It can be thought of as a tree of elements--containers of information.
Traditionally, from a markup perspective:
Articles, books, notes, poems, novels
Technical manuals, slip sheets, product packaging
More abstractly, documents are instances of information that typically have structure.
So, lots of things are documents:
Messaging, e-mails, web sites, etc.
Business transactions, invoices, statements, etc.
Log files, configuration files, install scripts, etc.
Ontologies, medical lab data, instrument experiment results, human genome annotations, etc.
XML is a useful common syntax for encoding these documents.