How to Pass the ISTQB CT-AI Exam: a Study Plan

To pass the ISTQB CT-AI exam, hold a valid CTFL certificate first, then study the current CT-AI v2.0 syllabus (about 19.5 hours of material), practise ML-metric calculations and generative-AI testing scenarios, and rehearse with full mock exams. The exam is 40 questions, 47 points, a 65% (31-point) pass mark, and 60 minutes (75 minutes if English is not your native language). With a focused four-week plan, that is very achievable.

What you need before you start

CT-AI is a Specialist certification, not an entry point. You must already hold the ISTQB Certified Tester Foundation Level (CTFL) — it is a formal prerequisite to register. ISTQB also recommends at least six months of practical experience in testing, software development, or data science. You do not need to be a programmer, but you should be comfortable with basic testing vocabulary. If you are still deciding whether the certification is right for you, start with our ISTQB CT-AI certification guide.

Know the CT-AI v2.0 exam format

The exam is closed-book and multiple-choice. Learn these numbers cold so nothing surprises you on the day:

• 40 questions worth 47 points in total (a mix of 1-point and 2-point items).

• Pass mark 65% — you need at least 31 of the 47 points.

• 60 minutes, extended to 75 minutes if you sit the exam in a non-native language.

• Questions map to cognitive levels K1–K3, so expect definitions, comparisons, and applied scenarios — not just recall.

v2.0 reorganised the syllabus into seven chapters around data, ML model, and system-level testing, and cut the standalone "Testing with AI" block (now part of CT-GenAI). For a full breakdown, see CT-AI v2.0 vs v1.0: what changed.

A 4-week study plan

Week 1 — Foundations and quality characteristics

Read chapters 1–2 of the syllabus: what makes AI-based systems different (non-determinism, self-learning, the data dependency), and the quality characteristics that matter — flexibility, autonomy, evolution, bias, transparency and safety. Build a one-page glossary in your own words. Anchor unfamiliar terms against the official ISTQB glossary.

Week 2 — ML metrics and the confusion matrix

This is where most people lose points. Learn to read a confusion matrix and compute precision, recall and F1 by hand. Understand the accuracy paradox on imbalanced data, when to favour recall over precision (and vice-versa), and the basics of RMSE vs MAE for regression. Practise five to ten calculation problems until they feel routine.

Week 3 — ML testing levels, GenAI and red teaming

Cover the two ML-specific levels new to v2.0 — input data testing and ML model testing — plus deployment testing (data drift, canary releases, monitoring). Then work through generative-AI content: LLM-specific challenges, prompt injection, hallucination, RAG evaluation (groundedness, faithfulness), and red teaming. Learn the black-box test techniques for AI: metamorphic testing, back-to-back testing, and adversarial examples.

Week 4 — Mock exams and gap-filling

Sit full 40-question mocks under timed conditions. After each attempt, review every question — including the ones you got right — and read the rationale. Log your weak topics and re-read only those syllabus sections. Aim to score 80%+ on mocks consistently before booking the real exam. You can practise with our free ISTQB CT-AI mock exams.

Topics that trip candidates up

• Metric maths under time pressure. Two-point calculation questions reward the people who practised precision/recall by hand.

• Confusing testing AI with testing using AI. CT-AI is about testing AI systems; the "testing with AI" material moved to CT-GenAI.

• Data-related defects. Know target/data leakage, label quality, and bias arising from historical data.

• AI-specific techniques. Metamorphic and back-to-back testing exist because AI outputs often have no single expected result.

Exam-day tactics

With 60 minutes for 40 questions you have about 90 seconds each. Answer the easy recall items first, flag calculation questions, and return to them. Never leave a question blank — there is no negative marking. Watch for "select two" multiple-response items and count your selections. Read qualifier words such as least, best, and not carefully; they change the correct answer.

Confirm the current exam rules and syllabus version on the official ISTQB CT-AI certification page before you book. Then treat mock exams as your final rehearsal — original questions with a written rationale for every answer teach far more than memorising a dump.