Here’s what WordPress AI for developers looks like now that WordPress 7.0 shipped a native AI client. Shipping it was the correct call. Not a safe call, not a trendy call — the correct one. I built a plugin on top of this API over the past few weeks, and the decision holds up under real use, not just in a press release.
Table of Contents
Every AI Plugin Was Solving the Same Problem Badly
Before 7.0, if you wanted AI in WordPress, you picked a plugin. Each one shipped its own API key field, its own settings screen, its own half-baked prompt system. Ten plugins meant ten different ways to authenticate, ten different rate limits, ten different failure modes when the vendor changed their pricing tier overnight.
None of that complexity had anything to do with your content. It was an integration tax, paid over and over, plugin by plugin. Core absorbing the AI client kills that tax. One connection, one credential flow, one contract every plugin can build against.
Vendor Lock-In Stops Being the Default Outcome
Third-party AI plugins tie your content workflow to whatever provider they picked. Switch providers, and you rebuild your prompts, your settings, sometimes your entire workflow. That’s lock-in dressed up as convenience.
A native client changes the leverage. WordPress owns the contract, not the vendor. Plugins built against that contract don’t care which model sits behind it. My plugin sends a system prompt and gets structured output back — it doesn’t know or need to know which provider WordPress is calling. That’s the whole point of an API boundary.
Users Stop Guessing What “AI in WordPress” Even Means
Before core AI, “AI-powered” meant something different in every plugin. One tool auto-generated alt text. Another rewrote your titles. A third summarizes posts into meta descriptions. Same label, three unrelated experiences, none of them consistent with how the rest of WordPress worked.
A native API gives every plugin the same primitives to build from: a system prompt, structured output, and a real connection to WordPress data. My plugin takes a title and an intent and returns a draft that lands in the WordPress editor like any other draft — no export step, no copy-paste, no separate app to babysit. That consistency comes from the platform, not from any single plugin author’s good taste.
What WordPress AI for Developers Means for You
If you’re a developer deciding whether to build against this API, the practical answer is yes — the contract is stable enough to ship on now, not “wait for 7.1.” My own plugin’s draft-generation endpoint is a standard REST API route built on top of the same primitives; there’s nothing exotic about the integration once you’ve seen it. WordPress’s own core development blog is the place to track where the contract goes next, since this is still a young API and the surrounding conventions are still settling.
The Fast AI, Slow Core Tension Is Real
Here’s the strongest case against what core just did: AI moves fast, WordPress core moves slowly. Models ship monthly. Core ships on a release cycle measured in months, sometimes longer. Baking AI into core risks locks WordPress into architecture decisions that age out before the next major release.
That risk is legitimate. A plugin can switch its AI provider next week if a better model becomes available. Core can’t move that fast, and it shouldn’t try to.
The API Contract Outlasts Any Single Model
But that argument confuses the model with the interface. WordPress 7.0 didn’t hard-wire a specific model into core — it defined a contract: how you send a prompt, how you get structured output back, how a plugin hooks into that exchange. The model behind the contract can change. The contract is what plugin developers build against, and a stable contract is worth more than whichever model happens to be state-of-the-art this quarter.
That’s the bet worth making. Models will keep changing — GPT this year, something else next year, something else after that. The plugin ecosystem doesn’t need WordPress to pick the best model. It needs WordPress to pick a contract stable enough to build on, and 7.0 did exactly that.