How our questions are made
A white paper on the origin, structure and verification of the question bank
Why this document exists
When you buy a question bank you are trusting someone else's homework. Most providers never tell you where their questions come from, who checked them, or what happens when a regulation changes. We think that is backwards: the only thing you are really paying for is trust in the content, so the content should be able to explain itself.
This paper describes exactly how every question in this bank was created, what it is anchored to, and the four checks it must survive before you ever see it.
1. Where the questions come from — and where they do not
The ECQB is confidential, and we do not have it
European theory exams for ATPL, CPL and IR draw on the European Central Question Bank (ECQB), a database of roughly 10,000 questions maintained by EASA. It is distributed only to national aviation authorities, only for running exams. EASA states plainly that its content is confidential and protected by intellectual property rights, and it investigates violations.
So there is no legal way to obtain it. Anyone who implies they have the "real questions" is either mistaken or describing something they should not have.
What is public: the Learning Objectives
What EASA does publish is the syllabus itself: the Learning Objectives (LOs), in AMC1 FCL.310; FCL.515(b); FCL.615(b); FCL.835(d), ED Decision 2020/018/R. This document — 609 pages — states, in hierarchical codes, exactly what a student pilot must know.
A Learning Objective looks like this:
050 01 02 02 (03) — Explain the effect of a constant ISA deviation on temperature during a step climb.
Each one carries a code, a topic, the exact knowledge required, and the licences it applies to (ATPL, CPL, IR, CB-IR, and the helicopter equivalents).
We extracted 4,667 Learning Objectives from that document, programmatically, and they are the seed of everything. Every question in the bank is written from an LO and carries its code.
The chain you can audit
EASA Learning Objective → question → public regulation
Open any question in the app and tap Source. You will see the LO code it was written from and the regulation the answer rests on — an ICAO Annex, a Chicago Convention article, an EU Regulation, a Certification Specification. Nothing is hearsay, nothing is copied from a competitor, nothing comes from someone's memory of an exam.
2. Coverage: all 13 subjects
The aeroplane ATPL syllabus is thirteen subjects. Here is the whole map, with the number of Learning Objectives in each:
| Code | Subject | LOs |
|---|---|---|
| 010 | Air Law | 507 |
| 021 | Aircraft General Knowledge — Airframe, Systems, Power Plant | 668 |
| 022 | Aircraft General Knowledge — Instrumentation | 460 |
| 031 | Mass and Balance | 86 |
| 032 | Performance — Aeroplanes | 243 |
| 033 | Flight Planning and Monitoring | 118 |
| 040 | Human Performance and Limitations | 363 |
| 050 | Meteorology | 465 |
| 061 | General Navigation | 119 |
| 062 | Radio Navigation | 328 |
| 070 | Operational Procedures | 388 |
| 081 | Principles of Flight — Aeroplanes | 497 |
| 090 | Radio Communications | 79 |
The fourteenth. For licences where communications are examined separately — 091 VFR and 092 IFR — the same material has to split. Every question in subject 090 is tagged vfr, ifr or general, so a VFR exam draws vfr + general and an IFR exam draws ifr + general. One body of content, three exams.
Helicopter subjects (034 Performance, 082 Principles of Flight) are a separate track.
3. The eight question archetypes
A bank of a thousand definition questions does not prepare anyone. Real exams test the same knowledge from different angles, so every Learning Objective is attacked through up to eight archetypes, across three difficulty levels:
Learning Objective
├── EASY
│ ├── conceptual the fundamental relationship
│ └── terminology the exact meaning of a technical term
├── MEDIUM
│ ├── scenario a concrete operational situation, real numbers
│ ├── calculation one computational step
│ └── misconception built around the mistake students actually make
└── HARD
├── multi_step two or more chained steps
├── edge_case a limit, an exception, a non-standard condition
└── reverse given the outcome, work back to the cause
Why this matters for you. Because every question is tagged with its archetype, the app can tell you what kind of thinking you are weak at — not just which subject. Getting the definitions right but failing every reverse question is a completely different problem from failing the calculation ones, and it needs different study.
Archetypes are not forced
Not every objective supports all eight. Air Law, Operational Procedures, Human Performance and Radio Communications contain no calculations; inventing one produces an artificial question that a student can smell immediately. Where an archetype does not genuinely fit, we record why it was skipped rather than padding the count.
Meteorology gets a special rule: its questions favour conceptual inversion (cause↔effect, increases↔decreases, high↔low), because that is precisely where students go wrong.
4. How a single question is built
Every question is four options, exactly one correct. But the three wrong answers are not filler — each has a defined job:
| Role | Purpose |
|---|---|
| correct | the answer |
| obvious | clearly wrong to anyone who studied |
| misconception | the classic documented error on this topic |
| plausible | technically credible, fails on one detail |
That structure is what makes your mistakes diagnostic. Choosing the misconception option means something specific and fixable; choosing the obvious one means you have not covered the material at all. The app knows the difference.
Other rules the writers work to:
- Stem length ~35 words. Long enough to set a real situation, short enough that it tests knowledge and not reading stamina.
- Options of similar length. A conspicuously long answer is a free hint.
- Numeric distractors derive from real errors — sign inverted, TAS used instead of GS, taxi fuel deducted twice, cosine used instead of sine — never from random numbers.
- No "All of the above", no "None of the above", no double negatives.
- Self-contained. If a question needs a METAR, a load table or a wind component, the data is in the question, not in a figure you must hunt for.
- English only. The professional exams are in English in all 30 EASA states (France alone also offers French), and ICAO English Level 4 is mandatory anyway. Translating the questions would train you in the wrong language.
5. Jurisdiction: what changes between countries, and what does not
Aviation law is layered. A statement can be perfectly true under an ICAO Annex and misleading for a European candidate, because EU law overrides it.
Every question therefore carries a jurisdiction tag:
easa_wide— the answer is identical in all 30 EASA states, because it comes from ICAO, an EU regulation, or physics. This is the overwhelming majority: meteorology, aerodynamics, performance, navigation and mass and balance do not change at a border.national_variant— the answer depends on the state, with a note explaining exactly what varies. Almost all of these live in Air Law and Operational Procedures.
Real examples of traps this discipline caught in our own drafts:
- Cabotage under Chicago Article 7 lets a state refuse — but not between EU states, where Regulation 1008/2008 gives Union carriers full rights.
- Domestic commercial air transport is not left to national rules in the EASA area: Regulation 965/2012 applies directly.
- The compensation bands of Regulation 261/2004 change for intra-Community routes, so a 6,000 km flight is not automatically in the €600 band.
Each of those was a question that read as correct and would have taught you something false.
6. The four checks every question survives
Writing a question is the easy part. Everything below happens before it reaches you, and a question that fails any step goes back and does not ship.
Step A — deterministic checks
No AI, just code: complete schema, four options, all distractor roles present, the correct-answer index actually pointing at the correct answer, stem within length, no banned constructions, no reference to a missing figure, the LO code existing in the EASA document, no duplicate IDs, no repeated options.
Step B — the fact checker
Recomputes every calculation independently, from scratch, in code. Checks the claim against the cited regulation. Checks the reference actually exists, is relevant, and is not invented or repealed. Verdict: is the key genuinely correct?
Step C — the adversary
A separate reviewer whose job is the opposite: prove the question is broken. It hunts for a second defensible answer, for a missing assumption in the stem that would flip the key, for wording that can be read two ways. A question it demolishes never reaches you. When in doubt, the adversary wins — a rewritten question costs us an hour; a wrong one costs you an exam.
Step D — exam-standard conformity
Does the item behave like a real professional exam question? One concept tested, correct ICAO/EASA terminology, no grammatical tell that gives away the key, homogeneous distractors, declared difficulty matching the real difficulty, and an explanation that stands on its own as teaching material.
Then, and only then, a human
Questions that clear all four reach a qualified theory instructor, who approves or rejects. Only instructor-approved questions are served to paying students.
Why four separate passes instead of one
A single reviewer asked to "check quality" tends to agree with what it reads. Splitting verify the fact from try to destroy it from check the form forces three readings with different goals — and those are exactly the three places errors hide: the arithmetic slip, the second acceptable answer, the ambiguous sentence.
On the first two batches through the chain, 29% of drafts were sent back. That number is the point of the system, not an embarrassment: those are the questions you never had to see.
7. Keeping it current
Aviation regulation moves. Rule numbers get renumbered — the all-weather operations amendment renumbered a whole block of Part-CAT — thresholds are revised, and conventions are amended for some states and not others.
Every question stores the source it depends on. When a source document changes, the questions that rest on it are automatically marked for re-verification rather than being quietly assumed to still be true.
8. What we will never do
- We will never copy, reconstruct or paraphrase the ECQB. It is confidential and it is not ours.
- We will never copy questions from another provider. Their wording is their copyright.
- We will never ask you to reproduce questions you saw in a real exam, and we do not build our content from exam recall.
- We will never present a question as verified when only a machine has read it.
- We are not affiliated with, endorsed by, or approved by EASA or any national aviation authority.
In one paragraph
Every question starts from a published EASA Learning Objective, is written against a public regulation you can look up, is attacked from up to eight different angles at three difficulty levels, carries distractors engineered to diagnose specific misconceptions, is tagged for the licences and the countries it applies to, must survive an automated check, an independent fact check, a deliberate attempt to demolish it, and an exam-standard review — and only then is signed off by a human instructor before you ever see it.
That is what you are paying for.
→ Take the sample test