practice, including generative AI coverage and updated test approaches. Professionals who want their certification to reflect current knowledge should consider taking the v2.0 exam.
Advance Your Career with CT-AI v2.0
The AI testing landscape has changed dramatically since the original CT-AI v1.0 was released. Generative AI systems, large language models, and increasingly complex ML pipelines are now part of everyday software delivery. The CT-AI v2.0 certification equips testers with the knowledge and techniques to validate these systems effectively, from assessing training data quality and detecting bias to testing model performance, managing drift in production, and applying red teaming to GenAI outputs.
As organizations continue to deploy AI across critical applications in healthcare, finance, transportation, and beyond, the demand for professionals who can test these systems competently and rigorously will only increase. CT-AI v2.0 certified professionals are well positioned to take on these roles and contribute to building trustworthy AI systems.
Recommended Study Materials for CT-AI v2.0
Coming soon. For v1.0 study material – visit this page. https://www.istqb.guru/ct-ai-study-guide/