What This Course Teaches

By the end of this course you should be able to open an iXBRL document, scroll for thirty seconds, and form a sharp opinion. Not "the file is valid", which is what a validator says. Something stronger: "the period type on this concept is wrong", or "the unit and the concept disagree", or "this extension was anchored to an abstract parent and the linkbase is going to fail at the regulator".

The path runs through ten chapters. The first three set up vocabulary you cannot skip: why structured reporting exists, what an XBRL fact actually is, and how concepts are declared in a schema. After that we move into linkbases (the five wires that hold a taxonomy together), dimensions and contexts (the grammar of "this number, this period, this segment"), Inline XBRL hosting rules (escape, hidden, continuation, transformation), validation profiles for ESEF and the EDGAR Filer Manual, prompt patterns for steering AI agents that author iXBRL on your behalf, and finally an integrated reading exercise that combines everything.

One mental model recurs so often that it is worth naming on day one: the graph. A taxonomy is not a list, it is a network. Concepts are the dots, and the relationships between them, this is a child of that, this adds up to that, this is labelled in Dutch, this is anchored to that base concept, are the lines that connect the dots. Almost every hard question in later chapters turns out to be a question about that network: which dots, which lines, which direction the line points. Chapter 4 makes the idea explicit and gives you the vocabulary (vertices, edges, directed, acyclic) that the linkbases, the Discoverable Taxonomy Set, and even the closing comparison to knowledge graphs all lean on. You do not need any of it yet. It is enough to know that the picture in your head should be a web, not a column.

What you will not learn is hand-authoring linkbases byte by byte. The market has decided that XML editors are not how production filings get written. AI agents and tooling do that work. Your job is to read the output well enough to spot when the agent is wrong, and to write a follow-up prompt sharp enough to fix it.

Two pedagogical scaffolds run through every chapter. Each section is tagged with a Bloom level (remember, understand, apply, analyze, evaluate, create), so you can tell at a glance whether the question is asking you to recall a definition or to make a judgement call. And each chapter ends with a short boss-level test where prior-chapter ideas combine with the current chapter; the questions are intentionally harder than the section quizzes and exist to confirm you can actually move between concepts, not just within them.

The final exam grades on a 0 to 10 scale. Pass it and you walk out with two things: a certificate, and the working memory to read structured reports the way analysts and auditors already do.

• You will learn to read iXBRL output and tell whether a number, a context, a unit, and a concept choice make sense together.

• You will learn to recognise the most common violations on sight rather than memorising every error code.

• You will learn how to talk to AI agents that generate iXBRL, including how to write prompts that produce valid output and prompts that ask an agent to fix specific violations.

• You will not learn to hand-write linkbases. You will learn to read them well enough to spot when they are wrong.

• Every chapter ends with a short boss-level test that mixes prior concepts. The final exam covers the whole course on a 0 to 10 scale.

Boss-level question

A multi-concept multiple-choice question at the end of each chapter that combines that chapter's material with at least one prior chapter's concepts.

Bloom's taxonomy

A staged hierarchy of cognitive levels (remember, understand, apply, analyze, evaluate, create) used here to scaffold sections from recall to judgement.