There are two related aspects to this pattern

Firstly, many web sites that are using frameworks or content management systems that are either difficult to customize or are operated by organizations that have little or no in-house technical staff. How can these sites be updated, with a minimum of effort, to support publishing of structured data?

Secondly, documents or files may be shared and copied across the web. This can result in a broken connection between the document and it's metadata which might only be available from the source website. How can the metadata be made available to anyone discovering a copy or mirror of the document?

Embed the structured data directly in the document itself rather than, or in addition to, publishing that data separately.

The most common scenario here is instrumenting an existing webpage to add some "semantic markup" that will enable a client to extract data directly from the HTML source. This typically involves changing just the templates used ot generate the web pages. By changing a small number of resources, it becomes possible to quickly and easily publish data about a large number of resources.

A less common scenario involves embedding data within a different document format. Typically this relies on using an existing extension mechanism that has been defined for that format. A tool may then inspect the file directly to discover the metadata or a link to its location.

At the time of writing there are a number of competing proposals for embedding metadata in XHTML/HTML documents, including RDFa, microdata and microformats. RDFa can also be used to embedded metadata in other XML formats

Options vary for embedding metadata in other formats, but Adobe XMP provides an option that can be used in a variety of formats.

Embedding metadata into existing resources, rather than requring changes to how content and data is published to the web is often much easier to achieve. At the time of writing there are a number of competing approaches for embedding metadata in XHTML/HTML. These typically offer the same basic features but vary in how much they require specific markup extensions in the original document.

While it can be very easy to quickly instrument a website with embedded metadata, the requirement to introduce all data into the same template means that it can become awkward to manage the competing concerns: e.g. clean, well-structured markup for styling and presentation, versus rich, detailed, semantic markup for machines. For more complex use cases its may be better to simply publish the metadata separately, or embedded only a subset of the data with links to additional resources.