We Want Better
ISTQB CT-AI v2.0: What Changed in the AI Testing Certification

ISTQB CT-AI v2.0: What Changed in the AI Testing Certification

November 02, 2026 3 min read

ISTQB's CT-AI v2.0 centers on testing AI-based systems, adds coverage of LLMs and red teaming, and changes how testers should prepare before version 1.0 retires in 2027.


ISTQB's Certified Tester AI Testing certification, or CT-AI, is in the middle of a major transition. Version 2.0 replaces the original syllabus and focuses on testing products that contain artificial intelligence, including machine learning models, data pipelines, and large language model features.

That does not mean using ChatGPT or another generative AI tool to write test cases. ISTQB places that work under its separate CT-GenAI certification. CT-AI v2.0 is for testers evaluating the behavior, quality, safety, data, and performance of AI-based systems. ([istqb.org](https://istqb.org/help/ct-ai-v2/))

What changed in CT-AI v2.0

The syllabus drops from 11 chapters to seven and follows the machine learning lifecycle more closely. Its sections cover AI fundamentals, quality characteristics for AI systems, machine learning basics, testing AI-based systems, input data testing, model testing, and machine learning development testing. ([istqb.org](https://istqb.org/help/ct-ai-v2/))

Generative AI and large language models now have dedicated coverage. Candidates will encounter exploratory testing for generative AI, test-oracle problems, red teaming, adversarial testing, metamorphic testing, model drift, overfitting, and back-to-back testing. Those topics connect the certification more directly to the issues teams face when AI features reach production.

ISTQB has added hands-on exercises to the accredited training structure as well. Courses can include building a simple machine learning model, preparing data, evaluating performance metrics, implementing a perceptron, and exploring an LLM. The exam remains multiple choice, so candidates will not write code, but they should be able to interpret basic machine learning metrics. ([istqb.com](https://www.istqb.com/certifications/ct-ai/))

The exam format and pass mark

CT-AI v2.0 has 40 multiple-choice questions and a standard 60-minute duration. The exam carries 44 points, with 29 required to pass. The retiring version 1.0 exam also has 40 questions, but it carries 47 points and requires 31. Candidates should use sample exams for the correct syllabus version instead of assuming older preparation materials reflect the current scoring model. ([istqb.com](https://www.istqb.com/certifications/ct-ai/))

Model testing is the largest section, with nine of the 40 questions. Machine learning and testing AI-based systems each account for seven, while input data testing accounts for six. The weighting points to a clear study plan: learn the fundamentals, then spend serious time on evaluating, challenging, monitoring, and comparing models.

What testers need to know about the transition

CT-AI version 1.0 remains available for a limited period. English-language training and examinations, including retakes, continue until April 21, 2027. Other languages remain available until October 21, 2027. After those dates, training and exams will use version 2.0. ([istqb.org](https://istqb.org/help/ct-ai-v2/))

Version 1.0 certificates remain valid, and ISTQB does not require certificate holders to retake the exam. There is no bridging exam either. Anyone seeking the version 2.0 credential must sit a separate certification exam; current version 1.0 holders can decide whether the revised scope fits their role.

Should you choose CT-AI or CT-GenAI?

The right choice depends on the work you expect to do. CT-AI v2.0 fits testers working on AI-enabled products, training and evaluation data, bias or drift investigations, model quality, and LLM behavior. CT-GenAI is for professionals using generative AI across testing work, including analysis, test design, automation, reporting, and maintenance. ([istqb.com](https://www.istqb.com/certifications/ct-ai/))

For QA leaders, the change signals that AI testing is becoming a distinct competency rather than a minor extension of conventional automation. For individual testers, the next step is straightforward: check the syllabus version before paying for a course, downloading sample questions, or booking an exam. The certification name has not changed much, but the material being assessed has.

Looking for tools to match this read?

Jump into our independent side-by-side comparisons.

Compare Tools