Big news for testers: ISTQB CT – GenAI is live
ISTQB has released a new Specialist Level certification called Certified Tester Testing with Generative AI, or CT GenAI for short. It focuses on how testers can use large language models and other generative tools to plan, design, execute, and report tests with the right safeguards. If your team is already experimenting with prompts or you want a safe and structured way to bring AI into quality work, CT GenAI is a timely guide.
The arrival of CT GenAI does not replace Certified Tester AI Testing, known as CT AI. The two syllabi serve different but complementary goals:
- CT AI teaches you how to test AI based systems. That means models, data, fairness, explainability, robustness, and specialized techniques like metamorphic and adversarial testing.
- CT GenAI teaches you how to test with AI. That means prompt engineering for test design, safe use and evaluation of outputs, and the nuts and bolts of running LLM powered testing at scale, including RAG, agents, fine tuning, and LLMOps.
Put simply, CT AI is about assurance of AI. CT GenAI is about assurance when using AI inside the testing process.
What changed with CT GenAI
Over the last two years, testers have shifted from curiosity to daily use of LLMs. People rely on generative tools to clarify requirements, design test ideas, generate data, write drafts of test scripts, analyze logs, and summarize results. The upside is real, but so are the risks. Hallucinations, privacy leaks, and inconsistent answers can add hidden cost if you do not manage them.
CT GenAI acknowledges that reality. It provides a common vocabulary for prompt patterns, a way to measure and iterate on output quality, and a roadmap for adoption that balances speed with governance. You learn when to reach for RAG versus agents, how to keep sensitive data out of prompts, and how to put human review in the loop without grinding flow to a halt.
Short answer for busy readers
- Work with LLMs and want immediate productivity gains with clear guardrails? Start with CT GenAI.
- Build or assure products that include machine learning models? You need CT AI.
- Lead a test organization? Plan to do both, in that order, so your team gets quick wins while you grow deeper AI assurance skills.
Side by side overview
Prerequisite for both: ISTQB Foundation Level.
Teaching time: CT AI takes about twenty five hours across eleven chapters. CT GenAI takes about thirteen and a half hours across five chapters, with the largest block devoted to prompt engineering.
Cognitive levels: CT AI goes up to K4 Analyze. CT GenAI goes to K3 Apply. Both include hands on exercises at H0 to H2, but the exams themselves remain knowledge based.
Scope:
- CT AI covers data quality, functional metrics, neural network coverage, explainability, bias, non determinism, drift, and test environments. It also includes AI for testing at a conceptual level, such as defect prediction and UI testing with computer vision.
- CT GenAI goes deep on prompt engineering, evaluation and iteration, multimodal prompting for visual and text artifacts, risk management for hallucinations and privacy, LLM powered test infrastructure like RAG and agents, and the real world rollout of LLMs in a team.
Deep dive on differences
1. Foundations
CT AI builds fluency with the machine learning ecosystem. You learn the common model families, the data pipeline from labeling through validation, and the metrics that tell you whether a model is fit for purpose. The syllabus shows how to reason about trade offs such as precision and recall, and how to read coverage signals in neural networks. You also explore classical and modern tooling along with the hardware that makes training possible.
CT GenAI builds fluency with large language models. You learn how tokens, embeddings, and context windows affect behavior. You compare foundation models, instruction tuned models, and models optimized for reasoning. You practice with multimodal inputs, where a screenshot, a wireframe, or a log excerpt can be part of a single prompt that produces usable test ideas.
What this means in practice: If your team ships features that rely on ML predictions, the CT AI foundation will help you think clearly about quality risks inside the product. If your team writes and runs a lot of tests and wants smarter tooling, the CT GenAI foundation will help you put LLMs to work without losing control.
2. Data and metrics versus prompts and output checks
CT AI treats data as the first class citizen. You work with labeling quality, dataset splits, and contamination risks. You pick and defend relevant metrics for classification, regression, and clustering. You learn when aggregate numbers hide local failure and how to drill into slices that represent real users.
CT GenAI treats the prompt as the program. You learn a simple scaffold for prompts that keeps the model on task and within bounds. You measure output quality, define acceptance checks, and iterate. Over time you build a library of reliable patterns for your team, just like you would with code snippets or test templates.
Bottom line: CT AI shows you how to quantify model performance. CT GenAI shows you how to quantify the results you get from prompts.
3. Testing focus
CT AI is a tour of the unique testing work that AI systems demand. You look at test levels for data, for the model, and for the whole system. You learn how to test AI specific quality characteristics such as bias, transparency, and resilience to distribution shift. You practice metamorphic testing, where you vary inputs in a principled way to see whether expected relations still hold even when you cannot compute a single exact answer.
CT GenAI is a tour of the unique testing work that LLMs enable. You generate candidate test cases and data sets. You ask the model to propose oracles and you check them. You triage failures with quick summaries and suggested next steps, then you verify those suggestions. You learn when a lightweight agent flow is enough and when you should add retrieval so the model stays grounded in your own documentation and code.
Takeaway: CT AI is about testing the intelligence in the product. CT GenAI is about using intelligence to test the product.
4. Environments and operations
CT AI covers simulation and virtual environments for complex or risky behavior, plus monitoring for drift once a model is in the wild. You learn how to decide when to retrain or roll back.
CT GenAI covers the infrastructure that lets a team use LLMs day in and day out. You compare agent patterns, you learn where retrieval fits, and you look at practical fine tuning. You learn what LLMOps looks like for a test group, from access control and prompt catalogs to evaluation sets and run logs.
5. Governance and risk
CT AI puts fairness, safety, and transparency at the center of testing AI systems. You learn how to probe for bias, how to demand and test for explanations, and how to communicate uncertainty to stakeholders.
CT GenAI puts safe use and accountability at the center of testing with AI. You learn to detect and mitigate hallucinations and reasoning errors. You see concrete ways to prevent sensitive data from leaving your boundary. You also consider the energy cost of model use and how that affects tool choices and frequency of use.
6. Who each syllabus is for
- Choose CT GenAI if you lead or coach a test team that wants to adopt LLMs responsibly. It is the fastest way to raise the floor on daily practice across analysis, design, execution, and reporting.
- Choose CT AI if you are responsible for assuring ML features, or you operate in regulated or safety critical settings where model behavior must stand up to scrutiny.
- Choose both if you set standards for a test organization. In that case, start with CT GenAI for near term wins, then invest in CT AI to build durable capability in testing the models that matter.
Suggested learning path for 2025
Step one. CT GenAI in four to six weeks
- Learn structured prompting and a small set of techniques you can remember under pressure.
- Build a prompt and output evaluation loop for two or three high value test tasks, such as deriving acceptance criteria, creating candidate test cases, or summarizing regression failures.
- Put guardrails in place for privacy, sensitive data, and non deterministic answers.
- Decide when to use retrieval and when a lightweight agent is enough.
- Write a short acceptable use policy and a run book for your team.
Step two. CT AI in six to ten weeks
- Strengthen your understanding of data quality and metric selection for your domain.
- Practice metamorphic and adversarial thinking against the models that matter in your product.
- Add explainability checks and set clear acceptance criteria for probabilistic behavior.
- Stand up drift monitoring and define re qualification triggers.
- Use simulation where direct testing would be too risky or expensive.
Step three. Operate and improve
- Treat prompts as you treat code. Review them, version them, and attach them to test artifacts.
- Track outcomes such as cycle time, defect discovery rate, coverage growth, and false positives.
- Extend governance as adoption grows. That includes model choice, data handling, and environmental footprint.
- Evolve LLMOps and MLOps together so you can reason about quality across both.
Frequently asked questions
Do I need both certifications
Not always. If your products do not include ML features, CT GenAI may be enough to raise team productivity and quality. If you ship ML capabilities, CT AI becomes essential. Leaders and architects benefit from both since they guide both the use of AI in testing and the testing of AI in products.
Which exam is harder
CT AI goes to K4 and expects you to analyze scenarios and select techniques. CT GenAI goes to K3 and expects you to apply prompt patterns, evaluate outputs, and explain safe use. Both are manageable with a focused study plan.
Which one should I do first
If you want visible improvements in team speed without sacrificing quality, start with CT GenAI. Then build deeper assurance with CT AI if your product includes models or you work in a high trust domain.
Sources and official pages
- CT AI overview and downloads: https://istqb.org/certifications/certified-tester-ai-testing-ct-ai/
- CT GenAI overview and downloads: https://istqb.org/certifications/gen-ai/
Contact us to help you pass
Need a clear study plan, proxy exam, or practice exams for CT GenAI or CT AI, We can help. Contact us and we will craft a plan that fits your timeline to take the exam.
Trust us to help you pass CT-GenAI exam on your first attempt. Guranteed!
Discover more from ISTQB Guru
Subscribe to get the latest posts sent to your email.
Have a question? Ask here.