Appearance
Study Plan (CT-AI Syllabus v2.0)
Syllabus Version
Aligned with the ISTQB® CT-AI Syllabus v2.0 (2026-04-17). See also: Official Resources
This plan covers 6 weeks of part-time study, about 5–6 hours a week. It is built for a tester who holds the ISTQB® Certified Tester Foundation Level (CTFL) certificate but has no machine learning (ML) background. The syllabus specifies 19.5 hours of instruction (1,170 minutes across 7 chapters). The plan adds time for the 8 hands-on exercises, the Topic Drills, and four timed mock exams.
Before You Start
- Download the official syllabus from ISTQB® and keep it open alongside this guide. The syllabus wording is what the exam tests.
- Bookmark the Glossary. The Chapter keywords ("know name and definition") come from it.
- Read the Exam Tips to learn the format (40 questions, 44 points, 60 minutes, pass mark 29).
- Optional, for the hands-on exercises: install Python 3 with
scikit-learn,pandasandmatplotlib. The hands-on objectives are not examined, but running the code makes Chapters 3, 5 and 6 much easier to remember.
How Each Week Works
For each lesson page: read it, do the hands-on exercise if there is one, take the Unit Quiz at the end, then do the Topic Drills for every section. Re-read any section where you score below 80% in a drill. The home page tracks your progress in this browser.
Recommended Schedule
Week 1: Foundations: what AI is and how its quality is judged (Ch 1–2)
| Day | Topic | Time |
|---|---|---|
| 1 | Ch 1: Introduction to AI, sections 1.1.1–1.1.4 | 1.5 hours |
| 2 | Ch 1, sections 1.1.5–1.1.8, and the Unit Quiz | 1.5 hours |
| 3 | Topic Drills 1.1.1–1.1.8 | 1 hour |
| 4 | Ch 2: Quality Characteristics, Unit Quiz, Topic Drills 2.1.1–2.2.1 | 1.5 hours |
| 5 | Review: the ISO/IEC 25059 characteristics and their parents; the acceptance-criteria table | 0.5 hours |
Chapters 1 and 2 are mostly recall and comparison, but together they carry 9 questions (20% of the points). Learn which quality characteristic is a sub-characteristic of which parent (for example, AI robustness is under reliability and intervenability under security). Practise classifying example acceptance criteria.
Week 2: Machine learning (Ch 3, the largest chapter)
| Day | Topic | Time |
|---|---|---|
| 1 | Ch 3a: Forms of ML and the ML Workflow, hands-on HO-3.1.3 | 1.5 hours |
| 2 | Ch 3b: Data and Performance Metrics, sections 3.2.1–3.2.3, hands-on HO-3.2.2 | 1.5 hours |
| 3 | Ch 3b, section 3.3.1 (K3), the worked examples, hands-on HO-3.3.2 and HO-3.3.3 | 1.5 hours |
| 4 | Ch 3c: Neural Networks, hands-on HO-3.4.2 (perceptron) | 1 hour |
| 5 | Topic Drills 3.1.1–3.4.3; redo the 3.3.1 drill until you get every calculation right | 1.5 hours |
Chapter 3 is 7 questions and 8 points, including one K3 question worth 2 points: calculating accuracy, precision, recall and F1-score from a confusion matrix. Learn the four formulas by heart. Practise until you can compute all four from raw counts in under three minutes, whichever way round the table is drawn.
Week 3: Testing AI-based systems (Ch 4)
| Day | Topic | Time |
|---|---|---|
| 1 | Ch 4a: Introduction to Testing AI-Based Systems: locked vs. adaptive, statistical testing, test oracles | 1.5 hours |
| 2 | Topic Drills 4.1.1–4.1.3 | 1 hour |
| 3 | Ch 4b: GenAI, Red Teaming and Test Levels, sections 4.2.1–4.2.2 (K3) | 1.5 hours |
| 4 | Ch 4b, section 4.3, and hands-on HO-4.2.3 (exploratory testing of an LLM) | 1.5 hours |
| 5 | Topic Drills 4.2.1–4.3.2 | 1 hour |
Chapter 4 is 7 questions and 8 points. The K3 question applies red teaming: know the five steps of the approach, the difference between security and safety red teaming, when to run it, and how automation extends coverage. Also learn the five ways of addressing the test oracle problem, and which test level covers which part of a machine learning system (MLS).
Week 4: Input data testing (Ch 5)
| Day | Topic | Time |
|---|---|---|
| 1 | Ch 5a: Risks, Bias and Data Pipelines: the risk table and testing for bias | 1.5 hours |
| 2 | Ch 5a, data pipeline testing; Topic Drills 5.1.1–5.1.3 | 1.5 hours |
| 3 | Ch 5b: Representativeness, Constraints and Labels, sections 5.1.4–5.1.5 (K3) | 1.5 hours |
| 4 | Ch 5b, section 5.1.6, and hands-on HO-5.1.7 | 1 hour |
| 5 | Topic Drills 5.1.4–5.1.6 | 1 hour |
Chapter 5 is 6 questions and 7 points. The K3 question applies dataset constraint testing. Be able to name the constraint type for any rule: single-value (missing, range, type), multi-value (sum, count, duplicate, useful, outlier) or comparison (greater than, correlate). Also learn the input data risk table: which mitigation goes with which risk.
Week 5: Model testing and development testing (Ch 6–7), first mock
| Day | Topic | Time |
|---|---|---|
| 1 | Ch 6a: Model Risks, Documentation, Performance and Adversarial Testing; Topic Drills 6.1.1–6.1.4 | 2 hours |
| 2 | Ch 6b: Metamorphic, Drift, Fitting, A/B and Back-to-Back Testing, section 6.1.5 (K3), hands-on HO-6.1.6 | 1.5 hours |
| 3 | Ch 6b, sections 6.1.7–6.1.10; Topic Drills 6.1.5–6.1.10 | 1.5 hours |
| 4 | Ch 7: ML Development Testing; Topic Drills 7.1.1–7.1.2 | 1 hour |
| 5 | Mock Exam 1, timed (60 minutes), then review every wrong answer | 1.5 hours |
Chapter 6 is the most heavily weighted: 9 questions and 10 points, or 23% of the exam. Its K3 question applies metamorphic testing, so practise deriving and judging metamorphic relations (MRs). Know the pairs the exam likes to contrast: data vs. concept drift, static vs. dynamic drift testing, overfitting vs. underfitting, and A/B vs. back-to-back testing. Chapter 7 adds 2 questions: the development risk table, and shadow vs. canary testing.
Week 6: Exam practice
| Day | Activity | Time |
|---|---|---|
| 1 | Official Sample Exam A: download the questions and answers (ISTQB® PDFs), answer the questions on paper in 60 minutes, then mark yourself against the answers | 2 hours |
| 2 | Mock Exam 2, timed, then review | 1.5 hours |
| 3 | Re-read the lesson sections behind your wrong answers; redo those Topic Drills | 1 hour |
| 4 | Mock Exam 3, timed, then review | 1.5 hours |
| 5 | Mock Exam 4, timed; aim for at least 33/44 (75%) before booking the exam | 1.5 hours |
The mock exams are deliberately harder than the official sample. If you score 29 or more (the real pass mark) on Mock Exams 3 and 4, you are in good shape. A score of 33 or more gives you a safety margin.
Study Tips
- Learn the syllabus's lists and tables, not just the ideas. Many questions ask which mitigation the syllabus pairs with a risk (tables in 5.1.1, 6.1.1 and 7.1.1), or which item belongs to a list (for example, the ML framework functions or the data preparation activities).
- Use your CTFL knowledge, but watch for redefined terms. "Regression" in ML means predicting a number, not re-testing after changes. "Bias" in a neuron is not the fairness bias of Chapter 5. "Performance testing" (speed) is not "ML functional performance testing" (accuracy, recall and similar).
- Practise the four K3 skills on paper. They are worth 8 of the 44 points: confusion-matrix metrics, red teaming, dataset constraints and metamorphic relations. The worked examples on each lesson page are templates you can reuse.
- Read scenario questions twice. Identify what is actually being asked (a risk? a test level? a mitigation?) before looking at the options. Distractors are usually real terms from the wrong context.
- Do the hands-on exercises even though they are not examined. Seeing validation accuracy differ from test accuracy, or a perceptron's weights change epoch by epoch, makes Chapters 3 and 6 far easier to recall under exam pressure.