I spent enough time building with Lovable to know why people like it.
You describe what you want, watch it build and a few prompts later you can have something that would have taken much longer to create the old way.
But I also learned something I wish I had paid more attention to earlier.
A website looking finished does not mean the technical SEO underneath it is finished.
I really started noticing this after I built the Marketur AI Readability Check and started running my own sites through it.
The problem wasn’t that the sites looked broken
That would have been easy.
If a button doesn’t work or a page is blank, you know something needs fixing.
The problems that bother me more are the ones a normal visitor never notices.
A site can have a Home link, About link, Pricing link and Contact link. You click them and everything appears to work.
But what URLs exist? What HTML does the server return for each route? Are the pages meaningfully different before JavaScript runs? Are internal links discoverable? Does each important page have the right title and metadata?
Those are different questions from whether the website looks good.
AI builders can interpret “make a page” differently than I do
This became one of my biggest lessons.
When I say, “Create a pricing page,” I am thinking about a dedicated URL with unique content, its own title, description, heading, canonical information and links from the rest of the site.
An AI builder’s job is to satisfy the prompt. If the result looks like a pricing page when I click Pricing, it may have technically done what I asked.
That does not guarantee it made every SEO decision I never mentioned.
Modern applications often use client-side routing. That is not inherently bad. Google has explicitly updated its JavaScript guidance over the years and can render JavaScript. Google also recommends clear crawlable links and careful canonical handling for JavaScript sites.
You can read Google’s JavaScript SEO basics for the technical details.
The issue for me is simpler: I should not assume a page is crawler-friendly just because it looks like a page to me.
AI search made this more important to me
Traditional SEO was already enough reason to care about crawlable architecture.
Now people are also discovering information through ChatGPT and other AI systems.
OpenAI distinguishes its search crawler from other bot functions and says OAI-SearchBot access matters for content that publishers want surfaced in ChatGPT search. OpenAI’s publisher documentation explains that distinction.
That made me want to know more than whether Google had indexed a URL.
I wanted to know what an outside crawler could actually fetch from my sites.
My own Lovable experience is what made this real
I am not claiming every site made with Lovable has bad SEO. That would be nonsense.
Lovable gives you a lot of control, and a developer who knows what to ask for can build around these issues.
My point is that I did not always know what I needed to ask for.
I was focused on getting the product working. I was prompting features, changing designs, fixing bugs and watching credits disappear. If the pages looked right, I moved on.
Then my own checker made me inspect the site from the crawler’s side instead of the builder’s side.
That is when the difference clicked for me.
What I would prompt for now
If I were building an SEO-dependent website with an AI builder today, I would explicitly ask for dedicated routes for every important page, unique page titles and meta descriptions, crawlable internal links, correct canonical URLs, sitemap inclusion and meaningful server-readable HTML for important content.
I would also test the finished result instead of assuming the prompt was followed exactly the way I meant it.
Google’s documentation is also clear that structured data helps it understand page content, so I would make sure the appropriate schema is present where it makes sense. Google explains structured data here.
This ties into why I moved back to WordPress
I wrote about what happens when AI builder credits run out and why I moved my work back to WordPress.
This is part of the same story.
I still want AI. I just don’t want to give up understanding and controlling the foundation of my website to get it.
WordPress gives me normal pages and URLs by default, and I can use AI to manage and build on top of that. For the way I work, I like that combination better.
Check your Lovable site instead of guessing
If you built a site with Lovable, don’t take my experience as proof that yours has a problem.
Check it.
I built the Marketur AI Readability Check for exactly this kind of question. It checks crawler access and server-readable content, samples internal pages and looks for signs that multiple routes are effectively returning the same underlying document.
It can also flag common client-rendering issues and give you information you can take back into your builder.
You may run your site and find everything is fine.
Or you may find what I found: a problem you would never have noticed by looking at the website.