A publisher adds a new article, but the feed still lists last month's entries. The page exists; the small document meant to announce updates has fallen behind. Checking the homepage alone would miss the discrepancy.
RSS, Atom, and JSON Feed give supporting readers and other clients a list of public updates. MachineRead looks for those documents and examines the dates it can parse. A feed can make updates easier to find, but its presence does not establish that a subscriber or crawler received them.
RSS and Atom describe published structure
RSS 2.0 is an XML syndication format. Its specification describes a channel that may contain any number of item elements, so a parseable channel alone does not tell you whether it lists any articles (RSS Advisory Board, "RSS 2.0 Specification").
Atom gives update dates a specific meaning. RFC 4287 defines an Atom feed document as the top-level atom:feed element with zero or more entries. It also defines atom:updated as the most recent change the publisher considers significant, and it explicitly says not every modification necessarily changes that value (Internet Engineering Task Force, "RFC 4287").
JSON Feed dates are optional
JSON Feed puts the same general publishing idea into JSON. The version 1.1 specification describes JSON Feed as a syndication format like RSS and Atom, but JSON-based instead of XML. It also marks date_published as optional and says the value, when present, uses RFC 3339 format (JSON Feed, "JSON Feed Version 1.1").
The optional part matters. A valid JSON Feed can have items without date_published. A target can also publish an invalid, stale-looking, or partly undated feed. That is still information, but it should become a caveat, not a verdict about the entire site.
What MachineRead checks
When feed discovery is in scope, MachineRead looks for common feed paths and homepage links with recognized RSS, Atom, or JSON Feed media types. These are candidate locations, not mandatory URLs. A publisher can host a feed elsewhere and link to it from the page.
The search-discovery check samples feed candidates, looks for items, and tries to parse their dates. Missing, invalid, or old dates become caveats. An empty feed provides less evidence of recent publication than one with dated entries, but neither result establishes the freshness of every page on the site.
How to read a feed finding
Compare an entry with the page it describes. Does the link open the intended article? Does the title match? Does its publication or update date describe a real content event? Then check whether a newly published page appears in the feed at all.
For example, a publishing job might regenerate article pages without regenerating the feed. Repeated audits would keep finding a perfectly parseable document, but its newest entry would never advance. The fix belongs in the publishing process, not in a fabricated date applied to every item.
If dates are intentionally omitted, keep that choice visible. If the site rarely publishes, an old latest entry may be accurate. Feed health is about whether the document reliably describes the updates it is meant to expose.
Works Cited
Internet Engineering Task Force. "RFC 4287: The Atom Syndication Format." RFC Editor, Dec. 2005, https://www.rfc-editor.org/rfc/rfc4287.txt. Accessed 9 Sept. 2026.
JSON Feed. "JSON Feed Version 1.1." JSON Feed, https://www.jsonfeed.org/version/1.1/. Accessed 9 Sept. 2026.
RSS Advisory Board. "RSS 2.0 Specification." RSS Advisory Board, 30 Mar. 2009, https://www.rssboard.org/rss-specification. Accessed 9 Sept. 2026.
See also
- Methodology reference - separates observed documents from runtime guarantees
- Indexing directives and snippet controls - another case where publication metadata should not be read as outcome proof
- Semantic HTML that agents can follow - why small public structures matter when their limits are kept visible