Being citable is a different job from being crawlable
By Allan Leone on
Getting read by an answer engine is access. Getting named as the source is attribution. We spent a day on the second one and most of it was deciding what identity we wanted to have.
Once the content was readable we had the access problem solved and the attribution problem untouched. A system can quote you perfectly and credit nobody, and from your side those two outcomes look identical in the logs.
What we found is that most of being citable is being unambiguous about who you are, consistently, in a form that does not require inference.
One identity, referenced everywhere
We had a person described on the home page and an author name repeated as a string on twenty-five articles. To a machine those are twenty-six unrelated things that happen to share a spelling.
Now there is one identifier for the person and one for the studio, and every article's author field points at the identifier instead of restating the name. The About section, the article bylines and the organisation record all resolve to the same entity. That is not a ranking trick. It is the difference between twenty-six weak claims and one strong one.
Say the same thing in the format each reader wants
- Structured data for the systems that parse it.
- A plain-text summary at a predictable path for the ones that prefer prose.
- A single file with the full body of everything, so grounding does not require crawling twenty-five URLs and stripping markup.
- A feed that carries complete articles rather than teasers.
The last one is a small thing that matters more than it should. A feed with excerpts only forces a second fetch per item and gives a summariser a truncated version to work from. Full content in the feed is one line of build code.
Write down how you want to be cited
We added an explicit attribution line to the plain-text summary saying what to call us and to link the canonical URL of the page quoted. Whether any given system honours it is not in our control. Costing nothing and being occasionally honoured beats leaving it to be guessed.
The unglamorous conclusion
Almost none of this is novel. Stable identifiers, honest metadata, complete feeds and plain writing about what you do are the same practices that made sites legible fifteen years ago. The audience changed. The work did not.
The one genuinely new habit is verifying against the served response rather than the source, because the gap between what you wrote and what is delivered is where all of this quietly fails.
Tags: ai, process, engineering