The problem being looked at
Websites publish files that are meant for machines rather than people: RSS and Atom feeds, JSON Feed, XML sitemaps, and structured data embedded in pages. Search engines, feed readers and AI systems rely on these files to understand what a site contains and when it changes.
These files tend to break quietly. A feed can be valid XML and still be missing dates, publish relative links, or stop updating after a redesign. A sitemap can list pages that redirect. Structured data can describe a business with details that no longer match the page. Nothing visible changes for a human visitor, so problems can go unnoticed.
What it might do
The working idea is a checker that takes a website address and:
- finds the feeds, sitemaps and structured data the site publishes;
- validates them against the relevant formats;
- checks that they agree with each other and with the pages they describe;
- explains each problem in plain language, with the specific fix.
The emphasis is on explanation. Validators already exist for most of these formats. The open question is whether a single, readable diagnosis across all of them is useful enough to justify a new tool.
Open questions
- Who has this problem often enough to care: publishers, developers, small business owners, or none of them?
- Which checks matter most in practice, and which are noise?
- Should it be a one-off check, a scheduled monitor, or something that runs in a build pipeline?
- How much can be checked reliably without logging into anything or touching private data?
- Does it duplicate free tools well enough that it should not be built at all?
What would need to be true to build it
Feed Doctor moves forward only if conversations with people who maintain websites show a recurring, specific problem that existing validators don't solve. If that evidence doesn't appear, the research will be written up in Notes and the idea shelved.
Talk about it
If you maintain feeds, sitemaps or structured data and have run into these problems, or think the idea is wrong, an email is welcome: brian@anvisco.com.