Custom Post Types and Taxonomies: Modelling Real Content

When something deserves its own post type, when a taxonomy is the wrong tool, and the archive URLs these decisions quietly create. A WordPress content modelling guide with the crawl consequences included.

Custom post types are the point where a WordPress site stops being a blog and starts being a system. They are also where a lot of sites acquire problems that take years to surface, because the decision is made in an afternoon and lived with for a decade.

The two failures are opposite and equally common. Forcing everything into posts and categories until the taxonomy is meaningless. Or creating a post type for every section of the site until nothing can be queried together.

The decision, stated simply

Something deserves its own post type when three things are true.

It has different fields. A case study has a client, an industry and a result. An event has a date, a venue and a capacity. If the data you need to store differs structurally, the content type differs.

It needs a different template. If it displays differently enough that you would write a separate template file for it, that is a signal.

It has a different lifecycle. Events expire. Job listings close. Case studies persist. Products come and go with stock. Different lifecycles mean different handling.

If all three are true, create the post type. If only one is, you probably want a category and some custom fields on an ordinary post.

The question people usually ask instead is whether the content feels different. That is not a reliable test, and it is how sites end up with nine post types that are all just articles.

Taxonomy or custom field

This one confuses people constantly, and the test is straightforward.

Use a taxonomy when you will need to list everything sharing that value. Industry, topic, product category, location served. These are things users browse by, so they need an archive and a shared identity.

Use a custom field when you only need to display the value on the item itself. Price, duration, client name, capacity, a date. Nobody browses all the events costing a particular amount.

Getting this wrong in the direction of taxonomy is the expensive error, because every taxonomy term creates a URL. A taxonomy of two hundred client names produces two hundred archive pages, almost all of them with one item on them, each one competing with the item itself.

Hierarchical or flat

Two kinds of taxonomy. Hierarchical ones behave like categories, with parents and children, selected from a list. Flat ones behave like tags, typed freely.

Use hierarchical for a controlled structure somebody designed: service lines, product categories, regions. Use flat for an open set where new values appear naturally.

The warning is about flat taxonomies and editors. A freely typed taxonomy will acquire near-duplicates within months. Pune, pune, Pune City, PUNE. Each one is a separate term with a separate archive page, and nobody notices until somebody audits the site and finds four hundred tags used once each.

If you give editors a free-text taxonomy, give them a governance rule with it, or make it hierarchical so they must choose from a list.

The part nobody thinks about: you are creating URLs

This is where content modelling becomes a search problem rather than a development one.

Register a public post type with an archive and you have created an archive URL, a paginated series of them, and one URL per item. Register a taxonomy and you have created one URL per term, each paginated. Attach that taxonomy to two post types and the combinations multiply.

On a large established site this is fine and often useful. On a new site with limited crawl allowance it is a real cost, because crawl effort spent on thin term archives is effort not spent on the pages you care about. We went through how that allocation works in crawl budget explained.

So decide deliberately, per post type and per taxonomy, whether its archives should exist publicly at all.

  • Does this archive serve a visitor? A services archive listing all services, yes. An archive of every author, on a site with two authors, no.
  • Would it rank for anything? A term archive with three items and no original text will not.
  • Does it duplicate something? A taxonomy archive that lists the same items as a hand-built landing page is competing with that page.

Where the answer is no, turn the public archive off at registration rather than noindexing it afterwards. A URL that does not exist is cheaper than one that exists and is excluded, which is the distinction we drew in robots.txt, noindex or canonical.

Custom post types get a URL prefix, and it is worth thinking about for thirty seconds rather than accepting the default.

Use the plural form, keep it short, and use the word your audience uses rather than your internal name. If your post type is internally called resource because that was convenient, your URLs do not have to say resource.

Decide whether taxonomy terms belong in the item URL. Including them reads nicely and creates a problem: when an item changes category, its URL changes, and you now owe a redirect. For content that moves between categories, keep the URL flat.

Whatever you choose, changing it later means redirecting every URL of that type. Treat it as close to permanent.

Register it in a plugin, not the theme

This is the rule worth repeating because the consequence is severe and delayed.

A post type registered in a theme disappears when the theme changes. The content stays in the database, orphaned, invisible, with every URL returning a 404 and the client convinced their case studies were deleted.

Put registrations in a small companion plugin. The theme decides how things look. The plugin decides what things are. This is the same boundary we set out in our piece on custom theme structure, and content modelling is exactly where it bites.

Changing your mind later is expensive

Converting content from one post type to another is possible and it is not a clean operation. The post type changes, the permalink changes, the old URLs break, and anything that referenced them by type stops working.

Which argues for two habits at the start. Spend an hour on the model before writing any code, with the actual content in front of you rather than imagined content. And when in doubt between a new post type and a category on an existing one, choose the category, because promoting a category to a post type later is less painful than the reverse.

The conceptual version of this exercise, applicable whatever system you use, is in content modelling for SEO and AI.

Sitemaps and what you are telling search engines

By default, public post types and taxonomies end up in your sitemap. Check what is actually in yours after you register anything new.

A sitemap should list the pages you want indexed. If it contains four hundred tag archives with one item each, you are explicitly asking for them to be crawled, which is the opposite of what you want on a site that is struggling to get its real pages seen. The structural side is in sitemaps that scale.

What is a waste of effort

A post type for a single page. Pages exist for that.

A taxonomy nobody filters by. If no template queries it, it is a field.

Mirroring your internal vocabulary in public URLs. Your users do not share your naming.

Building the model around a plugin’s demo content. It was built for somebody else’s site.

Leaving every archive public because it was the default. Defaults are not decisions, and this one creates URLs you will later pay to suppress.

A sensible sequence

  1. List your actual content with real examples in front of you.
  2. Group by fields, template and lifecycle. Those groups are your post types.
  3. For each shared attribute, ask whether anyone needs to browse by it. If yes, taxonomy. If no, field.
  4. Decide hierarchical or flat per taxonomy, and write the governance rule for anything flat.
  5. Decide, per post type and taxonomy, whether its archive should be public at all.
  6. Choose permalink structures in your audience’s vocabulary and treat them as permanent.
  7. Register everything in a companion plugin.
  8. After registering, check your sitemap and your crawl for URLs you did not intend to create.

If your site has accumulated term archives nobody designed, the related clean-up question is covered in faceted navigation and pagination.

We work through content modelling hands on in the training in Pune, and if you would rather have the model designed for you, get in touch.