Can ChatGPT read your site? Test it with curl.

If your server sends near-empty HTML and JavaScript draws the rest, AI bots see a blank page. One command shows what a bot receives, and what to change when the text isn't there.

By @ignapradosSEO and AI4 min read

Your site looks perfect in the browser, yet ChatGPT, Claude or Perplexity can't tell what it's about. The cause is often simple: the browser runs JavaScript and builds the page, but many bots only read what the server sends in its first response.

It's the typical case of single-page apps built with React and Vite, the base of many Lovable and Bolt projects. Here is how to tell if it happens to you, what the official sources say and what to ask your tool to fix it.

Google waits, AI bots don't

Google runs JavaScript, but through a queue. Its JavaScript SEO documentation says Googlebot queues every page with a 200 status for rendering, that a page may wait a few seconds or longer, and that blocked files and pages aren't rendered. It ends with a recommendation: server-side or pre-rendering is still a great idea, since not all bots can run JavaScript.

AI crawlers are the clearest example. In its December 2024 analysis, Vercel measured traffic from OpenAI, Anthropic and Perplexity crawlers and found that none of the major ones render JavaScript. ChatGPT and Claude do download .js files (11.5% and 23.84% of their requests) but don't execute them. The study is almost two years old, so treat it as a hint, not a guarantee. That's why the next step is checking your own site.

What a bot sees, in one command

Ask your server for the page the way a bot would, with no browser, and look at how much text comes back:

curl -s https://your-domain.com | grep -o "<h1[^>]*>.*</h1>"
curl -s https://your-domain.com | wc -c

If the first command doesn't print your main heading and the second shows only a few hundred characters, JavaScript supplies almost all the content. On a server-rendered site the <h1> shows up and the HTML weighs tens of thousands of characters. Try the home page and an inner page, because sometimes only the first one is sorted out.

What changes if your site started on Lovable

Lovable's documentation separates two cases. Projects created since May 13, 2026 use TanStack Start with server-side rendering: every request returns full HTML. Older React and Vite projects use on-request pre-rendering: when a verified crawler arrives, Lovable builds the page on the fly and returns the resulting HTML.

The word that matters is *verified*. Per the same docs, a bot that isn't on the list sees the regular single-page app unless it asks for the page as Markdown. So curl on an older project shows what an unverified client sees, and an empty result doesn't prove ChatGPT sees the same. The page itself suggests checking with Google's URL Inspection tool. And if your site isn't hosted on Lovable, none of this applies: pre-rendering is a hosting matter.

What to ask your tool for

The real fix is for the title, description and main text to be in the first HTML, through server rendering or pages generated ahead of time. Give this to your coding agent, changing structure only, not design:

My site is built with JavaScript in the browser and the HTML my server sends arrives almost empty. Make every page arrive rendered from the server (SSR or static generation; prerender in Vite, Server Components in Next.js) with the title, description and main text in the initial HTML. Don't change the design. When you're done, tell me how many characters of text the home page HTML has when fetched with curl.

Two warnings from Lovable's migration docs: some libraries only work in the browser and can break server rendering in ways the upgrade's checks don't catch, and your published site keeps serving the old version until you publish again. After migrating, run the curl again.

And mind your links: Google only discovers links that are <a> elements with an href, so a button with an onClick that navigates doesn't count as a link.

Checking every page

A fixed home page doesn't guarantee the rest. Slop Score Pro checks the whole site and, if the HTML arrives nearly empty, flags it as «Without JavaScript, your page is blank», with the prompt to fix it. To see why builder sites look so alike, the builders guide covers what each one decides, and the post on link previews covers another symptom of the same cause.

Sources

Frequently asked questions

Can ChatGPT read my site built with Lovable?

It depends on when you created it. Since May 13, 2026 new projects use server-side rendering. React and Vite ones get pre-rendering only if the crawler is verified.

Why isn't my site on Google if it looks fine?

It may be waiting in the rendering queue, be blocked, or generate content and links in a way Google can't discover. Start by checking the initial HTML with curl.