Building a website with AI: what you must fix yourself

You type one sentence and fifteen minutes later there is a website. Your own colours, tidy headings, a working contact form. That is no longer a promise from an advert, it actually works.

The question that comes next is a different one. Can Google read those pages? And can ChatGPT and Perplexity do anything with them? That does not depend on how the site looks, but on what the tool delivers under the hood.

I went through the official documentation of nine well known builders: Lovable, v0, Bolt, Framer, Wix, Squarespace, Hostinger AI Builder, Durable and Webflow. Not what review sites write about them, but what the makers put down themselves. The differences are big, and for two of the nine there is nothing at all on this subject.

This article covers what they do and do not take care of, which four things you have to check yourself, and where Google draws a line with content made using AI.

What the builders say themselves about findability

This is the state of the documentation on 22 September 2026. Empty means: not documented, not that it is missing.

Tool	Rendering	Setting titles and meta yourself	Sitemap and robots.txt	Schema markup
Lovable	Server-side for apps from 13 May 2026; older apps only pre-render for recognised crawlers	Yes	Not always created up front	Gets checked, not added automatically
Framer	Server-side rendered React	Yes, per page	Automatic	Manually via custom code
Wix	According to Wix server-side, but that only appears on a marketing page	Yes, generated automatically and editable	Automatic, robots.txt can be edited	JSON-LD, at most five per page
Squarespace	Not documented	Yes, metadata is generated automatically	Automatic	Automatic for certain page types
Durable	Not documented	Yes, per page	Automatic	Automatic, JSON-LD on every page
Hostinger AI Builder	Not documented	Through a prompt to the builder	Only with your own domain and after publishing	Not documented
Webflow	Not documented	Yes	Yes	Via page settings, paid plan
v0	Next.js, rendering model not documented	Not documented	Not documented	Not documented
Bolt	JavaScript, not documented further	Not documented	Not documented	Not documented

Two things stand out.

The first: the builders aimed at business owners take care of the basics largely by themselves. The builders aimed at developers, v0 and Bolt, do not document this subject. I went through the full documentation index of both and found no page at all about meta tags, sitemaps or rendering. That does not mean it is impossible, it means it is your job.

The second: the tools move fast. Lovable writes itself that apps created from 13 May 2026 render server-side: “New apps created from May 13, 2026 use TanStack Start with server-side rendering (SSR). Every request returns fully rendered HTML, for humans and crawlers.” Something else applies to older projects: they only render the page at the moment a recognised crawler comes by. Lovable itself writes what that means: “Third-party SEO scanner tools are not verified crawlers, so they may report your pages as empty even though Google, social platforms, and AI search engines see the full content.” So when your site was made determines what you get.

Why rendering is the point where it goes wrong

Rendering is the moment code turns into a page with text on it. That can happen on the server, after which the visitor gets ready made HTML. Or in the browser, after which JavaScript builds up the content.

Where the page gets built decides who can read it.

Google can handle the second. The documentation on JavaScript SEO says that Google works in three stages, crawling, rendering and indexing, and that it renders with a current version of Chromium. Google also names the catches: a page can stay in the queue, “The page may stay on this queue for a few seconds, but it can take longer than that.”

And then there is this, which I consider the most important advice in that whole document: “Keep in mind that server-side or pre-rendering is still a great idea because it makes your website faster for users and crawlers, and not all bots can run JavaScript.”

That last part is about the crawlers of AI search engines, and less is known about that than you would think. The crawler documentation of OpenAI, Perplexity and Anthropic describes which bots they use and how they deal with robots.txt, but none of the three says anything about JavaScript. Not that they run it, not that they do not run it. Which signals do weigh once such a system is inside, I describe in my article on how ChatGPT picks its sources.

The claim that “AI crawlers do not run JavaScript” is presented everywhere as fact. The only measurement with a visible method that I found is a study by Vercel and MERJ from December 2024: “The results consistently show that none of the major AI crawlers currently render JavaScript.” That same study says Gemini and AppleBot do render it, because they run on different infrastructure. It is almost two years old, and the party doing the measuring sells hosting for server-side rendering. So take it as an indication, not as law.

The practical conclusion is simple. Google documents that it renders JavaScript, the others do not document it. If you have to guess, HTML that is already finished is the safe choice. What else belongs to that technical basis is in my article on technical SEO.

Four things to check yourself after building

This takes half an hour and it saves you months of waiting.

Check that your pages really are in the HTML. Open your page, right click and choose view page source. Do you see your own headings and paragraphs in that code? Then nothing is wrong. Do you only see a few lines of script and an empty div? Then your content is built up in the browser first. That is the signal to look further.
Check titles and meta descriptions per page. Automatically generated titles are often the page name plus the site name. That is not a mistake, but it is not a choice either. Write them yourself, and watch out for duplicate titles; that is the most common mistake on sites that were set up quickly.
Look up your sitemap and your robots.txt. Type your domain plus /sitemap.xml and plus /robots.txt in the browser. Do you get a 404? Then there is nothing. With some builders those files only appear once you have connected your own domain and really published the site, not in the preview version.
Check that your preview URL is not joining in. Nearly all builders put your site on an interim address first. Lovable is clear about that: only apps that are really published can be indexed. After connecting your own domain, check whether the old address is still reachable too, because then you have the same site online twice.

What you achieve with that is not a position in Google. It is a site whose pages are in the index and readable for crawlers. That is the minimum, not the end point.

Where Google draws the line with AI content

This is the question I get most often: is this even allowed by Google? The answer has been there for years, and it is clearer than the discussion about it.

Google does not ban AI. The post Google published on this subject in 2023 and still uses as a reference says: “Appropriate use of AI or automation is not against our guidelines.” And a little further on: “Using AI doesn’t give content any special gains. It’s just content.”

The line sits in the volume and the intent. The spam policies call that scaled content abuse, and the first example in that list is literally: “Using generative AI tools or other similar tools to generate many pages without adding value for users.”

In practice: a ten page site you built with AI and whose texts you checked yourself is not a problem. Three hundred city pages where only the place name differs is. Google even says what you should do with such pages: “If you’re hosting such content on your site, exclude it from Search.”

There is one more piece of advice that almost nobody follows, and it is in Google’s documentation on generative AI content: “If you’re automatically generating content, consider adding information on how your content was created in a way that makes sense for your audience.” So just say that a text was set up with AI, if that is the case.

When an AI builder is enough

I sell websites, so you may read this section with a pinch of salt. But I do not tell everyone that it could be better.

An AI builder is enough if you need a simple site, you write the texts yourself, you have checked the basics above and you do not want to expand the site every quarter. For a one person business that mainly wants to be findable for one location and one service, that gets you a long way.

It gets harder with three things. The first is a lot of pages that look alike, because then it is about the structure of your site and about how those pages link to each other, no longer about building. The second is moving: not all builders let you take the code with you, and Framer simply says no to that. The third is speed, because there the platform decides more than you do. If you run into those, that is the moment to look at web development instead of at the next tool.

The real cost question, by the way, is not in the building. It is in what gets added every month after that. What that costs in the Netherlands, I looked into separately in my article on what a website costs.

In short

Of the nine builders I checked, most take care of the basics themselves; the builders for developers, v0 and Bolt, do not document findability at all.
Lovable renders server-side for apps from 13 May 2026; a different approach applies to older projects. The date your site was made determines what you get.
Google renders JavaScript, but still advises server-side or pre-rendering itself, because not all bots can run JavaScript.
OpenAI, Perplexity and Anthropic say nothing about JavaScript in their crawler documentation. The well known claim that they do not run it comes from a single measurement in December 2024.
After building, check four things: is your text in the HTML, are your titles your own, do the sitemap and robots.txt exist, and is your preview address offline.
Google does not ban AI content. The spam policies target many pages without added value.

Frequently asked questions

Is a website built with AI bad for SEO?

Not by definition. Google says itself that AI gives no bonus and no penalty: “It’s just content.” What counts is whether the pages are readable for crawlers and whether they add something. The technology of the builder you pick makes more difference than the fact that AI is involved.

Can Google read a site that leans entirely on JavaScript?

In most cases, yes. Google documents that it renders JavaScript with a current Chromium. There are conditions: scripts must not be blocked, and with an error status the rendering can be skipped. Google still recommends server-side rendering.

Can ChatGPT and Perplexity read my site if it leans on JavaScript?

Nobody knows for sure, and that is the honest answer. OpenAI, Perplexity and Anthropic do not document whether their crawlers run JavaScript. As long as that is the case, HTML that is already rendered is the safe choice.

Should I block AI crawlers?

That depends on what you want. If you block the search crawlers, you can no longer appear in those answers; OpenAI and Perplexity both write that you have to allow their search bot to show up in their results. Watch out for the lists behind such a button too. Squarespace has one checkbox to block known AI crawlers and publishes the list alongside it; PerplexityBot and OAI-SearchBot, the two search bots, are not in it. Squarespace itself writes alongside it: “Requesting that known AI crawlers exclude your site doesn’t guarantee they will, but it’s the best option currently available.”

Do I have to say that my texts were made with AI?

Google advises it, in the words “consider adding information on how your content was created in a way that makes sense for your audience”. It is not a requirement and it is not a ranking signal. It is more honest towards your reader, though.

Sources

All sources consulted on 22 September 2026.

Not sure whether your site is set up right?

In a free GEO scan I look at your technology, your findability in AI answers and the basics underneath. You get the findings with the source alongside, so you can check them yourself.