The Content Model Decides the Page

The Model Is The Constraint

An editor can only enter what the model has a box for. If there is nowhere to record who wrote a piece, nobody records it, and no amount of editorial standards fixes a missing field.

So the content model is where the extraction rules from Optimizing for AI Discovery either become the default or stay a thing people are asked to remember. Defaults win.

The Four Fields

A summary, required, written as an answer. Not an excerpt derived from the first paragraph. A separate field with help text saying what it is for, rendered directly under the heading. On a collection of two hundred, this is two hundred first sentences and it is the highest-value field on the model.

An author reference, not a string. A relation to a person record with a biography and a link that resolves somewhere. A name in a text field is a name; a reference is an identity a system can follow.

A date for the claim, separate from the publish date. Most models have created and updated timestamps and neither says when the thing being claimed was true. One field, printed where the claim is, answers the freshness question instead of leaving it to be guessed.

A source on anything numeric. Where a content type carries figures, a repeatable field of source and date beside them, rendered inline rather than as a footnote nobody links.

Make The Template Enforce The Order

Adding fields does nothing if the template renders them in the wrong place.

Summary directly under the heading, body beneath it, provenance at the end of the piece rather than in a sidebar that gets lifted out of context. Decorative blocks after the substance. One template decision, applied to every item of the type, which is the same leverage One Template, Many Pages describes with a visual builder.

Backfilling Without A Project

Adding a required field to a collection with three hundred existing entries is where this stalls.

Do not backfill all of them. Sort by whatever traffic or revenue signal you have, take the top twenty, write those by hand, and let the rest fill in as they are next edited. The twenty that carry your commercial questions are the twenty a category question would retrieve, and the tail was never going to be cited.

Never generate the summaries automatically from the body. A generated first sentence is the first sentence of the body with the words rearranged, which is the problem restated rather than solved.

One More Field Worth Considering

A boolean for whether the entry should be in the sitemap.

Most models make everything public by default and rely on somebody remembering to unpublish. An explicit field turns "should this be findable" into a decision made once, at the moment somebody knows the answer, rather than an audit somebody does later with less context.

Key Takeaways

• The content model decides what every page of a type can say, so editorial standards cannot fix a missing field.
• Four fields carry most of it: a summary written as an answer, an author reference, a date for the claim, and a source per figure.
• The template has to render them in the right order, or the fields exist and the answer is still fourth.
• Make the summary required. An optional field on a busy team is an empty field.
• Backfill the twenty entries that carry your commercial questions and let the rest fill in as they are edited. Never generate summaries from the body.