Every topic page lists the papers that assess one topic. A paper gets there in one of two ways, and each was checked on its own.
Every question part of IAL Mathematics (all 258 papers, 16 units) and IAL Physics (all 128 papers, 6 units) was read with its mark scheme and tagged with the topics its own marks credit, from the topics the unit’s specification sets out. A part that only uses a result found earlier, or only sets up a later part, keeps the topics of its own marks.
So on these papers a topic page lists the questions, with their marks, not just the paper. “needs (a)” names an earlier part the listed parts build on, and “then (c)” a later part that continues from them, so a question can be worked through whole.
IAL Mathematics. Blind checks of 350 randomly chosen question parts (270 from Mechanics M1–M3, 80 from the other units), by readers who had not seen the tags, found 343 tagged with the right topic (95% confidence interval 95.9–99.0%). For the units other than Mechanics, second readers re-read 52 of the 190 papers independently and found 11 tags, of 1,320, that the first reading had missed.
IAL Physics. Blind checks of 160 randomly chosen question parts, 80 from Unit 4 and 80 from the other five units, by readers who had not seen the tags, found 155 tagged with the right topic (95% confidence interval 92.9–98.7%). Second readers re-read 33 of the 128 papers independently and found 20 tags, of 1,173, that the first reading had missed. On the practical papers, Units 3 and 6, every mark spent on an experiment counts towards that experiment’s topic.
Subjects not yet read are matched automatically: a paper is listed under a topic when the wording of its questions or its mark scheme matches the patterns written for that topic. That finds practice quickly, but it cannot say which question, and it misses papers whose wording does not match.
International A Level. A hand-check of 65 randomly chosen matches, read against the source question papers, found 53 that genuinely assess the topic and 12 that do not — 81.5% (95% confidence interval 70.4–89.1%). That is a good way to find practice on a topic, not proof a paper covers it, and a paper missing from a topic’s list may still assess it. A separate check found real papers that assess a topic without being tagged for it.
International GCSE. A hand-check of 100 randomly chosen matches, each read against the source question paper and the specification's own statements for the topic, found 99 that genuinely assess the topic and 1 that does not — 99% (95% confidence interval 94.6–99.8%). That is a good way to find practice on a topic, not proof a paper covers it, and a paper missing from a topic’s list may still assess it. The topics are the specification’s headings, which are broad: one question on any statement under a heading was enough.
On a read unit’s page, each topic shows how many of the unit’s papers have a question on it, and how many marks those questions carried on a paper: the fewest, the most and the average. A question part on two topics counts towards both, so the topics of one paper can add up to more than its total.
A paper missing from a topic page has not been shown not to assess it. On a read unit the second readings above found few missed tags; on a keyword-matched one a miss is common, and a separate check found real papers that assess a topic without being matched to it.