← All field notes

Does this client really need headless WordPress?

A new frontend can be the right answer. It can also give an ordinary content site a second deployment, a second set of credentials and a second place for something to fail.

Before choosing headless WordPress, name the requirement that the existing theme cannot reasonably meet. “We can generate a nicer first screen” is not enough. The editor and the person maintaining the site have to live with the architecture afterward.

Keep the definition simple

In a headless setup, WordPress stores or manages content while a separate application renders the visitor-facing experience. That application reads content through an API. WordPress may remain the familiar editing interface, but the published page is no longer necessarily rendered by its theme.

Lovable’s WordPress connector documentation describes this sort of CMS-backed application. A JavaScript app built with another tool can also use APIs. Those connections make a workflow possible; they do not automatically recreate every theme or plugin behavior.

The benefit should be concrete. Perhaps the same content feeds several interfaces, or the site needs a custom interaction that deserves a separate application. Write that reason into the proposal so the added maintenance has a purpose.

Test the editor’s ordinary week

Ask an editor to create a draft, preview it, change an image and publish a correction. Check when the frontend reflects each change. Does preview show unpublished content safely? Does a content update need a rebuild or cache purge? Who notices if that step fails?

Custom fields need their own rendering contract. ACF data arriving in an API response is not the same as an editor’s layout appearing correctly. Agree which fields the frontend uses, how missing values behave and who maintains changes to that structure.

Multilingual publishing, redirects and social metadata also need deliberate handling. A good homepage does not prove that every article language has the right canonical or preview image.

Inventory the behavior supplied by plugins

List the things the current site does: forms, search, member access, checkout, bookings or protected downloads. Determine which behavior still executes in WordPress and which must be implemented in the new application.

Do not assume that an API endpoint reproduces a plugin’s complete customer flow. Test permissions and transactional behavior against the actual setup. A contact form that looks the same but never reaches the inbox has not survived the rebuild.

For stores, review cart state, taxes, payment redirects and order creation as a dedicated project. The public page’s appearance is only one part of that work.

Make the ownership map before signing

  • Who owns the WordPress hosting and its backups?
  • Who owns the frontend deployment and domain configuration?
  • Where do credentials live, and who can revoke them?
  • Who tests publishing and recovers a failed release?
  • Which ongoing services have separate usage charges?

Keep credentials server-side where required. Never expose an administrator’s application password in a public bundle just because the integration initially worked in a development session.

An exported repository is useful, but restoring code does not restore every database, file upload or external account. Practice recovery at the same scope as the system you are delivering.

Use a native theme when it meets the requirement

For an editorial or brochure site whose team already uses WordPress well, a native implementation can reduce the handoffs you must maintain. That is an architectural preference, not a claim that headless sites cannot perform well or be indexed.

IrmaWP can manage the connected WordPress side, including supported operations and AI-assisted work. It does not replace general deployment and monitoring of every external frontend. If you choose headless, assign those responsibilities separately.

Read the Lovable comparison and Bolt comparison with that boundary in mind. Then write a small acceptance test: an editor publishes one real content change, a visitor uses the affected feature, and a teammate can recover the release. If that handoff is awkward, fix it before adding more pages.


Have a correction or a workflow we should cover? Tell the IrmaWP team.