Autumn 2026: nine new tests, nine rebuilds, and the debrief

It has been a long time since we had a month this busy on Pilotest. Since mid-August, nine tests have been born and nine more have been rebuilt from top to bottom. The question banks have been gone through one by one, and the result screen now offers you a full debrief of your run. The site currently serves 169 tests and 85,437 questions.

First of all, we want to thank warmly every candidate who takes the time to write to us on the way back from their assessment. It is thanks to your feedback, and to nothing else, that we know what needs fixing and in which direction. Almost everything below came out of your emails. Please keep them coming, they are precious to us!

"See my debrief": finally knowing where you went wrong

To our mind this is the most useful change of the season. On a test's result screen, a "See my debrief" button now opens a full review of your run: every question as you saw it, the answer you gave, the correct one, and on some tests the explanation behind it.
Until now your score told you where you stood, but not why. Yet a candidate stuck on a reasoning test needs to know whether they are tripping on one particular kind of logic or simply running out of time everywhere. The work to do afterwards is clearly not the same! The debrief arrived test by test and does not cover the whole catalogue yet. The button only shows up where it has something to show you.

Nine new tests

All of them are playable right now. Do bear in mind, though, that a new test displays "calibration in progress" until enough people have sat it: your raw score is correct, but the mark out of 9 that places you against everyone else needs a population before it means anything. A little patience, then, as it fills up on its own.
One practical note: most of these are not translated yet, so their instruction pages will show in French for the time being. The tasks themselves are visual or numerical and play just as well.

• Aeronautical general knowledge, Air France cadets: we long thought a test was unnecessary here, given how freely the material is available online. We have changed our minds. The test follows the format you describe to us: 4 options, a timer on each question, 3 marks for a correct answer and -1 for a wrong one. Its real distinguishing feature lies elsewhere, though. Every correction shows the right answer, an explanation and a source, so that the test teaches you something instead of merely marking you. 1,363 questions in the bank, spread over nine subject areas, from principles of flight right through to the Air France fleet.
  A crucial word on the level, because this is where many people misjudge it. Your feedback from the September 2026 session is unanimous: the paper sits well above basic introductory level. Do not stop there. The right target is the PPL theoretical syllabus, and that is the level we pitch our questions at.
• Working memory: reverse digit span: digits stream past at one per second, and you then have 10 seconds to type them back… in reverse. Sequences run from 4 to 9 digits.
• Working memory II: a 2-back applied to arithmetic. You have to hold one result in mind while the next calculation is already on screen. Brutal.
• Angles with free entry and angles to tick: our angle series is now complete. Multiple choice estimation came first, then free entry, and now multiple selection, which forgives neither approximation nor guesswork.
• Chairs (fixed viewpoint): a rotation in space, but seen from an imposed position. Turning the object over in your head is no longer enough, you have to walk around it. It is the cousin of our cars test.
• Tangram with free answer: the tangram, but with no options to pick from. So much for any elimination strategy…
• Colour 2-back: the n-back moved onto colour rather than symbols. The switch sounds minor. It really is not: the little verbal anchors that help you along on digits stop working entirely here.
• Attention 3 with disruptors: the hard version of our attention 3, with interruptions that break the rhythm of your scanning.

Nine tests rebuilt from top to bottom

The rest of the work went into our own generators. When several of you report the same thing on the way back from a session, we put the test back on the workbench. That happened nine times this summer. Be aware that this has a consequence for you: when a test changes this much, its version changes too, and its scoring goes back into calibration. Your older results on those tests therefore no longer count towards your averages. It is frustrating at the time, we know, but it is the only honest way of not comparing runs that are no longer comparable.

• ENAC-EPL psychomotor test: rebuilt twice in a row to reach version 4, and by far the biggest job of the lot. Events now arrive in bursts, with a minimum distance between two appearances, and the colour scheme has been reworked.
• Successive shapes and colours: rebuilt from scratch following your feedback in September.
• Sliding and overlapping shapes, version II: our plates had grown far busier than they ought to have been. The test had therefore drifted towards the harder end, which misleads you just as surely as the opposite would. That is now fixed.
• Cars (sequential): the pace has been tightened considerably. Reckon on 5 seconds per rotation and 8 seconds to designate the final position, instead of 15 seconds for everything.
• Attention 1, Object location, Mental arithmetic 3, Mathematics and Reading comprehension round off the list.

Other tests were touched up without their scoring moving at all, because only their presentation changed. Your results on those remain perfectly comparable. They are the Tangram, the Arithmetic grid, the Sliding shapes first version, and the Dominoes tutorial, which we rewrote entirely around the 28 rules the test uses.

The questions gone over with a fine-tooth comb

A knowledge test is obviously worth no more than its questions are worth. So we audited our three large banks this summer, and the outcome was a good deal less flattering than we had hoped…

The reading comprehension test went from 98 to 302 passages, which is 906 questions and 4,530 options, each one with its own written explanation. Writing those explanations doubled as an audit: you cannot write down why an option is wrong without first checking that it really is! We turned up correct answers sitting on options the passage contradicted, questions that in fact admitted two defensible answers, and even date typos that made the right answer wrong. All of it has been corrected.

The English test got the same treatment: 5,313 questions read one by one, and 285 withdrawn. Most of them were items where two options were equally acceptable. That is the worst possible flaw in a multiple-choice question, since it penalises precisely the candidates with the best command of the language. The mathematics test, finally, has had its generator rewritten and now serves you 52,075 problems, with variations in form the old model simply did not produce.

Tracking your progress

The results page for a selection has been redesigned as well. It now plots your eight ability curves on a single chart with a date filter, then breaks down the dimension you select test by test. The aim is to answer at a glance the only question worth asking a month before the assessment: what still needs work?
Our FAQ has also gained some fifty entries, drawn from the emails you send us most often.

And under the bonnet…

One last piece of work that you will not see but may well feel: the interface of all 169 tests has been rebuilt. Every test now fills the screen without ever scrolling the page, adapts to the shape of your display instead of being drawn for one single size, and the on-screen keypads you answer with all follow the same model from one test to the next. Sign-ins to the site are also protected against bots now, which should put an end to a few untimely disconnections.

That is it for this autumn! As always, do get in touch if anything strikes you as wrong, out of date or simply improvable. We wish you every success and good flying! ✈