Would I trust an AI website builder with a real client website?
After one Hostinger WordPress trial a few months ago, my honest answer is: not as a complete delivery workflow. Not yet. That is different from saying every AI builder is useless. This was a few months ago, so the tools may have changed since then.
Returning to heavier WordPress work
For several years my WordPress work was mostly smaller updates on existing client sites. This year I returned to more substantial WordPress development and started looking for places AI could make me faster without removing control, or removing the chance to review before launch.
I never left WordPress entirely. I just spent a stretch doing less of the bigger builds.
The Hostinger WordPress trial
Roughly three or four months before writing this, I was setting up a WordPress test site with a dummy domain through a client’s Hostinger account.
That is where I ran into Hostinger’s AI tools for WordPress: the AI Website Builder for WordPress (opens in a new tab) flow, plus a chatbot I remember as Kodee. Hostinger’s current docs describe Kodee as the earlier name for Hostinger Agent (opens in a new tab). I am going by what I remember from the trial; the rename is just a present-day clarification.
Importantly, this was inside a WordPress setup on Hostinger. It was not me reviewing Hostinger’s separate standalone Website Builder as if they were the same product.
My first, slightly embarrassing assumption was that the AI builder was somehow a native WordPress feature. Then I realised it came from the hosting provider. Of course it did. I still had that half-second of “wait, when did WordPress start doing that?” before the more useful thought arrived: this is a Hostinger workflow sitting on top of WordPress.
Compressing a real brief into a small box
I hit a character limit when describing the website. I had written a substantial brief, and compressing it into the allowed space felt wrong compared with the context documents I normally use when I work with an agent in Cursor. The memory that stuck was the frustration of shrinking requirements I already considered important.
The first generated website did not reflect that brief as closely as I wanted. The shortened prompt, or asking for too much in one go, may have contributed. Either way, the output was a website, and it was not the website I intended.
Follow-ups that felt like more work, not less
I did try requesting changes.
I remember asking for structural updates across pages, asking for a homepage redesign, and finding that at least one requested cross-page change did not complete as expected. I do not remember the chatbot’s exact wording or the precise technical limitation.
What I do remember is the accumulating burden: look at the result, spot the gaps, explain them to the chatbot, check again. Review is normal. I already review AI-generated work elsewhere. The question here was whether this tool reduced the total effort compared with the way I already deliver WordPress sites.
In that trial, it did not feel like it did.
The gap after the first generation
I also felt there was limited guidance on what still needed doing after the initial site appeared.
Permalink settings are an ordinary example of the kind of WordPress decision I still think about. I wanted clearer help with the remaining work between “a site exists” and “this is ready.”
Generating a first version quickly is not the same as finishing.
Why the local agent path still suited me
What I preferred then, and still prefer now, is working in a local WordPress environment with Cursor: richer context and instructions, adjusting a custom theme where that is appropriate, working across the whole site, and reviewing before anything moves to hosting.
Local and WordPress Studio are different tools, and Cursor does not magically handle every publish or migration step without setup. What I care about is the shape of the workflow: keep the brief intact, keep the site’s conventions visible, and keep a human review before launch.
What I want from WordPress AI, longer term, is an ongoing agent that can keep making meaningful changes while respecting the site’s context. That is the direction I am looking for, not a verdict that nothing currently exists.
Still interested, still not convinced by that trial
The trial was months ago. Hostinger’s tools, and everybody else’s, may have improved since then. I am willing to revisit Hostinger and look at other providers. AI builders and hosting assistants feel increasingly loud in website marketing, and I have not audited every host. I also suspect there may be work helping owners repair or refine AI-generated sites.
I am genuinely interested in using these tools myself if they speed up setup, make collaboration easier, let a client build an initial version and feel involved, and still leave room for me to provide consulting, refinement and technical support. A first draft and a finished delivery are different jobs.
So would I trust an AI website builder with a real client website?
On the strength of that Hostinger WordPress trial: not as the whole process. I want tools that improve delivery end to end. A fast first site is only useful if it leaves less work afterwards, not more.
