A feed is a tidy, machine-readable list of what a site has published, with titles, dates, links and often the full text. For twenty years that was useful mainly to feed readers. In the last couple of years it has become useful to a new kind of software: AI assistants and the automated agents built on them.
What has changed
When an assistant needs to know what is new on a site, it has two options. It can load the pages and try to work out which parts are articles, or it can read the feed. The feed is smaller, cleaner and dated, so it is cheaper to process and less error-prone. Developers building “daily digest” or “monitor these sources” tools reach for RSS first, and several assistants can now be pointed at a feed directly.
Uses we have seen
- Summaries. A tool reads your subscriptions every morning and writes a short digest of what matters.
- Monitoring. Researchers, journalists and companies follow dozens of feeds and get alerted when a topic appears.
- Grounding. An assistant that answers questions about a site can use its feed to know what is recent and avoid citing outdated pages.
What it means for publishers
If you publish a site, a good feed is now one of the simplest ways to make sure software understands what you publish. A few practical points:
- Keep accurate dates, including update dates when you revise a post.
- Include the full text, or at least a real summary rather than the first line.
- Use clear titles; they are often all that gets shown.
We wrote about the formats in RSS, Atom and JSON Feed; any of them works.
What it means for readers
For people, the value of a reader has not changed: you choose the sources, you see everything, in order. AI digests can help when you follow too much, but they also reintroduce something feeds were designed to avoid, a layer that decides what is important for you. Use them as a filter on top of your list, not a replacement for it.
