Most of an XML document is elements and text, but three other constructs show up regularly, and each has a specific job. Knowing them on sight stops them from reading as noise.
CDATA sections
A CDATA section holds text that should be taken literally, without the parser interpreting < and & as markup. It is written <![CDATA[ ... ]]>, and everything between the brackets is raw character data:
<summary><![CDATA[Uses <tags> & symbols freely]]></summary>
Without CDATA, that content would have to be escaped as Uses <tags> & symbols. CDATA is a convenience for blocks that contain a lot of characters that would otherwise need escaping, such as embedded code or markup samples. The one thing it cannot contain is the sequence ]]>, since that is what closes it. Importantly, CDATA changes only how the text is written, not what it means: the parsed content is identical to the escaped version.
Comments
A comment is written <!-- ... --> and is ignored as content. It exists for human readers, to annotate or temporarily disable part of a document. Comments cannot be nested and cannot contain a double hyphen -- inside them. Because a comment is not data, a well-behaved consumer never treats its text as part of the document's information.
Processing instructions
A processing instruction (PI) carries a directive for an application that consumes the XML, written <?target data?>. The target names who the instruction is for, and the data is whatever that application understands. The classic example is a stylesheet association: <?xml-stylesheet type="text/css" href="style.css"?>. The XML declaration at the top of a file looks like a PI but is technically a special case, not one.
Why they are grouped
These three share a common trait: none of them is an element, so none of them appears in the element tree as a node with children. They sit alongside elements as siblings of a sort, annotations and instructions and literal-text escapes woven through the document. When you are mapping out a document's structure, it helps to set them aside from the element hierarchy: the elements carry the data model, while comments, PIs, and CDATA are, respectively, notes, directions, and a writing convenience.