Lovable Site Not Showing on Google? Diagnose It First

Before any of that, find out which of Lovable's two rendering paths your project is on, because they fail differently and they test differently.

The 60-second output test

Four checks, all free, in the order that tells you the most for the least effort.

• ✓
  View source, not Inspect. On the live page press Ctrl+U, then Ctrl+F for your main headline. Inspect shows the page after JavaScript has run, which is the one view that cannot answer this question.
• ✓
  Fetch it as an ordinary client. Run the curl below and read what comes back. On an older React or Vite project you will see the single-page app shell: a near-empty body whose only real element is a mount point, usually <div id="root"></div>, with your content nowhere in the response. That exact string is what to search the output for.
• ✓
  Ask Google what it actually received. Search Console, URL Inspection, then View crawled page. This is the authoritative answer for Google and it beats every third-party scanner.
• ✓
  Pull the Agentic Browsing checks in PageSpeed Insights. The category is experimental, needs Chrome 150 or later, and reports a ratio of checks passed rather than a score out of 100.

1curl -sS -A "GPTBot" https://yoursite.com/ | head -c 2000

Step 1: confirm the project is eligible to be indexed

Before debugging anything, establish whether the URL is allowed to be indexed at all. Lovable's documentation states that unpublished, private and branded workspace URLs are never indexable, so if that is the address you have been checking, nothing you do to the content will help. Confirm the published canonical host, its status code, the robots directives and the sitemap contents.

• ✓
  Publish the current version to a public URL. Lovable says unpublished, private and branded workspace URLs are not indexable.
• ✓
  Open the intended canonical host and confirm the page returns the expected content and a successful status.
• ✓
  Check for a sitewide noindex meta tag, X-Robots-Tag header or robots.txt rule that blocks the page.
• ✓
  Confirm that the sitemap exists, uses the intended host and contains the page's current canonical URL.
• ✓
  Choose one primary custom domain and redirect alternate connected domains where the current product supports it.

Step 2: use Lovable's current review

Lovable's SEO and AI search review checks metadata, Open Graph data, structured data, content structure, robots.txt, sitemap, indexing, performance, accessibility and mobile usability. Published projects receive additional live checks. The scan does not rerun automatically on every publish, so run it again after a material change and confirm the result reflects the latest code.

Lovable also documents a Google Search Console workflow when its connector is enabled. Verify the intended property, publish the latest routes and submit the current sitemap. A passing platform scan is useful technical evidence, not a promise that a page will rank. Treat the performance part of that scan the way our repeatable PageSpeed process describes, with the profile and run count written down.

Step 3: inspect what Google fetched

Use Search Console URL Inspection for the exact canonical URL. Review index status, the user-declared and Google-selected canonical, crawl state and the crawled page. Test the live URL when appropriate. This is more probative than concluding that Google saw an empty page because a third-party scanner or a normal curl request returned the older app's SPA shell.

The crawl, index and rendered-output group in our competitor ranking gap analysis works from the same evidence, on any stack.

Lovable says its older-stack pre-rendering is limited to verified crawlers, while unverified agents receive the regular single-page app. Newer TanStack Start apps return server-rendered HTML more broadly. First identify the project stack and hosting path, then test the correct output.

Step 4: validate page-level signals

Once the page can be indexed, six signals decide whether it deserves to rank. Unique accurate titles and descriptions, one clear topic per route with meaningful internal links, an accurate self-referencing canonical, structured data only where it matches visible content, and a mobile experience that is measured rather than eyeballed. The last question is the honest one: does this page answer the query better than the pages already indexed?

• ✓
  Write a unique, accurate title and useful search description for each important route.
• ✓
  Use one clear page topic, descriptive headings, visible factual copy and meaningful internal links.
• ✓
  Set an accurate self-referencing canonical unless a deliberate consolidation strategy says otherwise.
• ✓
  Use structured data only where it matches visible content and the relevant search-engine policy.
• ✓
  Give important images descriptive alt text and keep the mobile experience usable and accessible.
• ✓
  Check whether the page actually satisfies the query better than indexed competitors.

The mobile experience in that list is measured, not judged by eye. Our guide to Core Web Vitals covers which field metrics to read and at what percentile.

Step 5: separate indexing from ranking

An indexed page can still receive no impressions because demand, intent, content quality, competition, links, location or SERP presentation work against it. Compare pages and queries in Search Console. Do not diagnose a rendering failure from “zero traffic” alone. If the change overlaps a Google rollout, check the dates against the record of the March 2026 core update.

Google says JavaScript sites should use unique titles and descriptions, meaningful HTTP status codes, crawlable links and careful canonical handling. Server-side or pre-rendering can help users and crawlers, but Google also processes JavaScript. Architecture is one input, not an automatic index or rank decision.

When a migration is actually justified

Four conditions have to hold at once, and the search consequences are covered separately in what a migration does to search visibility and in what goes into costing a website migration.

• ✓
  The platform you are on cannot meet a requirement you have written down about rendering, routing, integrations, accessibility, speed or governance.
• ✓
  The residual problem remains after current Lovable publishing and SEO controls are configured and verified.
• ✓
  The team accepts the new editing workflow, hosting, maintenance, security and total operating cost.
• ✓
  The migration plan covers the routes, the data, forms, logins, integrations, analytics, consent, redirects, testing and how to roll back.

Next.js can be an appropriate destination, but it does not automatically produce correct metadata, fast pages, rankings or citations. Preserve stable URLs where practical, document every URL disposition and define measurable acceptance in the signed scope. The stack we build on is described in our note on pairing Next.js with Sanity, and the route-by-route method in migrating a site to Next.js.

Delivery sits under custom engineering and, for a CMS-backed rebuild, our migration service. Scope tiers are on the pricing page and finished builds on the work page.

Do ChatGPT and Claude crawlers run JavaScript?

Design as though they do not, and be suspicious of anyone stating it as a current fact in either direction. The most-cited measurement is Vercel's crawler analysis, which found that GPTBot and ClaudeBot fetched JavaScript files, 11.5 and 23.8 percent of their requests, without executing them. That is a real finding from a large sample. It was also published in December 2024, which is a long time ago in this particular area.

Neither OpenAI nor Anthropic documents current rendering behavior for its crawlers. OpenAI's bots page names each agent and what it is for and says nothing about JavaScript. Treat every confident blanket claim about what AI crawlers can render as undated until someone shows you the date. The safe design does not change either way: keep the content that matters in the HTML your server sends, then read your own logs to see which agents actually requested the page. The wider glossary entry sits in our AEO and web performance glossary.

Do AI discovery files guarantee citations?

No. Lovable includes AI-oriented checks and documents crawlable output for named AI crawlers. An llms.txt file, Markdown view, schema block or framework can make information easier to discover or interpret, but no such control guarantees crawling, use, recommendation or citation by an external AI product. What we do build in, and why, is set out in our answer-engine playbook and on our AI information page.

Primary sources

• Lovable: optimize your app for SEO and AI search
• Lovable: publish your project
• Lovable: custom domains
• Google Search Central: URL Inspection
• OpenAI: crawlers and bots
• Vercel: the rise of the AI crawler (December 17, 2024)

Correction, July 24, 2026: an earlier version of this article said Lovable, Bolt and Replit all deliver an empty shell to every crawler, and presented migrating to Next.js as the fix. Lovable's documentation contradicts that, and the byte counts quoted were not reproducible. The article was rewritten around the documented behavior. Sources rechecked July 31, 2026.

Frequently asked questions