IB Math AI SL Past Paper Analysis 2021–2025 + May 2026 Predictions

Methodology: Every question from 40 IB Math AI SL papers (Paper 1 and Paper 2) across 20 exam sittings — 2021–2025 plus the specimen — has been analysed: 353 questions, 3,200 marks. Marks are allocated to the 5 IB syllabus topics. When a question spans multiple topics, marks are distributed equally to prevent double-counting.

SL Paper Structure: Both AI SL papers allow a GDC. Paper 1 — 80 marks, 90 minutes, short-response, technology required. Paper 2 — 80 marks, 90 minutes, extended-response, technology required. Total: 160 marks per sitting. No Paper 3 at SL.

What IB Actually Tests — Sub-Topic Breakdown

16 syllabus sub-topics ranked by % of total marks. Colour-coded by IB syllabus topic — this is the chart that matters most for revision planning.

Statistics & Probability Leads — 31.2% of All Marks

AI is the data-driven Math route, and the numbers show it: nearly a third of every paper is statistics and probability. Here's how those marks break down.

Statistics & Probability Sub-topics

Statistics & Probability: P1 vs P2

Bivariate statistics and regression lean heavily on Paper 2. Hypothesis testing (χ² and t-tests) and probability are spread evenly — the GDC does the heavy lifting on both papers.

5 Syllabus Topics — Overview

The 5 broad IB syllabus topics and their marks share across every AI SL paper.

Marks Share by Topic

Paper 1 vs Paper 2

Final Long-Question Appearances by Topic

AI papers are not split into Sections A/B — but each Paper 2 ends with extended investigation-style questions. This counts how often each topic anchors those final long questions. Differentiation leads with 14 appearances.

Year-over-Year Trends (questions per year)

Geometry, Trig & Algebra — Sub-topic Deep Dive

Breaking down the other high-weight topics into their tested components.

Geometry & Trigonometry Sub-topics

Number & Algebra Sub-topics

What Stands Out

Five years of AI SL data, six findings worth memorising before May 2026.

Biggest Topic

Trigonometry tops the paper at 13.5%

431 marks across 48 questions — the single most-tested sub-topic. Sine/cosine rule, 3D solids and bearings recur in virtually every sitting, on both papers.

Biggest Group

Statistics & Probability is 31% of the exam

Distributions, hypothesis testing (χ² and t-tests) and regression together dominate — the defining, GDC-driven character of the AI route.

Calculus

Differentiation is the calculus workhorse

10.6% of all marks (338 marks) — far ahead of integration at 5.5%. Rates of change and optimisation drive nearly every calculus question.

Unique to AI

Financial maths carries 5.9%

Loans & annuities account for 189 marks and lean on the GDC finance solver — almost always on Paper 2. A reliable, high-frequency AI-only topic.

Distinctive

Voronoi diagrams hold 5.0%

The graph-theory / geometry topic unique to AI appears as full long questions, worth up to 19 marks — a common anchor for the toughest exam item.

Long Questions

Differentiation leads the final long questions

14 appearances as the extended investigation, ahead of trigonometry and integration (8 each). Expect a calculus-heavy finale on Paper 2.

May 2026 Predictions — Math AI SL

Paper 1 (GDC · short response)

Trigonometry: sine/cosine rule, 3D solids or bearings
Functions & models: linear, exponential or piecewise
Finance or sequences: annuities / geometric growth
Differentiation: gradient, tangent, optimisation

Paper 2 (GDC · extended)

Statistics: a distribution + a χ² or t hypothesis test
Bivariate stats: regression line + prediction
Voronoi / graph theory: a full modelling question
Calculus finale: differentiation/integration (15–20+ marks)

Style — applied throughout

Real-world contexts on every question
GDC expected on both papers — no no-calculator section
Accessible "Write down" / "Find" entry parts
Extended final questions worth 15–20+ marks

Watch These Topics

Trigonometry: the single most-tested topic (13.5%)
Hypothesis testing: χ²/t-tests rising — 7 Qs in 2025
Differentiation: the reliable long-question topic
Loans & annuities: financial maths on nearly every P2

Student Difficulty Heat Map

How well candidates handled each AI SL topic, session by session, drawn straight from what the examiners flagged. Synthesised from 17 official IB examiner subject reports (N21–N25).

Topic	N21	M22 T1	M22 T2	N22	M23 T1	M23 T2	N23 T1	N23 T2	M24 T1	M24 T2	N24 T1	N24 T2	M25 T1	M25 T2	M25 T3	N25 T1	N25 T3
Trigonometry	Mixed	Poor	Poor	Good	Mixed	Mixed	Good	Mixed	Mixed	Good	Poor	Mixed	Good	Poor	Mixed	Good	Poor
Hypothesis Testing	Mixed	Poor	Poor	Poor	Poor	Good	Poor	Poor	Mixed	Mixed	Mixed	Poor	Good	Mixed	Mixed	Poor	Mixed
Functions	Poor	Mixed	Mixed	Mixed	Poor	Poor	Poor	Poor	–	Poor	Poor	Mixed	Poor	Poor	Poor	Mixed	Poor
Integration	Mixed	Poor	Good	Mixed	Mixed	Poor	Mixed	–	Poor	Poor	Mixed	Mixed	Good	Mixed	Poor	Mixed	Mixed
Probability Distributions	–	Poor	Mixed	Poor	Mixed	Mixed	Mixed	Mixed	Mixed	Poor	Mixed	Poor	Mixed	Poor	Poor	Mixed	Poor
Loans & Annuities	Poor	–	Poor	Poor	Poor	Mixed	Poor	Mixed	Mixed	Mixed	Poor	Mixed	Mixed	Mixed	Poor	Mixed	Poor
Differentiation	Poor	Poor	Poor	–	Poor	–	Poor	Poor	Mixed	Poor	Poor	Poor	Mixed	Poor	Good	Mixed	Poor
Univariate Statistics	Mixed	Mixed	Mixed	Good	–	Poor	Poor	–	Poor	Poor	Poor	Mixed	Poor	Mixed	Poor	Mixed	Mixed
Probability	Mixed	Mixed	Mixed	Poor	Mixed	Poor	Poor	Poor	Poor	Poor	Poor	–	–	Poor	–	Poor	Mixed
Sequences & Series	Mixed	Mixed	Mixed	Good	–	Mixed	–	–	Good	Mixed	Poor	Mixed	Good	Good	Good	Good	Mixed
Bivariate Statistics	Good	Mixed	–	–	Mixed	Mixed	Good	Good	Good	–	Poor	Poor	Good	Poor	–	Mixed	Mixed
Trig / Sinusoidal Graphs	Mixed	Poor	Poor	Poor	Poor	–	–	–	Poor	–	Poor	Poor	–	Poor	Poor	Poor	Poor
Voronoi Diagrams	Poor	–	Mixed	Mixed	–	Mixed	Poor	Mixed	Mixed	–	–	–	Mixed	Mixed	Mixed	Mixed	Mixed
Approximation & Error	Mixed	–	–	–	Poor	Good	Poor	Poor	Poor	–	–	Mixed	–	Poor	Mixed	Mixed	Poor
Exponents & Logarithms	Mixed	Poor	Mixed	Poor	Mixed	–	–	–	–	Poor	Poor	–	Poor	–	Mixed	–	Mixed

Well done Mixed Poorly done Not tested

Common Errors & Misconceptions

The ten mistakes IB examiners flag most often across five years of AI SL subject reports — the marks students give away before they even reach the maths.

Error 1

Premature rounding & wrong accuracy

By far the most-flagged error every session: rounding or truncating intermediate values, then failing to give the final answer correct to 3 significant figures. Keep unrounded values on the GDC and round only at the very end.

Error 2

'Show that' worked backwards

Starting from the given result and verifying it, rather than deriving it. Circular reasoning earns no credit — you must progress towards the stated answer using independent working.

Error 3

Hypotheses stated imprecisely

Vague or reversed χ² and t-test hypotheses that confuse population with sample, omit the population mean, or state the significance level only after computing the p-value. Define parameters precisely before testing.

Error 4

Final answer in the wrong form

Giving a coordinate pair when one value is required, leaving fractions unsimplified, using calculator notation (e.g. 2.51e-07), or a decimal where an exact value in terms of π was asked. Match the form the question demands.

Error 5

Conclusions not in context

Writing a generic "reject H₀" or "there is a correlation" instead of interpreting the result in the scenario given. Every hypothesis-test and statistics conclusion must reference the real-world context.

Error 6

Missing or unconverted units

Omitting units altogether, or failing to convert them — especially the squared/cubed factors when moving between cm²/m² or cm³/m³. This routinely costs the final accuracy mark.

Error 7

Probability rules confused

Adding probabilities instead of multiplying along tree branches, treating dependent (without-replacement) events as independent, mishandling "given" conditions, and mixing up binompdf with binomcdf. Some answers even exceed 1.

Error 8

No working shown

Giving a bare answer — often straight from the GDC — so that a wrong result scores zero with no method marks recoverable. Always record the values and steps that lead to the answer.

Error 9

χ² and t-test procedure slips

Confusing the χ² goodness-of-fit test with the test for independence, omitting the expected matrix or degrees of freedom, and choosing the unpooled instead of pooled setting for the t-test p-value.

Error 10

Finance solver mis-set

Not giving PV and FV opposite signs in the GDC finance app, forgetting to set FV = 0, or using the compound-interest formula where the financial application was expected. Small sign slips derail the whole calculation.

IB Grade Boundaries — Math AI SL

Official IB grade boundaries (overall, % of total marks). Source: official IB subject reports, N21–N25 TZ3 session.

Grade 7 Boundary Over Time (% required)

Full Grade Boundary Table — Math AI SL (All Sessions N21–N25 TZ3)

Session	G2	G3	G4	G5	G6	G7
N25 TZ3	13	28	45	58	70	83
N25 TZ1	13	26	39	53	67	80
M25 TZ3	11	22	35	49	63	77
M25 TZ2	9	18	27	40	53	66
M25 TZ1	10	20	29	42	55	70
N24 TZ2	14	27	42	57	70	82
N24 TZ1	14	27	42	57	70	82
M24 TZ2	12	24	34	47	61	76
M24 TZ1	13	25	38	51	65	77
N23 TZ2	12	21	34	48	64	78
N23 TZ1	12	21	34	48	64	78
M23 TZ2	11	22	33	48	63	77
M23 TZ1	9	19	30	43	58	71
N22	12	19	32	47	62	77
M22 TZ2	8	14	23	37	53	67
M22 TZ1	6	13	23	35	50	67
N21	5	13	25	38	52	68

Source: official IB subject reports (grade-boundaries page), N21–N25 TZ3. Values are the minimum overall mark (out of 100) for each grade.

Grade 7

66%–83% range for a 7

AI SL Grade 7 boundary ranged from 66% (May 2025 TZ2) to 83% (Nov 2025 TZ3). Average across all sessions: 75%.

Grade 5

35%–58% gets you a 5

AI SL Grade 5 boundary ranged from 35% to 58%, averaging 47% of total marks — reachable by securing the routine method and calculator marks.

Trend

Boundaries climbing since N24

The most recent sessions (N24–N25) sat at the top of the range (76%–83% for a 7), well above the 75% long-run average. If that harder-mark style holds, expect a Grade 7 near — or above — 75% for May 2026.

IB Math AI SL Past Paper Analysis 2021–2025