The speed of online publishing often makes the delay required to manually art direct a story’s presentation unacceptable. And with so many devices mandating so many different presentations, the expense involved is huge. So manual art direction isn’t feasible, and the problem is getting worse.

The only option in my mind is to replace manual art direction with rule-based art direction. These rules, though imperfect, would be designed to do the same thing a human art director does: clarify or enhance a story’s meaning by providing context for it and making its presentation visually appropriate. For example, we might have rules like: “When an article is satire, use Typeface X.” or “When an article includes key people whom the reader might not know, offer background information on those people in a tooltip.”

Metadata comes in because it’s what enables the computer to apply these rules automatically; if the computer doesn't know what topics the content is about, how important it is, who it’s talking about, etc. then it can’t customize the presentation accordingly. So structured content + rules is the replacement for traditional art direction.

And this is beginning to happen (more). The Guardian’s iPad app is a recent example: each story is assigned an importance score that dictates how prominently it will be displayed.

Since publishing this article, I’ve become more conscious of the limitations of rule-based art direction compared to manual art direction. There are a lot of valuable things that human art directors do that rule-based art direction can’t because they only apply on a case-by-case basis. Or, if those things are driven by general principles, it would take too much work enable the computer to implement those principles automatically. Things like setting a quote in Courier to reinforce that it came from a government document, borrowing color (or type) directly from the content, using formatting conventions to separate different actors in a story, or aligning an image so that it’s “coming out of” one edge of the screen because it’s cropping warrants such a treatment.

So rule-based art direction can’t get us all the way to the kind of layouts we’d like and, where it leaves off, something like content tailoring should probably jump in. But, whenever we see a visual treatment that is generalizable, we should prefer to implement it with rule-based art direction rather than story-by-story CSS customizations, because using metadata will allow other customizations in the same spirit (e.g. as you refine your design ideas or as your content is forced to show up in new, unanticipated contexts) to be applied retroactively to all relevant stories.