AI for AKM Preparation in Indonesia
AI helps Indonesian teachers prepare students for AKM by turning ordinary subject content — a Science passage, a Social Studies case, a Math word problem — into reading-literacy and numeracy stimulus items in the exact format AKM tests, since the assessment measures cross-curricular competency rather than memorized subject facts.
That reframing matters more than any single tool choice. A teacher generating "AKM practice" that looks like a subject quiz is practicing the wrong skill entirely.
Quick Answer: AKM (Asesmen Kompetensi Minimum) tests reading literacy and numeracy through short stimulus texts and applied problems, sampled from students in Grades 5, 8, and 11 rather than sat by every student. AI's best use is converting whatever a class is already studying into AKM-style literacy or numeracy items — not drilling isolated facts. Indonesia's PISA 2022 scores (reading 359, mathematics 366, science 383, each well below the OECD average) are part of why AKM was designed around applied competency in the first place (OECD, 2023).
Indonesia's Asesmen Nasional (National Assessment) replaced the old Ujian Nasional (UN) starting in 2021, after a 2020 pilot was delayed by the pandemic. AKM is its core cognitive component, run alongside two non-cognitive surveys that round out the picture of a school.
This guide sits inside the broader AI in Education Around the World: A 2026 Regional Guide and focuses specifically on AKM: what it measures, and where AI realistically helps a teacher prepare for it.
What AKM Actually Measures
AKM tests two domains — Literasi Membaca (reading literacy) and Numerasi (numeracy) — and neither is tied to a specific subject's syllabus. Every item is built from three dimensions working together: content, cognitive process, and context.
| Dimension | Reading Literacy | Numeracy |
|---|---|---|
| Cognitive process | Finding information, interpreting and integrating, evaluating and reflecting | Understanding (pemahaman), applying (penerapan), reasoning (penalaran) |
| Context | Personal, socio-cultural, scientific | Personal, socio-cultural, scientific |
| Format | Stimulus text plus a set of linked questions | Applied problem, often with a real-world stimulus |
Why "Content-Free" Doesn't Mean "Anything Goes"
AKM items aren't tied to one subject's curriculum, but they still need real, coherent content behind them — a passage about a real phenomenon, a problem grounded in a plausible scenario. Generic, made-up filler text produces a weak stimulus, even when the question format is technically correct.
Five Question Formats, Not Just Multiple Choice
Unlike a typical subject quiz, AKM items appear in five formats: Pilihan Ganda (multiple choice), Pilihan Ganda Kompleks (complex multiple choice, more than one correct option), Menjodohkan (matching), Isian Singkat (short answer), and Uraian (a short essay response). Practice material that only uses simple multiple choice misses most of what a real AKM item looks like.
How AKM Is Actually Delivered
AKM runs on ANBK (Asesmen Nasional Berbasis Komputer), Indonesia's computer-based national assessment platform. Schools with enough working computers test in a fully online mode; schools that don't often use a semi-online mode, downloading the test package to a local server in advance and syncing results once a connection is available.
- Some smaller schools without a working computer lab arrange to test at a nearby school with better facilities, a practice commonly called menumpang.
- The semi-online option means a shaky internet connection on testing day doesn't automatically block a school from participating.
- AI-generated practice material should mirror this reality: exportable, printable sets matter as much as anything designed for an on-screen quiz format.
How AKM Differs From a Normal Exam
AKM is deliberately not a pass/fail test for any individual student, and that single design choice changes what "preparation" should even mean for a teacher.
| Feature | Ujian Nasional (pre-2021) | AKM (2021–present) |
|---|---|---|
| Who sits it | Every student in a graduating cohort | A sample of students in Grades 5, 8, and 11 |
| What it tests | Memorized subject-specific content | Cross-curricular reading literacy and numeracy |
| Individual stakes | Fed directly into a student's own academic record | None — no individual score is reported back |
| Purpose | Ranking students and schools | Diagnosing how well a school is building foundational competency |
The Four Proficiency Bands
Results are reported in four bands per domain, at the school level rather than the individual level: Perlu Intervensi Khusus (needs special intervention), Dasar (basic), Cakap (proficient), and Mahir (advanced). A school's distribution across these bands, not any one student's answer sheet, is what actually gets used.
Where the Results Actually Go
AKM results feed into Rapor Pendidikan (the Education Report Card), a platform schools and local education offices use to see their own literacy and numeracy standing and plan improvement. A teacher generating AKM-style practice is ultimately contributing to that school-level picture, not chasing an individual grade.
The Two Non-Cognitive Surveys Alongside AKM
AKM is only one of three instruments inside the Asesmen Nasional, and the other two shape how a school's results actually get read. Neither is something AI-generated academic practice can substitute for, but understanding them clarifies what AKM prep is really contributing to.
| Instrument | Who Completes It | What It Captures |
|---|---|---|
| AKM | Sampled students | Reading literacy and numeracy competency |
| Survei Karakter | Sampled students | Alignment with the Profil Pelajar Pancasila (character and values framework) |
| Survei Lingkungan Belajar | Students, teachers, and the principal | School climate, safety, and instructional quality |
Why This Context Matters for Teachers
A school's Rapor Pendidikan reads all three instruments together, not AKM in isolation. A class strong in literacy and numeracy but reporting a weak learning-environment picture still shows up as a school needing attention — a reminder that AI-generated AKM practice supports one piece of a larger diagnostic picture, not the whole of it.
Where AI Genuinely Helps With AKM Prep
AI's clearest use in AKM preparation is converting existing lesson content into a properly formatted stimulus-and-question set — something that would otherwise take real time to build from scratch for every unit, in every subject.
| Prep Task | Traditional Approach | AI-Assisted Approach |
|---|---|---|
| Building a literacy stimulus | Search for an existing passage that fits the topic | Draft a fresh, content-accurate passage at the right reading level |
| Practicing numeracy reasoning | Rework a textbook word problem by hand | Generate several applied variations at a stated difficulty |
| Covering all five question formats | Write each format manually, one at a time | Request a mixed-format set matched to a single stimulus |
| Explaining a wrong answer | Re-mark and hope the student sees the pattern | Ask for a walkthrough of the specific reasoning error |
Turning Any Subject Into a Literacy or Numeracy Item
Because AKM is cross-curricular, a Science lesson on plant growth or a Social Studies unit on local government can both become a literacy stimulus — a short passage plus questions testing inference — without touching the subject content itself. The skill being tested is reading, not photosynthesis or civics.
The same logic applies to numeracy: a Social Studies lesson on a regional population table can become a numeracy item asking a student to interpret and reason from the data, not just recall the population figure.
Explaining the Reasoning, Not Just the Answer
AKM's cognitive-process dimension rewards evaluation and reasoning over recall, so a bare answer key undersells what the assessment actually measures. Asking an AI tool to explain why a specific answer choice reflects weaker reasoning — not just which option was correct — builds the transferable skill AKM is designed to detect.
A Practical AKM Prep Workflow
AKM preparation works best woven into regular subject teaching across the year, not run as a separate track competing for lesson time.
- Pick a piece of content you're already teaching — a passage, a data table, a real scenario — rather than starting from a blank prompt.
- Decide which domain and cognitive process you're targeting, since a vague request produces a vague, hard-to-classify item.
- Generate a stimulus plus a mixed set of question formats, not just multiple choice, so students see the full range AKM actually uses.
- Ask for a plain-language explanation of any wrong answer, framed around the reasoning error, not just the correct choice.
- Rotate contexts — personal, socio-cultural, and scientific — so practice doesn't drift toward one context repeatedly.
- Check the finished item against your current Kurikulum Merdeka learning outcomes (Capaian Pembelajaran) for the subject it was drawn from, so the content stays accurate even though the skill being tested is cross-curricular.
Why the Framework Stays Stable Across a Curriculum Transition
Many Indonesian schools are still moving between Kurikulum 2013 and Kurikulum Merdeka at different paces, and that transition can make subject-content planning feel unsettled from term to term. AKM's own framework doesn't shift with it — reading literacy and numeracy are measured the same way regardless of which curriculum a school is mid-transition on.
That stability is genuinely useful for AI-assisted prep. A stimulus-and-question set built around the content, process, and context dimensions above stays valid even if the specific subject unit it was drawn from gets resequenced the following term.
A Concrete Illustration: A Grade 5 SD Reading-Literacy Item
Say you teach Grade 5 at an SD (elementary school) and your class just finished a Science unit on the water cycle. You could ask an AI tool for a 120-word passage about a local flooding event, framed in a socio-cultural context, followed by a mixed set: two multiple-choice items testing recall of the sequence, one complex-multiple-choice item testing inference about cause, and one short-answer item asking the student to justify their reasoning.
That single passage gives you literacy practice across three cognitive processes without writing three separate exercises. EduGenius can generate a stimulus-and-question set like this from a class profile specifying Grade 5 and the source subject, exported as a printable worksheet with an answer key.
A Second Illustration: A Grade 8 SMP Numeracy Item
Now say you teach Grade 8 at an SMP (junior secondary school), and a Social Studies lesson covered a table of provincial population figures. You could ask an AI tool to turn that same table into a numeracy item requiring students to calculate a rate, compare two provinces, and justify which comparison is misleading if read carelessly.
That last part — spotting a misleading comparison — sits squarely inside AKM's reasoning process, and it's a genuinely different skill from simply reading a number off a table. Reusing content students already know keeps the numeracy practice grounded instead of abstract.
A Third Illustration: A Grade 11 SMA Mixed-Format Set
Say you teach Grade 11 at an SMA (senior secondary school), and you want your sampled students genuinely comfortable with all five AKM formats before assessment day, not just multiple choice. Take a single Language Arts passage already in your term's plan.
You could ask an AI tool to build one item in each format from that same passage — a Pilihan Ganda Kompleks item on its main argument, a Menjodohkan item pairing claims with supporting evidence, and a short Uraian prompt asking students to evaluate the argument's weakest point.
Working through one passage across all five formats, rather than five different passages in one format each, tends to build format familiarity faster — and it keeps the underlying reading skill, not the format itself, as the actual focus of the practice session.
Building a Reusable Bank Across a School
A single well-built stimulus set is useful once. Saved and tagged by domain, cognitive process, and grade level, it becomes useful across parallel classes and future cohorts, since AKM's underlying framework doesn't change even as the specific subject content teachers draw from does each term.
Choosing Tools for AKM Preparation
Different tools solve different pieces of AKM prep, and no single one replaces a teacher's judgment about whether a generated item actually fits the framework.
| Tool Type | Strength | Limitation |
|---|---|---|
| General AI chat assistant | Flexible, good for one-off passages or explanations | Format and cognitive-process labeling drift without careful, repeated prompting |
| Dedicated content-generation platform | Consistent, exportable, mixed-format sets from one class profile | Narrower scope than an open-ended conversation |
| Official Kemendikdasmen AKM sample materials | Most accurate calibration of real item style | Limited quantity, updated on a fixed schedule |
Where EduGenius Fits
EduGenius is designed to generate differentiated practice materials — including quizzes, worksheets, and short passages — from a described class profile, with answer keys included automatically and export to PDF, DOCX, or PowerPoint for offline printing. For a teacher building AKM-style practice across several subjects, that adaptability is a genuine time-saver over drafting each stimulus by hand.
Official sample materials published by Indonesia's education ministry remain the most format-accurate reference, since they're built by the body that designs the real assessment — AI-generated practice works best as a volume supplement between them, not a replacement.
Related Contexts Across This Series
AKM is one of several non-traditional, sample-based assessments covered across this series. For a broader regional view, see AI in Education Around the World: A 2026 Regional Guide, and for a genuinely individual, high-stakes contrast, AI for ECAT and Engineering Entry Tests covers Pakistan's engineering entrance exam.
Kenya runs a comparable diagnostic-first, no-individual-stakes assessment of its own at the primary level — see AI for KPSEA Preparation in Kenya for how that comparison plays out. For a very different, high-stakes board-exam landscape, AI Tools for CBSE vs ICSE vs State Board Syllabus covers India's parallel school-board systems, and AI Lesson Planning for Large Pakistani Classrooms addresses a class-size challenge common across the region. For the numeracy domain specifically, Best AI for Math Problems in 2026 (Benchmarked) compares how different AI tools actually handle applied problem-solving.
Pro Tips for AI-Assisted AKM Preparation
- Always start from real content your class already covered — a generated stimulus grounded in a genuine topic reads more naturally than one built from scratch.
- Name the cognitive process you're targeting, not just "make a reading question," so the output actually tests inference or evaluation instead of recall.
- Request all five question formats across a unit, not just multiple choice, so students aren't surprised by Uraian or Menjodohkan items on the real day.
- Ask for the reasoning behind a wrong answer, since AKM rewards evaluation over recall far more than a typical subject test does.
- Cross-check generated items against Kemendikdasmen's official sample materials periodically, since item style can shift slightly between cycles.
- Rotate personal, socio-cultural, and scientific contexts deliberately, rather than letting every generated item default to the same setting.
- Remember AKM is school-level, not individual, so frame practice as competency-building for the whole class rather than pressure on the sampled few.
What to Avoid
- Generating subject-content quizzes and calling them AKM practice. A Science fact-recall question is not a literacy or numeracy item, no matter how well-written it is.
- Sticking to multiple choice only. AKM uses five distinct formats, and unfamiliarity with Uraian or Menjodohkan items costs real time on assessment day.
- Treating a sampled student's performance as a personal grade. AKM carries no individual stakes, and framing it that way creates unnecessary pressure.
- Skipping the reasoning-explanation step. A bare correct-answer key undersells what a cognitive-process-based assessment is actually designed to build.
Key Takeaways
- AKM (Asesmen Kompetensi Minimum) tests reading literacy and numeracy, sampled from Grades 5, 8, and 11, as the cognitive core of Indonesia's Asesmen Nasional.
- It replaced the Ujian Nasional starting in 2021, shifting from a census-based, subject-content exam to a sample-based, cross-curricular competency assessment.
- Every item combines three dimensions — content, cognitive process, and context — and uses one of five formats, not just multiple choice.
- AKM runs alongside two non-cognitive surveys — Survei Karakter and Survei Lingkungan Belajar — and a school's Rapor Pendidikan reads all three together, not AKM in isolation.
- Results are reported in four school-level bands (Perlu Intervensi Khusus, Dasar, Cakap, Mahir) and feed into the Rapor Pendidikan platform, not an individual grade.
- AI's strongest role is converting existing subject content into properly formatted literacy or numeracy stimulus items, not generating standalone subject quizzes.
- Indonesia's PISA 2022 results (reading 359, mathematics 366, science 383) sit well below the OECD average, part of the rationale behind AKM's competency-based design (OECD, 2023).
- Explaining the reasoning behind a wrong answer matters more here than in a typical exam, since AKM's cognitive-process dimension explicitly rewards evaluation over recall.
- EduGenius can generate a stimulus-and-question set from a class profile, useful for building AKM-style practice across several subjects without starting from scratch each time.
Frequently Asked Questions
What does AKM stand for, and what does it test?
AKM stands for Asesmen Kompetensi Minimum (Minimum Competency Assessment), the cognitive core of Indonesia's Asesmen Nasional. It tests reading literacy and numeracy — applied, cross-curricular skills — rather than memorized content from any single subject.
Does every student in Indonesia take AKM?
No. AKM samples students from Grades 5, 8, and 11 rather than testing every student, and results are reported at the school level. No individual student receives a personal AKM score.
How is AKM different from the old Ujian Nasional?
The Ujian Nasional was a census-based exam testing memorized subject content, with results feeding a student's own academic record. AKM samples students instead of testing everyone, tests cross-curricular reading literacy and numeracy instead of subject recall, and carries no individual stakes at all.
Can AI-generated content be used directly for AKM practice?
It can be a strong starting point if it's built from real content the class already covers and matched to AKM's specific format and cognitive-process framework, but it should be checked against Kemendikdasmen's official sample materials before being treated as fully calibrated practice.