Integration · Lovable & AI builders

Lovable SEO & blog: Get found by Google and ChatGPT.

Many sites built with Lovable, Bolt or Replit are single-page apps: the text only appears once the browser runs JavaScript. Google renders that later, AI crawlers like GPTBot or ClaudeBot don’t run JavaScript at all. Since May 2026 Lovable renders new projects on the server, older ones only for crawlers it verifies. Claire runs your blog on blog.your-domain.com, rendered on the server, with no code in your builder.

Book a demo
No credit cardSet up in 15 minutesCancel monthly
Two founders reviewing printed drafts at a work table, a laptop beside them
Publishing on blog.your-domain.com …
Server-rendered page, full text in the HTMLdone
Sitemap, RSS feed and Article schemadone
What does an online course cost per student?Approve

Companies whose blogs grow with Claire

TrustypostPlanimolREA CardTomorrow UniversityKUGUPaforySybilionNubibase

The problem, measured

What AI crawlers see on your site. We counted.

Whether ChatGPT, Claude or Perplexity can quote your site depends on the HTML your server sends, not on what you see in your browser.

01

Why AI-built sites often arrive empty

Lovable (until May 2026), Bolt, Replit and Base44 build websites as React apps that run in the browser. The server sends an almost empty HTML file: a title, a description, an empty container and a link to the JavaScript. Headings, copy and links are assembled by the browser.

People never notice. A crawler that doesn’t run JavaScript only sees the title. To it, your page has no content it could understand or quote. That is the whole point of single-page application SEO: a single-page app needs server-side rendering or pre-rendering to be readable without JavaScript.

02

Our test: 41 websites from six builders

On October 5, 2026 we fetched the home pages of 41 public websites that were demonstrably built with AI builders: twice raw, without JavaScript like a crawler, once identifying as GPTBot and once as a regular browser, and once in a real browser. We counted the words in the page content.

The pattern is clear. Where the page is built on the server or at build time, the text arrives: on five of six v0 sites, four of five new Lovable projects and six of seven sites built with Claude Code. Where it is only built in the browser, almost nothing arrives: none of the older Lovable sites and none of the Bolt sites delivered content without JavaScript, and four of six Replit sites fewer than 200 words. Base44 sends bots a separate, rendered version.

A second, larger sample from a public showcase of Lovable projects confirms it: of 65 older projects on Lovable hosting, 52 delivered fewer than 100 words without JavaScript; of 19 new projects, 17 delivered at least 100.

Built withRaw as GPTBotRaw as browserIn a browser
Lovable, old stack (5 sites)00408
Lovable, new stack (5)786786910
Bolt (6)00585
v0 (6)616616788
Base44 (6)872379859
Replit (6)107321,203
Claude Code (7)379483504
Comparison: byclaire.co subpage858858885
Median words on the home pages, measured on October 5, 2026. Raw means the HTML the server sends, without JavaScript. Old stack: React with Vite; new stack: TanStack Start with server-side rendering since May 2026. We deliberately don’t name domains. Lovable sends older projects a pre-rendered version only to crawlers it verifies itself; our request identifying as GPTBot wasn’t one of them. On the three sites on Bolt hosting, Cloudflare blocked the request identifying as GPTBot; they count as zero words.
03

Google renders later, AI crawlers don’t

Google runs JavaScript and does index single-page apps, just with a delay: pages first wait in a queue to be rendered. Even so, Google itself recommends server-side rendering or pre-rendering because, in the words of its documentation, “not all bots can run JavaScript.”

The big AI crawlers are among them. A Vercel and MERJ study from December 2024 found that GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot and PerplexityBot download JavaScript files but don’t execute them. Gemini uses Google’s infrastructure and does render. OpenAI, Anthropic and Perplexity say nothing about it in their own crawler documentation.

04

What Lovable changed in 2026

Lovable has responded. Projects created since May 13, 2026 run on TanStack Start and are rendered on the server. Older projects stay React apps built with Vite. For those, Lovable pre-renders pages at request time, but only for crawlers it recognizes as genuine, such as those of Google, Bing, ChatGPT, Perplexity, Claude and Gemini. “Third-party SEO scanners and other unverified agents see the regular single-page app,” Lovable writes.

This is described for sites on Lovable hosting. For exported projects you host yourself, Lovable mentions no pre-rendering. Since July 2026 older projects can be upgraded to the new stack; according to Lovable that usually costs 10 to 35 credits.

05

Our own website had the same problem

Until October 5, 2026, byclaire.co was itself a single-page app built with React and Vite. Fetched as GPTBot, it returned a title, meta tags and tracking scripts, and an empty content area: zero words. The website of an AI content agent was invisible to AI crawlers.

Since then we pre-render every public page at build time. Our German homepage now delivers about 1,460 words without JavaScript, almost as many as the browser shows. And the build fails as soon as any page drops below 60 words, so this can’t happen again unnoticed.

06

How to check your own site

Open your site in the browser and view the page source: Ctrl+U in Chrome, Cmd+Option+U on a Mac. Search for a sentence from your page. If it isn’t in the source, a crawler without JavaScript doesn’t see it either. On older Lovable projects the source shows the empty shell, even though Lovable sends verified crawlers a pre-rendered version.

It’s quicker with our free AI crawler checker. It tests whether your robots.txt lets AI crawlers in and whether any text arrives without JavaScript, and tells you what to do about it.

Run the AI crawler checker

Lovable SEO: what you can do yourself

Six steps inside your builder. No agency, often one prompt.

You can do most of this in Lovable itself, with a prompt or the built-in SEO review. In Bolt, Replit or Base44 the menus have different names; the steps are the same.

01

Check your stack

If your Lovable project was created before May 13, 2026, it runs as a single-page app. Upgrading to TanStack Start gives it server-side rendering, so the text is in the HTML for everyone, not just for crawlers Lovable recognizes. Test the site thoroughly after the upgrade, because a stack change is a bigger rebuild. And look at the page source: one of the new sites in our test delivered only a loading screen despite server-side rendering, because it loads its content via JavaScript.

02

A title and description for every page

In a single-page app, subpages often share the title from the start file. Give every page its own title with its main search term and its own meta description. Lovable’s SEO review finds missing tags and fixes most of them in one click.

03

Add a sitemap and robots.txt

Lovable doesn’t create either automatically; its docs say they are “not always generated up front.” Have it create a sitemap.xml listing every public page and a robots.txt that lets search engines and AI crawlers in and points to the sitemap. Base44 generates both automatically on a custom domain.

04

One URL per topic

A long one-pager with jump links like #pricing is a single page to search engines. Services, pricing and the questions people search for need their own subpages with their own URL. Links between them should be real links, not buttons that redirect via script.

05

Custom domain and Search Console

Custom domains come with Lovable’s paid plans, Bolt Pro and Base44 Starter and up. Base44 states explicitly that its SEO setup is designed for custom domains, not its free addresses. Then verify the domain in Google Search Console and submit the sitemap.

06

Copy that machines can quote

One main heading per page, clear subheadings, real text instead of text in images, alt text for images. Spell out what you offer, for whom and at what price. AI search quotes sentences that answer a question directly, not slogans.

Lovable blog

A blog for your AI-built site. Three ways, and what they cost you.

None of the big AI builders ships a blog. For articles that Google and AI search can find, you have three options.

01

Option 1: articles inside the builder

You have every article built as its own page by prompt. For three or four articles that works. After that it gets tedious: every new article and every fix is a code change you have to review and republish, and every change can shift something somewhere else.

An overview page, categories, a feed and a sitemap that grows with the blog all need building too.

02

Option 2: connect a CMS

Lovable offers connections to Contentful, Storyblok, Notion and WordPress, Base44 to WordPress.com and Contentful. You keep the articles there and your site pulls them in. That’s clean, but you run two systems and still write everything yourself.

If your site runs as a single-page app, the articles also only load in the browser. Without server-side rendering or pre-rendering they stay just as empty for AI crawlers as the rest.

03

Option 3: a blog on a subdomain

The blog lives at blog.your-domain.com on a separate system that renders every page on the server. Your builder site stays untouched. You add one DNS record and a “Blog” link to your navigation.

A subfolder like your-domain.com/blog would need a reverse proxy in front of your site instead. Lovable allows that, but setting it up and maintaining it is on you.

04

What an article needs to get quoted by AI search

Wherever your blog lives, what gets quoted is what clearly answers a question.

  • A real question people search for as the topic.
  • The answer in the first paragraph, details after.
  • Numbers with a source and a date.
  • Subheadings, lists and tables where they help.
  • Upkeep: prices, features and numbers go out of date.

Claire and Lovable

How Claire adds a blog to your site. No code in your builder.

Claire finds topics your customers search for, writes the articles in your voice and serves them herself: on a subdomain of your domain, every page rendered on the server. Your Lovable site stays exactly as it is.

Connect in three steps

1

Enter your blog address

In Claire, open the “Site builder without a blog?” tile under Integrations, enter an address like blog.your-domain.com and click “Create blog”. Claire reads your website and suggests your brand color, logo and links to your legal pages.

2

Add one DNS record

Wherever your domain is managed, add a CNAME record: name “blog”, target blogs.byclaire.co. If you bought the domain through Lovable, you can do that in Lovable under Workspace domains. “Check connection” shows whether everything is in place. It can take up to an hour; the HTTPS certificate is issued automatically.

3

Style it and approve

Set color, font, logo, name and tagline in a live preview that works even before the DNS record is in place. Read every finished article in Claire and publish it with “Publish to blog”.

What Claire’s blog delivers

Server-side rendering
Every page is built on the server. Headings, text, images and links are in the HTML, including for crawlers that don’t run JavaScript. A test article on our own hosted blog arrived with 1,716 words without JavaScript, exactly as many as in a browser.
URLs
blog.your-domain.com for the overview, every article at a readable URL built from its headline. If you withdraw an article and publish it again later, it keeps its URL.
Page title, description, previews
From Claire’s page title and meta description, plus a canonical link and Open Graph tags for previews in messengers and social networks.
Structured data
Article schema for every article with headline, date, image and publisher, Blog schema on the overview.
Sitemap, robots.txt, RSS
/sitemap.xml with every article, /robots.txt pointing to it, /feed.xml with the latest 50 articles. All of it grows with every published article.
Visibility
One switch: “Visible to Google and AI search”. While it’s off, the blog stays closed to search engines, which is handy while you prepare.
Design
Color, font, logo, name and tagline, links to your main site and your legal pages. For developers: custom CSS plus your own header and footer, for example with your main site’s menu. Claire strips out scripts.
Languages
Your main language sits directly under the blog address, every other one under a prefix like /de/, with a language switcher and hreflang links.
Your Lovable site
Stays untouched; Claire needs no access to it. You add the “Blog” link to your navigation once yourself, for example with a prompt in the builder.

When an article goes live

Approve with one click

Every finished article waits in Claire. You read it, change whatever you like and publish it with “Publish to blog”. Only then does it appear at your blog address.

Take it back any time

With “Not published after all” you remove an article from the blog, the sitemap and the feed. It stays in Claire, and so does its URL.

Good to know

The hosted blog is meant for websites without a CMS. If WordPress, Shopify, Webflow or Shopware is already connected, Claire publishes there directly. On the hosted blog, every article goes online only once you approve it.

How Claire finds topics, researches and writes is shown on the homepage.

See how Claire works

Bolt, v0, Base44, Replit, Claude Code

Not built with Lovable? Here is how the others compare.

Whether a crawler sees text doesn’t depend on the builder’s name but on where the page is rendered. An overview based on the vendors’ documentation as of October 2026 and on our test.

01

Bolt

According to its docs Bolt builds with many JavaScript frameworks, according to its website with React and Vite by default. The docs say nothing about server-side rendering or pre-rendering. In our test none of the six Bolt sites delivered content without JavaScript. Bolt hosting also blocks requests that identify as GPTBot or ClaudeBot and had neither a robots.txt nor a sitemap in our test. Whether the block also hits the real crawlers can’t be checked from outside.

For SEO, Bolt offers templates in its Prompt Library and an SEO/GEO skill; custom domains come with Pro.

02

v0 by Vercel

v0 generates Next.js apps, and according to Vercel, published sites are server-rendered by default. Our test shows the same: on five of six v0 sites the full text was in the HTML. Of the builders here, that’s the best starting point. v0 doesn’t mention a sitemap or robots.txt, so have those created as well.

03

Base44

Base44 apps are React apps built with Vite that run in the browser. According to its docs, Base44 sends crawlers a rendered version. In our test, the request identifying as GPTBot got a median of 872 words, about as many as the browser shows; a regular request without JavaScript got only 379. On a custom domain Base44 creates the sitemap, robots.txt and canonicals itself, but no per-page meta descriptions. Custom domains start with the Starter plan.

04

Replit

New Replit apps have started from a React template since January 2025; server-side rendering isn’t built in. In our test two of six Replit sites had their text in the HTML. On the other four, fewer than 200 words arrived without JavaScript, compared with 674 to 1,549 in a browser. Replit’s paid SEO Agent adds a robots.txt and a sitemap, among other things.

05

Claude Code

Claude Code isn’t a website builder but a coding agent: it builds what you ask for. Whether the site is readable depends on the framework you choose. In our test six of seven sites, built with Astro, Next.js or as static HTML, had their text in the HTML. The one plain React app built with Vite delivered zero words. So say explicitly when you build it that every page should be pre-rendered at build time or rendered on the server.

06

The same for all: no blog

None of these builders ships a blog. Claire’s hosted blog works with any website whose domain lets you add a DNS record, whatever it was built with.

What you need

Included from Starter. €199 a month.

Free trial, no credit card. Cancel monthly, no setup fee. An additional language and the full SEO strategy come with Growth (€299).

Compare all plans

Starter

cancel monthly

€199/ month

  • Up to 30 articles a month
  • Hosted blog on your subdomain
  • Sitemap, RSS feed and structured data
  • One-click approval for every article

Questions

Good to know.

Does Lovable have good SEO?

Better than in early 2026. Projects created since May 13, 2026 are rendered on the server; older ones get a pre-rendered version for crawlers Lovable verifies, on Lovable hosting. Titles, descriptions, sitemap and robots.txt don’t appear by themselves, so check them with the SEO review. And rankings need content: a landing page on its own answers hardly any search.

Can ChatGPT see my Lovable site?

For new projects, usually yes: the text is in the HTML. Still check the page source, because content loaded later via JavaScript is missing there too. For older projects on Lovable hosting, Lovable says it sends verified AI crawlers a pre-rendered version. If you host an exported project yourself, AI crawlers see an empty page without pre-rendering, because they don’t run JavaScript.

Is SSR better than a single-page app for SEO?

For pages that should be found, yes. With server-side rendering or pre-rendering the content is in the HTML right away, for people, Google and AI crawlers alike. A single-page app is fine for logged-in areas like dashboards; for marketing pages and blogs it’s the weaker choice.

Can I add a blog to my Lovable site?

Yes, as pages in code or through a connection to Contentful, Storyblok, Notion or WordPress. Lovable has no built-in blog with an overview, a feed and approvals. For regular articles a blog on a subdomain is usually easier, because not every article is a change to your app.

Subdomain or subfolder for the blog?

With an AI builder, a subfolder like /blog needs your own reverse proxy; a subdomain needs one DNS record. That’s why Claire uses a subdomain. Link blog and main site both ways so visitors and search engines see they belong together.

Do I have to change anything in Lovable for Claire to blog?

No. Claire needs no access to your Lovable project. You add one DNS record and, if you like, a “Blog” link to your navigation. Claire serves everything else.

Is SEO still worth it in 2026, now that everyone asks ChatGPT?

Yes. ChatGPT, Perplexity and Google’s AI Overviews base their answers on pages they can read. If your HTML shows nothing, you don’t appear in those answers. Readable pages and articles that answer questions are the foundation for both.

More integrations

Using a different system? Claire works with that too.

What else Claire does for you besides the blog is on the homepage.

Try it on your own website. Set up in 15 minutes.

Enter your website, connect your blog, read the first draft. Claire takes care of the rest.

No credit cardSet up in 15 minutesCancel monthly