Lovable and Durable Review: Are AI Website Builders Bad for SEO and AEO?

Are websites built with Lovable, Bolt, or Replit invisible to Google and AI search?

Not to Google, but possibly to AI search. Tools that generate code, such as Lovable, Bolt and Replit, often produce React apps that render in the browser unless you ask for, or the tool defaults to, a framework with server rendering. When a crawler requests such a page, it receives a small HTML file and a bundle of JavaScript. The content appears only after the script runs.

Google handles this: Googlebot renders JavaScript, though rendering can add delay. Most AI crawlers do not. A December 2024 study by Vercel and MERJ found that none of the major AI crawlers it examined, including OpenAI's, Anthropic's and Perplexity's, rendered JavaScript. Google's Gemini uses Googlebot, and Applebot renders pages. A client-rendered site can therefore rank on Google and still be an empty page to ChatGPT's search crawler. How to fix that generally is covered in how to make JavaScript-rendered pages readable to AI crawlers.

What changed in Lovable in 2026?

Lovable's documentation now describes three layers:

• Server-side rendering for new apps. Lovable apps created from 13 May 2026 use TanStack Start with server-side rendering, so pages arrive as complete HTML.
• Pre-rendering for older apps. Older React and Vite apps get on-request pre-rendering on deployed public URLs. It is served only to verified search and AI crawlers, which Lovable lists as Google, Bing, social-preview bots, and AI engines such as ChatGPT, Perplexity, Claude and Gemini. It runs at request time, so dynamically loaded content is included. Older projects can be upgraded to TanStack Start for full server-side rendering.
• An SEO and AI search review. Free on all plans, it audits sitemaps, robots.txt, metadata, structured data, semantic HTML, alt text, canonical tags, indexing and AI readiness. It can connect Google Search Console, verify the site and submit the sitemap. Semrush-powered keyword research is included at no extra cost until 30 November 2026.

Lovable's own documentation also notes that sitemaps, robots.txt and metadata are not always generated up front, and that the review must be run again after changes because it does not rerun automatically.

Why is my Lovable-built website not visible in AI search?

Usually one of these:

1. An older project on pre-rendering only. Crawlers outside Lovable's verified list receive the JavaScript shell. Upgrading to server-side rendering removes that dependency.
2. Missing basics. No sitemap, a default robots.txt, or the same title and description on every route.
3. Still on the lovable.app subdomain. Lovable recommends a custom domain to build search presence on a domain you control.
4. Not connected to search tools. No Search Console or Bing Webmaster Tools verification, so pages are discovered slowly. Bing matters because it feeds Copilot.
5. Nothing to cite. A new site with generic copy and no links or mentions elsewhere gives assistants little reason to use it, however well it renders.
6. Blocked crawlers. A CDN or security setting blocking AI crawlers. Cloudflare, for example, changed its defaults in 2026 to block training and agent bots on some new domains.

How long a new site takes to be picked up is covered in how long Google and AI search engines take to pick up a new website.

How does Durable compare?

Durable is a different kind of tool: an AI builder for small businesses that generates a hosted website, plus a CRM and marketing tools, rather than code you own. Its blog says Durable sites have a sitemap at /sitemap.xml, crawlable links and automatic indexing, and that you verify Search Console by pasting a tag into the site's head code. Its pricing page says all plans include SEO and hosting on Cloudflare's network, and its paid Launch plan adds "Advanced SEO & GEO" with a weekly visibility ranking, custom code and a free custom domain. Higher tiers include daily AI blog posts.

What Durable does not document on the pages we read is how pages are rendered for crawlers. That is easy to test, as shown below. Two other cautions apply. Daily AI-generated blog posts are the kind of commodity content Google's guidance warns adds little, so review anything published under your name. And check how you would leave: Durable says it will transfer your custom domain on request if you cancel, but the site itself lives on its platform.

How do you handle SEO and AEO for a vibecoded website?

A checklist that works for any builder:

1. Check what crawlers receive. Open "view source" on a key page, not the browser's inspector, and look for your headline and main text. Or fetch the page with a crawler's name, for example curl -A "OAI-SearchBot" https://yoursite.com/pricing, and search the output for your content. On an older pre-rendered Lovable project this test always returns the shell, because a faked user agent is not a verified crawler; use the live test in Search Console's URL Inspection instead, or move the project to server-side rendering.
2. Use server rendering or pre-rendering if the content is missing. In Lovable, new projects already do this; older ones can be upgraded.
3. Give every page its own URL, title and description. Avoid hash-based routes such as /#/pricing.
4. Generate a sitemap and a sensible robots.txt, allowing Googlebot, Bingbot, OAI-SearchBot and other search crawlers you want.
5. Use a custom domain and verify it in Search Console and Bing Webmaster Tools.
6. Link pages with real anchor links, not only buttons that change state.
7. Write pages that answer real questions: what you offer, for whom, where, at what price and why you.
8. Add structured data where it describes real content, such as your organisation, products or local business.
9. Recheck after every major change, since generated code can quietly change routing or metadata.

Where does each fall short?

• Lovable: pre-rendering on older projects covers a named set of crawlers only; SEO basics may need generating and rechecking.
• Durable: rendering method not documented; AI-generated content needs editing; the site is tied to the platform.
• Both: a well-rendered site with generic content still has little reason to be cited.

We have not built and tested sites on Lovable or Durable ourselves; this review draws on their documentation, pricing pages and blogs, and on published crawler research.

What might this look like in practice?

Consider an illustrative bookkeeping firm, Tallwood Accounts, whose founder built the site in Lovable in early 2026. Google indexed it, but ChatGPT never cited it. The project predated server-side rendering. The founder's curl test with OAI-SearchBot's user agent returned an almost empty page, but that test cannot show what real crawlers get on a pre-rendered Lovable project, because Lovable serves pre-rendered HTML only to crawlers it can verify. The founder upgraded the project to TanStack Start, moved it from the lovable.app subdomain to a custom domain, ran Lovable's SEO and AI search review to generate a sitemap and distinct page titles, and rewrote the services page to state fees and specialisms. A month later, ChatGPT's answers to "bookkeeper for Etsy sellers in Bristol" began citing the services page.

What do people ask about AI website builders and SEO?

Does Google index Lovable sites?

Yes. Googlebot renders JavaScript, and new Lovable projects are server-rendered anyway.

Can ChatGPT read a client-side rendered site?

Often not. Research has found that OpenAI's crawlers do not render JavaScript, so content must be in the HTML.

Is Durable good for SEO?

It covers basics such as sitemaps, meta tags and Search Console verification. Test what crawlers receive, and edit AI-written content.

Should I rebuild my site on another platform?

Only if you cannot get content into the HTML crawlers receive. Server rendering or pre-rendering usually fixes the core problem.