The Commission Quietly Changed Its Internal Competition Test — Here's the Full Breakdown

If you are preparing an internal competition at the European Commission — AD7, AD5 or the brand-new AST3 — stop for a second, because the first hurdle no longer looks like anything you have trained for.

Since 2026, the Commission's internal competitions open with an adaptive cognitive assessment delivered on the SHL platform. Not the familiar EPSO-style multiple choice. Not verbal reasoning passages. A single 36-minute sitting, 24 interactive questions, three reasoning types — and a pass mark set at the 60th percentile of the benchmark population.

We spent the last weeks verifying every detail of the new format: the official notices, the SHL product documentation, the official practice assessments — the complete pipeline, item by item. Here is what actually awaits you, and how to turn it into an advantage.

What changed, exactly

Three internal competitions have adopted the new model in 2026:

• COM/AD/107/2026 — Administrators AD7 (generalist, also open to temporary agents): published 31 March, its candidates already sat the new test in June.
• COM/AD/108/2026 — Administrators AD5: applications closed 30 June; the cognitive assessment is expected in September 2026.
• COM/AST/109/2026 — Assistants AST3: the newest one — applications are still open until 31 July 2026, with tests following shortly after.

In all three, the structure is the same: the cognitive assessment comes first, as an eligibility gate. Only candidates at or above the 60th percentile continue to the traditional stages — the EU Knowledge MCQ (30 questions, 30 minutes), a written assignment, and the oral. The EU Knowledge and written stages you know from previous internals are still there; the entrance door is what changed completely.

Inside the 36 minutes

This is the part almost nobody tells you. The assessment mixes three sections in one continuous, adaptive flow — roughly eight questions each:

Deductive reasoning — the section that replaces classic verbal reasoning. You get rules and facts, and you must make everything fit: reschedule a meeting across four colleagues' calendars (client meetings cannot move), fit tasks with durations into the gaps of a day, assign people to rooms respecting capacity and adjacency, work out shift calendars over a month, rank six colleagues from five clues. No options to choose from — you paint the slots, drag the badges, pick the dates.

Inductive reasoning — the closest cousin of abstract reasoning, with a twist: you build the answer. Four frames are shown; you construct the fifth by tapping regions of a figure (each tap cycles blank → green → dotted). Or you extend a sequence of labelled bubbles by literally drawing the line to the next one, once you have cracked the hidden rule.

Numerical reasoning — workplace data (allowance rates, project revenues, team sizes) turned into actions: order six people with ranking badges, approve or reject each tab against a threshold, drag pie-chart handles until the percentages match the facts, set bar totals and splits, plot a monthly series node by node.

Three properties shape the whole experience: the test is adaptive (correct answers bring harder questions — struggling at the end is a good sign, not a bad one), there is no going back (once you press Next, the answer is locked), and there is no negative marking — but leaving items unfinished lowers your score, so you always finish.

Why it catches people off guard

From the official practice materials and the notices, a few patterns stand out — the kind that punish an unprepared first-timer:

• Rules that bind per line, not per total. A meal-allowance item can reject a claim whose total fits comfortably, because one single line exceeds its rate. Read the rule twice; it is worded deliberately.
• Boundary cases on purpose. Project revenue at 2.02× the investment is "Funded"; at 1.98× it is "Deferred". The items are built around the threshold.
• The obvious answer is trapped. In calendar items, the tempting early gaps tend to be client meetings — the one thing you cannot move.
• Directions flip. One ranking asks 1 = highest, the next 1 = lowest. The stem tells you; momentum betrays you.

None of this is hard once you have seen it. All of it is expensive to discover live, with the timer running and no way back.

The calendar is tight — and that is your opportunity

AD7 candidates have already gone through it. AD5 internal candidates sit the test around September. AST3 follows right behind. That leaves a preparation window measured in weeks — and because the score is a percentile, you are effectively compared against colleagues who mostly discovered the format the day of the test.

A percentile system has a beautiful property: preparation compounds. Understanding the mechanics, knowing the item types, having burned the classic traps in practice — every bit moves you up the distribution, and the 60th percentile stops being a wall.

Train for it today — in your language

This is why we built it: EU·Now is the first preparation platform to cover the three SHL-style sections of the Commission's new internal test — deductive, inductive and numerical reasoning — with the verified format, every documented item type, the traps and the method, available in English, French, German, Spanish and Italian.

The SHL Lab is already live on EU-now.com. Try the format right now with the free demo simulator — 12 interactive questions across the three sections, no account needed. And when you are ready to train seriously, the full Internal Competition Prep is open: dedicated trainers for deductive, inductive and numerical reasoning with step-by-step explanations, the full-length adaptive simulator (24 questions, 36 minutes, like the real thing) and the EU Knowledge module — all included in the 2026 Pass: €99 one-time, valid for the whole cycle until 31 January 2027 (see pricing). Being early matters twice here: percentile scoring rewards whoever prepares first. Don't let the September window close on you.

EU·Now is not affiliated with, or endorsed by, SHL Group Ltd or the European Commission. All practice items on our platform are original, created in the style of the assessment; we never reproduce SHL content. Format details verified against the official notices of competition, SHL's public product documentation and the official practice assessments — sources below.