Reading time: 13 minutes
A site built on Lovable can show up on Google and ChatGPT, as long as the crawler receives the text in the HTML. It’s the first check for Lovable SEO, and it depends on the stack and hosting. Since May 2026, new apps ship with server-side rendering. But older React + Vite apps depend on Lovable’s pre-rendering. Here’s how to test and fix it.
In short
- Apps created on Lovable since May 13, 2026 use TanStack Start and deliver fully rendered HTML. Earlier ones use React + Vite and build the page in the browser.
- On Lovable, older apps serve pre-rendered HTML only to verified crawlers, including AI search crawlers. Once exported, however, we infer that they lose this feature.
- Googlebot executes JavaScript, but the AI crawlers tested in 2024 did not. A report of a change at OpenAI remains unconfirmed.
- For an older app, the most direct SEO fix is upgrading to TanStack Start inside Lovable. It costs 10 to 35 credits in most cases.
Why doesn’t a site built on Lovable show up on Google?
There are four common causes. To begin with, the crawler may not receive the text in the HTML, which depends on the stack and the hosting. The other three are a URL that can’t appear in search, generic metadata without a sitemap, and navigation built with buttons instead of links.
According to Lovable’s documentation, published apps deliver HTML that search engines and AI crawlers can read. Apps created since May 13, 2026 (June 22 on Enterprise plans) use TanStack Start, with server-side rendering. Older apps, on the other hand, get on-the-fly pre-rendering, only for verified crawlers. These include Google, Bing and the AI search crawlers of ChatGPT, Perplexity, Claude and Gemini. The change appeared in Lovable’s changelog on May 18, 2026.
| Stack | Where the page is built | Who receives HTML with the text |
|---|---|---|
| TanStack Start | On the server | All visitors |
| React + Vite on Lovable | In the browser | Only verified crawlers |
| React + Vite outside Lovable | In the browser | No one (our inference) |
The last row is our inference, because the documentation describes pre-rendering only for addresses published on Lovable.
To find out the stack, open package.json. According to Lovable’s external hosting guide, TanStack Start projects list the @tanstack/react-start package.
What are the other common Lovable SEO problems?
Besides the HTML, check the other three causes:
- A URL that can’t appear in search. Unpublished projects, projects with restricted access and branded workspace URLs never show up.
- Generic metadata and no sitemap. The starting metadata is generic, and Lovable doesn’t always create the sitemap.
- Links the crawler doesn’t follow. Buttons that switch pages with JavaScript are not links, so the crawler doesn’t follow them.
Before the May change, in February 2026, Andrea Saez reported that a crawl test showed her Lovable site as empty to Googlebot. She also found 20 buttons where links should have been.
Can ChatGPT and other AI crawlers read a Lovable site?
They can, as long as the text is already in the HTML. Googlebot executes JavaScript, but in a queue. In a December 2024 study, however, the crawlers from OpenAI, Anthropic and Perplexity did not execute it. A September 2026 report suggests a change at OpenAI, but without official confirmation. Either way, Lovable says it serves server-rendered or pre-rendered HTML to these crawlers.
According to Google, every page with a 200 status goes into a rendering queue, for a few seconds or longer. In addition, AI Overviews and AI Mode cite pages from Google’s index. The guide on how AI Mode and AI Overviews choose their sources explains this.
In the Vercel study with the consultancy MERJ, from December 17, 2024, none of the major AI crawlers tested rendered JavaScript, except Gemini. Note that Vercel sells hosting for Next.js.
| Crawler | Executes JavaScript? | Basis |
|---|---|---|
| Googlebot | Yes, with a queue | |
| Gemini | Yes, through Googlebot’s infrastructure | Vercel and MERJ, 2024 |
| OAI-SearchBot and GPTBot | Not in 2024; report of a change in 2026 | Vercel and MERJ; unconfirmed report |
| ChatGPT-User, ClaudeBot and PerplexityBot | No, in the test | Vercel and MERJ, 2024 |
The choice between rendering on the server or in the browser decides what each crawler sees. Ryan Siddle, from MERJ, summed up the study’s message:
“Brands must ensure that critical information is server-side rendered and that their sites remain well-optimized to sustain visibility in an increasingly diverse search landscape.”
Ryan Siddle, Managing Director of MERJ (2024)
Did OpenAI’s crawlers change in 2026?
The report of a change dates from September 30, 2026. The owner of a Japanese marketplace built on Next.js published his own logs on DEV Community and on the Hisashi Space blog. He also checked the IPs against OpenAI’s lists. According to him, since September 25, OAI-SearchBot and GPTBot have rendered pages on his sites. However, he says the change seems to apply only to some sites. He found no announcement, and OpenAI’s crawler documentation does not mention JavaScript. So treat the report as an unconfirmed signal.
How do you test the SEO of your Lovable site?
The first SEO test on a Lovable site is whether the crawler receives the text in the HTML. There are four ways to see what arrives without JavaScript: curl with the crawler’s user agent, the page source, JavaScript turned off and Search Console’s URL Inspection.
In an older app hosted on Lovable, however, the first three show an empty page. That is why Lovable recommends URL Inspection to check the pre-rendered version.
- Run curl with the crawler’s user agent. The strings below come from Google’s and OpenAI’s documentation, and OpenAI warns that the version may change. Then search the saved HTML for a sentence from the middle of the page.
- Open the page source with Ctrl+U. An empty block with the id root indicates a page built by JavaScript.
- Turn off JavaScript. In Chrome, for example, open the developer tools, press Ctrl+Shift+P, choose “Disable JavaScript” and reload.
- Use URL Inspection. With the site verified, the live URL test shows a screenshot and the page code as Google processed it, according to Search Console Help.
# Googlebot Smartphone (W.X.Y.Z is the Chrome version)
curl -sL -A "Mozilla/5.0 (Linux; Android 6.0.1; Nexus 5X Build/MMB29P) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/W.X.Y.Z Mobile Safari/537.36 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)" https://www.yourdomain.com/ -o googlebot.html
# GPTBot, from OpenAI
curl -sL -A "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.4; +https://openai.com/gptbot" https://www.yourdomain.com/ -o gptbot.html
# OAI-SearchBot, the ChatGPT search crawler
curl -sL -A "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36; compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot" https://www.yourdomain.com/ -o searchbot.html
# Lines with the sentence in each file (0 = the text is not in the HTML)
grep -c "a sentence from the middle of the page" googlebot.html gptbot.html searchbot.html
# Markdown version that Lovable serves to clients that ask for it
curl -sL -H "Accept: text/markdown" https://www.yourdomain.com/In Windows PowerShell, use curl.exe instead of curl and Select-String "your sentence" googlebot.html -SimpleMatch instead of grep.
How do you read the test result?
For an older app hosted on Lovable, expect an empty curl result. After all, anyone can spoof the user agent, as Google warns, and Lovable only serves pre-rendered HTML to verified crawlers. URL Inspection shows what Google reads, since Google also executes JavaScript.
| Curl | URL Inspection | Diagnosis |
|---|---|---|
| With the text | With the text | Every crawler receives the text |
| Empty | With the text | Google reads it; AI crawlers depend on pre-rendering |
| Empty | Empty or incomplete | Fix the rendering |
It’s worth repeating the test on product and pricing pages, because that’s where you find what usually disappears without JavaScript.
What we found when we tested 228 Lovable template sites
We tested the homepage of the 228 published sites in Lovable’s template gallery. They are demo templates, not company websites. In 211, the HTML had at most 20 words of text in the body. In 15, the text was already there.
With the Googlebot and GPTBot user agents, the result was the same, except for one site that blocks GPTBot. That doesn’t mean Google sees an empty page. After all, our test is not a verified crawler. Lovable also warns that third-party SEO tools may see an empty page where Google and AI search engines read the full content.
When we asked for Markdown, 221 served that version. In our test, 204 of the 211 responded with a 200 status to a nonexistent address. If Google gets the same response, the page may become a soft 404. The 15 with text in the HTML, however, responded with a 404. Only 9 of the 228 had a valid sitemap.xml, and 207 had a robots.txt file.
Finally, a retest of 31 of these sites confirmed the numbers; only one started serving Markdown.
How do you fix Lovable SEO: pre-rendering, SSR or exporting the code?
There are four paths: upgrade to TanStack Start on Lovable, keep the automatic pre-rendering, use an external pre-rendering service or export the code to a framework with server-side rendering, such as Next.js. The choice depends on the hosting and on how much control you want.
| Path | Pros | Cons |
|---|---|---|
| Upgrade to TanStack Start | Fully rendered HTML for everyone, without leaving Lovable | Costs credits; browser-only libraries may break |
| Keep React + Vite on Lovable | Automatic, at no extra cost | Only verified crawlers; late-loading content may be missing |
| External pre-rendering | Fixes an exported app without rewriting it | One more service; Google calls it a workaround |
| Export to Next.js | Full control of the code | Rewriting routes and maintaining a server |
Upgrade to TanStack Start
According to Lovable, the upgrade runs inside the project, on all plans, and costs 10 to 35 credits in most cases. The published site doesn’t change until the next publish, and you can roll back through the version history. Libraries that only work in the browser can break server-side rendering, so test every page before publishing.
External pre-rendering
On Lovable’s hosting, the company says you don’t need an external service. So services like Prerender.io are for those who moved an older app to another server. According to Prerender.io’s own documentation (the company sells the service), it intercepts the crawler’s request at the CDN or the server. Then it renders the page, caches it and returns the HTML.
Google calls this technique dynamic rendering and treats it as a workaround. It prefers server-side rendering, static pages or hydration. Still, it does not consider it cloaking when the content is similar. Cost, setup and when pre-rendering pays off depend on the size of the site and the hosting.
Export the code and migrate to Next.js
The code is yours. According to Lovable’s external hosting guide, Git sync is available on all plans, and the .zip download requires a paid plan. An older app becomes static files. The guide doesn’t mention pre-rendering outside Lovable, so we infer that the exported app loses it. A TanStack Start app, on the other hand, needs a server that runs the rendering.
Migrating to Next.js means rewriting the React Router routes and the data fetching, which is developer work. After the migration, rendering and metadata in Next.js need their own configuration.
How do you set up Lovable SEO: title, description, sitemap and robots.txt?
Through the chat and Lovable’s SEO review, in the SEO & AI tab. The review checks the sitemap, robots.txt, metadata, structured data and indexing. It can also fix most issues with one click. All plans include running the review, but the fixes use credits.
- Title and description. Lovable generates both for each page, but the starting point is generic. So ask for a change in the chat and publish again.
- robots.txt. Check that it allows Googlebot, Bingbot and OAI-SearchBot. The list of AI crawlers is in the guide on what GEO is.
- Sitemap. Ask Lovable to create it, then check that it lists only public pages, with full URLs on the final domain.
- Custom domain. It requires a paid plan. Set a primary domain, and the others redirect to it.
- Search Console. The Lovable connector verifies the site with a tag in the HTML and submits the sitemap through the chat.
- Bing Webmaster Tools. Import the site already verified in Search Console. This step matters because Microsoft says Bing powers AI tools like ChatGPT, the subject of the guide on how to show up in ChatGPT.
- Real links. Ask Lovable to replace navigation buttons with links that have an href on public pages.
After the fix, track indexing in Search Console and see how to measure your brand’s mentions and citations in AI answers. Visits from assistants show up in GA4, in an AI traffic channel.
Is it worth leaving Lovable because of SEO?
Only if the problem is bigger than rendering. That’s because, for a site hosted on Lovable, upgrading to TanStack Start solves crawler access without switching tools. Migrating makes sense when the team needs a CMS to publish content often or when the company has already decided to leave Lovable.
- Stay on Lovable if the site has few pages and the test shows the text in the HTML.
- Move to WordPress or Webflow if marketing publishes every week and needs an editor.
- Move to Next.js if there is a development team and the product mixes a website and a logged-in area.
In other builders, such as Bolt and v0, check the framework in package.json and run the same test.
Rendering is only one part of the SEO that supports brand visibility on Google and in AI. If crawlers already read the site and AI still doesn’t cite the brand, the cause lies elsewhere, such as a lack of mentions in third-party sources.
Frequently asked questions
Not necessarily. Apps created on Lovable since May 13, 2026 use TanStack Start and deliver fully rendered HTML to all visitors. On the other hand, older apps built with React + Vite build the page in the browser. They depend on Lovable’s pre-rendering, which only applies to verified crawlers.
According to Lovable, upgrading to TanStack Start costs 10 to 35 credits in most cases and is available on all plans. The published site doesn’t change until the next publish, and you can roll back through the version history.
Because older apps hosted on Lovable only serve pre-rendered HTML to verified crawlers, and third-party SEO tools are not among them. That is why Lovable itself warns that these tools may see an empty page. To check what Google reads, use Search Console’s URL Inspection.
Yes, with care. An exported older app becomes static files, and Lovable’s documentation only describes pre-rendering on addresses published on Lovable. That is why we infer that the exported app loses this feature and needs external pre-rendering or server-side rendering.
Want to know how your brand shows up in AI answers today?
NotHuman is opening spots for agencies and brands to track how they appear in AI answers. Join the waitlist and we’ll let you know when a spot opens up for you.
Main sources
- Lovable, Optimize your app for SEO and AI search (2026).
- Lovable, Upgrade a project to TanStack Start (2026).
- Google Search Central, Understand JavaScript SEO Basics (2026).
- Vercel, The rise of the AI crawler, with MERJ (2024).
- OpenAI, Overview of OpenAI Crawlers (2026).
