Checked 29 September 2026. From ISTQB's Exam Structures & Rules Tables v1.19 (19 Aug 2026), Exam Structures and Rules v1.2 and the CT-AI v2.0 syllabus and FAQ. ISTQB revises the tables a few times a year; the CT-AI v1.0 figures lapse when v1.0 is retired (English 21 Apr 2027).
The ISTQB Certified Tester AI Testing exam has 40 questions but 44 points, and the difference decides how many mistakes you can afford. In the v2.0 exam, 36 questions are worth one point and four K3 questions are worth two points each. The pass mark is 29 of 44: ISTQB's rule is "at least 65%", 65% of 44 is exactly 28.60, and that rounds up to 29. The v1.0 exam, which you can still sit in English until 21 April 2027, is different: 47 points, pass mark 31, with K1 recall and K4 analysis questions that v2.0 no longer has. This guide sets out ISTQB's own blueprint for both versions, chapter by chapter, and ends with a calculator that tells you whether your expected answers clear the bar.
ISTQB Exam Structures & Rules Tables v1.19, overview table.
Where other pages get this wrong
Pass mark depends on the version: 29 of 44 on v2.0, 31 of 47 on v1.0.
Commonly published instead: A single exam table with 44 points and a pass mark of 29, shown to readers who are also told they can still sit v1.0 — including on our own CT-AI notice-board page, which we are correcting here.
ISTQB's overview table lists the two exams separately. A v1.0 candidate who aims for 29 points has aimed two points too low and is working to the wrong total: v1.0 has 47 points, not 44.
v2.0 and v1.0 side by side
| Item | CT-AI v2.0 | CT-AI v1.0 |
|---|---|---|
| Questions | 40 | 40 |
| Points | 44 | 47 |
| Exactly 65% of the points | 28.60 | 30.55 |
| Pass mark (at least 65%) | 29 | 31 |
| Time | 60 min (75 with +25%) | 60 min (75 with +25%) |
| Questions by K-level | K1 0 · K2 36 · K3 4 · K4 0 | K1 4 · K2 29 · K3 3 · K4 4 |
| ISTQB's approximate answering time by K-level | 48 min | 58 min |
| Two-point questions | 4 (the K3 questions) | 7 (3 K3 + 4 K4, our reading of the chapter totals) |
| Examinable until | current version | 21 Apr 2027 in English, 21 Oct 2027 in other languages |
Source: ISTQB Exam Structures & Rules Tables v1.19 (overview and per-version tables) and the CT-AI v2.0 FAQ. The v1.0 two-point count is our arithmetic: its chapter totals only add to 47 when K3 and K4 questions carry two points each.
ISTQB's FAQ describes the v2.0 pass mark simply as "65%". Both statements are right: the general rule in ISTQB's Exam Structures and Rules (5.3.1) is "a score of at least 65%", and 29 is the first whole number of points at or above 65% of 44. On 29 points you have scored 65.9%.
The v2.0 blueprint, chapter by chapter
ISTQB fixes how many questions each chapter gets and at which K-level. The table below is the official v2.0 distribution with the minimum training time ISTQB sets for each chapter in syllabus section 0.10. The share-of-points column is our arithmetic.
| Chapter | Training minutes | Questions | K2 | K3 | Points | Share of 44 points (ours) |
|---|---|---|---|---|---|---|
| 1. Introduction to AI | 120 | 6 | 6 | 0 | 6 | 13.6% |
| 2. Quality characteristics for AI-based systems | 45 | 3 | 3 | 0 | 3 | 6.8% |
| 3. Machine learning | 375 | 7 | 6 | 1 | 8 | 18.2% |
| 4. Testing AI-based systems | 195 | 7 | 6 | 1 | 8 | 18.2% |
| 5. Input data testing for ML systems | 180 | 6 | 5 | 1 | 7 | 15.9% |
| 6. Model testing for ML systems | 225 | 9 | 8 | 1 | 10 | 22.7% |
| 7. ML development testing | 30 | 2 | 2 | 0 | 2 | 4.5% |
| Total | 1,170 (19.5 h) | 40 | 36 | 4 | 44 | 100% |
ISTQB Exam Structures & Rules Tables v1.19. Chapter 6 alone is 10 points; chapters 3, 4 and 6 together are 26 of 44.
Three things stand out. Chapter 6, model testing, is the single biggest chapter at 10 points from 9 questions. Chapters 3, 4 and 6 together carry 26 points — more than half the paper. And chapters 2 and 7 together are only 5 points, even though chapter 2 carries the vocabulary (ISO/IEC 25059 quality characteristics) that the rest of the syllabus leans on.
Where training time and points disagree
| Chapter | Share of training time (ours) | Share of points (ours) | Points per training hour (ours) |
|---|---|---|---|
| 1. Introduction to AI | 10.3% | 13.6% | 3.0 |
| 2. Quality characteristics for AI-based systems | 3.8% | 6.8% | 4.0 |
| 3. Machine learning | 32.1% | 18.2% | 1.3 |
| 4. Testing AI-based systems | 16.7% | 18.2% | 2.5 |
| 5. Input data testing for ML systems | 15.4% | 15.9% | 2.3 |
| 6. Model testing for ML systems | 19.2% | 22.7% | 2.7 |
| 7. ML development testing | 2.6% | 4.5% | 4.0 |
Our arithmetic on ISTQB's minutes and points. Chapter 3 takes 32.1% of the teaching time (375 of 1,170 minutes) and also contains five of the eight hands-on exercises, but returns 8 points (18.2%). Hands-on exercises are not examinable, so a candidate who self-studies can spend less time building models and more on chapters 4 to 6.
The four K3 questions carry 8 of the 44 points
In v2.0 there are only four learning objectives at K3 ("apply"), and each is examined by one two-point question. They are the only K3 (apply) questions in the blueprint; the other 36 are K2 questions, which can still ask you to compare, classify, explain or infer.
| Learning objective | What you must do (syllabus wording) | Chapter | Points |
|---|---|---|---|
| AI-3.3.1 (K3) | Calculate common ML functional performance metrics from a given set of confusion matrix data | 3 | 2 |
| AI-4.2.2 (K3) | Implement red teaming for GenAI systems | 4 | 2 |
| AI-5.1.5 (K3) | Apply dataset constraint testing | 5 | 2 |
| AI-6.1.5 (K3) | Use metamorphic testing to derive test cases for a given scenario | 6 | 2 |
From the CT-AI Syllabus v2.0 learning-objective list and the Exam Structures & Rules Tables v1.19. The K3 guide in this series works through all four.
The arithmetic is worth doing once. If you get all four K3 questions right, you need 21 of the 36 one-point questions (21 + 8 = 29). If you get none of them right, you need 29 of 36 — about 81% of the rest. Each K3 question you drop costs you two one-point questions elsewhere.
Question types and what the exam does not have
How CT-AI questions work
- All questions are multiple choice and may need one or more answers. In ISTQB's Sample Exam A v2.2, 38 of the 40 main questions are select-one and two (Q2 and Q39) are select-two, each worth one point. ISTQB warns that real exams may include a wider variety of question types than the sample.
- No partial scoring. ISTQB's exams FAQ: on a question that needs two options you score only if you select both correct answers.
- No K1 questions in v2.0. Pure recall has gone from the question mix, but definitions still matter. The two ISTQB documents word it differently: the CT-AI v2.0 syllabus (0.5) says keyword names and definitions "shall be remembered (K1)"; the Exam Structures and Rules (4.1.4) make keywords examinable "for understanding their definition (K2)". Either way, learn them; in the v2.0 question mix they can only arrive as K2 questions.
- No K4 questions in v2.0. v1.0 had four K4 ("analyse") questions worth two points each.
- Not every learning objective appears in every exam. In chapter 1, eight K2 objectives share six questions; in chapter 3, seven K2 objectives share six. Chapters 2, 4, 5, 6 and 7 have one question per objective.
- Negative marking: no ISTQB document we read states one. The rules define points earned and a 65% pass mark only, so do not leave questions blank.
ISTQB Exam Structures and Rules v1.2 (4.1.4, 5.3.1), Exam Structures & Rules Tables v1.19, CT-AI Syllabus v2.0 (0.5), CT-AI Sample Exam A v2.2, ISTQB exams FAQ.
| ISTQB exam rule | What it says | Section |
|---|---|---|
| Pass mark | A score of at least 65% is required to pass | ESR 5.3.1 |
| Points per question (Specialist) | K2 = 1 point; K3 "in general" 2; K4 "in general" 3 | ESR 5.2.2 |
| Suggested time per question | K1 1 min, K2 1-2 min, K3 3 min, K4 4 min (a guideline for average time, not a per-question limit) | ESR 6.1 |
| Extra time | +25% when the exam is not in the candidate's native language | ESR 6.2.1 |
| Keywords | All syllabus keywords are examinable for understanding their definition (K2) | ESR 4.1.4 |
| Minimum length | At least 40 questions | ESR 5.1.1 |
ISTQB Exam Structures and Rules v1.2 (2 May 2025). The CT-AI tables apply these rules: v2.0 uses K2 = 1 and K3 = 2 throughout. v1.0's K4 questions come out at 2 points each in its chapter totals, which is why ISTQB says "in general".
On timing, ISTQB's own estimate is 48 minutes of answering time for the v2.0 question mix, which leaves about 12 minutes of the 60 for review. With the 25% extension, which ISTQB gives when the exam is not in the candidate's native language, you have 75 minutes; ask your exam provider how to claim it when you book.
Check your score
CT-AI score calculator: will your answers reach the pass mark?
Pick the syllabus version you are sitting, then enter how many one-point and two-point questions you expect to get right. The calculator adds up your points and compares them with ISTQB's pass mark for that version.
Points per question from ISTQB's Exam Structures & Rules Tables v1.19 (v2.0: K2 = 1 point, K3 = 2; v1.0: K1 and K2 = 1, K3 and K4 = 2, our reading of the chapter totals). ISTQB's exams FAQ: there is no partial scoring, so a pick-two question scores only when both picks are right. Our arithmetic; the exam provider's result is final.
| Two-point questions right | One-point questions needed, v2.0 (pass 29) | One-point questions needed, v1.0 (pass 31) |
|---|---|---|
| 4 (v2.0 max) | 21 of 36 | 23 of 33 |
| 3 | 23 of 36 | 25 of 33 |
| 2 | 25 of 36 | 27 of 33 |
| 1 | 27 of 36 | 29 of 33 |
| 0 | 29 of 36 | 31 of 33 |
Our arithmetic on ISTQB's pass marks. v1.0 has seven two-point questions, so a v1.0 candidate with 5, 6 or 7 of them right needs 21, 19 or 17 of the 33 one-point questions.
What the v1.0 blueprint looked like
| v1.0 chapter | Training minutes | Questions | Points |
|---|---|---|---|
| 1. Introduction to AI | 105 | 4 | 4 |
| 2. Quality characteristics for AI-based systems | 105 | 4 | 4 |
| 3. Machine learning — overview | 145 | 4 | 5 |
| 4. Machine learning — data | 230 | 4 | 4 |
| 5. ML functional performance metrics | 120 | 3 | 5 |
| 6. ML — neural networks and testing | 65 | 2 | 2 |
| 7. Testing AI-based systems overview | 115 | 4 | 5 |
| 8. Testing AI-specific quality characteristics | 150 | 4 | 5 |
| 9. Methods and techniques for testing AI-based systems | 245 | 6 | 8 |
| 10. Test environments for AI-based systems | 30 | 1 | 1 |
| 11. Using AI for testing | 195 | 4 | 4 |
| Total | 1,505 (25.1 h) | 40 | 47 |
ISTQB Exam Structures & Rules Tables v1.19 (CT-AI v1.0 table) and CT-AI Syllabus v1.0. If you are sitting v1.0, chapter 9 (methods and techniques) is the biggest at 8 points, and chapter 11 (using AI for testing, 4 points) is content v2.0 dropped entirely.
How to use the blueprint
Turning the numbers into a plan
- Revise in points order, not page order: chapter 6 (10), chapters 3 and 4 (8 each), chapter 5 (7), chapter 1 (6), chapter 2 (3), chapter 7 (2).
- Drill the four K3 skills until they are mechanical — they are 18% of the points for 10% of the questions.
- Learn every keyword definition: with no K1 questions, definitions arrive as K2 questions that ask you to tell similar terms apart.
- Take ISTQB's free Sample Exam A v2.2: time questions 1 to 40 at 60 minutes, and do the six appendix questions (A1 to A6) separately as extra practice. ISTQB warns it "may both be more difficult or less difficult than any official exam".
- Budget about 1.2 minutes per question on the first pass (48 of 60 minutes by ISTQB's estimate) and use the rest to review flagged questions.
Our suggestions, built on ISTQB's published tables.
Practise it, don't just read it
The ProSyllabus quizzes follow the v2.0 chapters, so you can check each chapter against its share of the 44 points.
The ProSyllabus CT-AI library has 10 quizzes of 10 questions written on the v2.0 syllabus and covering all seven chapters: Introduction to AI (two quizzes), quality characteristics, machine learning (two: forms and workflow; metrics, neural networks and coverage), testing AI-based systems (two: test oracles and test levels; generative AI and red teaming), input data testing, and model testing plus ML development testing (two). They are our own questions, not ISTQB exam items, and they are single-answer only, while ISTQB's official Sample Exam A v2.2 also includes two select-two questions, so practise that format there.
How many questions are in the ISTQB CT-AI exam?
40 multiple-choice questions in 60 minutes, in both v1.0 and v2.0. Candidates sitting in a non-native language get 25% more time, 75 minutes.
What is the CT-AI passing score?
v2.0: 29 of 44 points. v1.0: 31 of 47 points. Both are ISTQB's rule of at least 65%, rounded up to a whole point.
Why is CT-AI out of 44 and not 40?
Four of the 40 v2.0 questions are K3 (apply) questions worth two points each: 36 × 1 + 4 × 2 = 44.
Which CT-AI v2.0 chapter carries the most marks?
Chapter 6, Model Testing for ML Systems: 9 questions, 10 points. Chapters 3 and 4 follow with 8 points each.
Is there negative marking in CT-AI?
No ISTQB document we read states any negative marking. The rules define points earned and a 65% pass mark, so answer every question.
Are there multiple-answer questions?
They can occur. ISTQB's Sample Exam A v2.2 has two select-two questions (Q2 and Q39), each worth one point, and 38 select-one. ISTQB says there is no partial scoring: you must pick both correct options to score.
Does v2.0 have K1 or K4 questions?
No. v2.0 is 36 K2 and 4 K3 questions. v1.0 had 4 K1, 29 K2, 3 K3 and 4 K4 questions.
Which are the K3 learning objectives in v2.0?
AI-3.3.1 confusion-matrix metrics, AI-4.2.2 red teaming for GenAI systems, AI-5.1.5 dataset constraint testing and AI-6.1.5 metamorphic testing.
Are the hands-on exercises examined?
No. The syllabus says all sections are examinable except the introduction, hands-on objectives, references and appendices.
How long does ISTQB expect answering to take?
Its tables estimate about 48 minutes for the v2.0 question mix and 58 minutes for v1.0, within the 60-minute exam.
Sources
- official — ISTQB — Exam Structures & Rules Tables v1.19 (19 Aug 2026, PDF)
- official — ISTQB — Exam Structures and Rules v1.2 (2 May 2025, PDF)
- official — ISTQB — CT-AI Syllabus v2.0 (GA 17 Apr 2026, PDF)
- official — ISTQB — Help: CT-AI Version 2.0 (FAQ)
- official — ISTQB — Certified Tester AI Testing (CT-AI) certification page (v2.0)
- official — ISTQB — CT-AI Sample Exam set A, Questions v2.2 (PDF)
- official — ISTQB — CT-AI Sample Exam set A, Answers v2.2 (PDF, released 20 Jul 2026)
- official — ISTQB — Help: ISTQB Exams (no partial scoring on multiple-answer questions)
- official — ISTQB CT-AI Syllabus v1.0 (GA 1 Oct 2021) — Indian Testing Board copy (PDF)
official = a document published by the conducting body. reported = a news or coaching site we could not check against an original. Where sources disagree this page says so rather than picking one.
