IB MYP + DPAssessment × AI

How IB assessment works in a world with AI

Analysis of all current IB subject briefs – how much must be supervised, how much can be open, and what happens when students use AI. Based on DP briefs (Groups 1–6 + Core) and MYP briefs (10 subjects).

Prepared for International School Stockholm · September 2026 · Source: IB subject briefs “Assessment at a glance” tables + IB AI integrity policy.

The big picture

70–80%typical DP supervised share in exam subjects
20–30%typical DP open coursework (IA)
100%open in TOK, EE, DP Arts
5 / 10MYP subjects fully supervised on-screen
Supervised exam / live oralOpen coursework (take-home)

IB splits everything into External (IB-marked) vs Internal (teacher-marked + IB-moderated). That split is not the same as supervised vs open. Example: a DP Arts portfolio can be External but still 100% open; a language oral is Internal but fully supervised.

Method note: ibo.org blocks bulk scraping (Cloudflare). All numbers below are from the brief PDFs’ assessment tables via IB pages + search-indexed brief text + downloadable school mirrors, cross-checked across SL/HL. Full subject list is in the tables.

DP at a glance – supervised vs open

Supervised = timed written paper + live oral / live performance. Open = coursework over weeks/months, even if teacher-guided. Assumes students use AI where they can.

SubjectSupervised / OpenSplitVerdict
Biology / Chemistry / Physics SL·HL
80 / 20 – P1 36% + P2 44%; IA investigation 3,000 wordsrobust anchor
Maths AA / AI SL·HL
80 / 20 – P1+P2 (+P3 HL 20%); Explorationrobust anchor
Economics / Business / Global Politics / Digital Society
SL 70/30, HL 80/20 – IA portfolio / projectrobust anchor
History / Geography / Philosophy (SL)
75/25 – IA investigation / fieldworkrobust anchor
Psychology new 2027 (SL)
70/30 – IA = proposal onlyIA fragile
Language A Lit / Lang & Lit SL
100/0 – P1 35 + P2 35 + oral 30very robust
Language A HL
80/20 – HL essay openrobust
Language B / ab initio
100/0 – writing + listening/reading + oralvery robust
Classical Languages
80/20 – dossier openrobust
Literature & Performance SL
70/30 – live performance saves itmixed
Computer Science SL / HL
SL 70/30, HL 80/20 – 35h solution openIA fragile
Design Tech SL
60/40 – 40h projectmost open exam subject
ESS SL / SEHS
75/25 (HL 80/20)robust anchor
Visual Arts / Film / Music / Theatre / Dance
0/100 – portfolio + video, no timed paperfully exposed
TOK / Extended Essay / CAS
0/100 – essay / exhibition / projectfully exposed

HL is usually more supervised than SL (extra paper). Exception: Social & Cultural Anthropology HL 75/25 vs SL 80/20. Design Tech SL has the largest open share of any exam subject (40%).

MYP at a glance

100% supervised on-screen (2h, external)

  • Language & Literature
  • Individuals & Societies
  • Sciences (Bio/Chem/Phys/Integrated)
  • Mathematics
  • Interdisciplinary Learning (mandatory for certificate)

Secure browser, rich media, simulation tasks. AI-proof if phones are out.

100% open ePortfolio (teacher + moderated)

  • Arts
  • Physical & Health Education (+3-min video)
  • Design (design folder A–D)
  • Personal Project (~25h, mandatory)

Highly AI-exposed without process evidence + defence.

Hybrid: Language Acquisition – on-screen listening/reading/writing ~75% + moderated live speaking ~25%.

How well does it hold up with AI?

1. Robust – supervised exams & live orals

70–100% of most DP exam subjects + 5 MYP subjects. Timed papers, unseen sources, no internet, plus 10–15 min orals with follow-up questions. A teacher can probe understanding in a way AI cannot fake. Keep these offline and phone-free – they are the validity anchor.

2. Fragile – open IAs worth 20–30%

Science investigations, Maths explorations, Econ commentaries, History investigations, Business projects, CS solutions, DT projects, Geography fieldwork. AI drafts RQs, cleans data, writes analysis, debugs code – exactly the Conclusion/Evaluation marks (often 50% of IA). Does not collapse the diploma (80% exam anchors it) but inflates 1–2 grade boundaries and rewards AI fluency.

3. Very exposed – 100% open

TOK essay/exhibition, EE, all DP Arts, MYP ePortfolios + Personal Project. Pure take-home prose/portfolios are LLM-native tasks. Citation rules don’t solve paraphrased drafting. Practical/video parts (playing, performing, making) are better than written rationales – embodiment is harder to fake – but artist statements and comparative studies are easily polished.

Paradox: the most IB-aligned work – inquiry, process, reflection, creativity – is the least defensible as take-home assessment. IB’s own defence (work over time + check-ins + viva) only works if schools enforce it.

What schools should do now

  1. Viva everything open. No IA / EE / TOK / portfolio passes without a 10-minute oral defence. Fail incoherent explanation, not polished prose.
  2. Witness the process. RQ approval in class, data collection witnessed, first draft started in class, commits/photos/rehearsal video required.
  3. Traffic-light AI rules per task. Green: brainstorm/grammar with disclosure + bibliography entry. Red: no AI in supervised work. Teach bias-checking as ATL skill.
  4. Weight supervised work locally. IB final weights are fixed, but predicted grades and internal reports can overweight supervised formatives.
  5. Design for AI. Prefer personal/local data, live sources, in-class stimuli; assess decisions and revisions, not just final polish.

IB policy reference: IB will not ban AI; AI use must be credited; school must authenticate work as the student’s own; human examiners remain (IB AI in learning, teaching and assessment).

Bottom line

DP and MYP survive AI because the supervised core survives – roughly three-quarters of most subjects. The open layer (20–30% IAs, plus 100%-open core, Arts, and MYP portfolios) does not survive as pure take-home product. Assume AI was used there, and assess what AI cannot do for the student: explain live, show process, defend choices.