SEO for Lovable sites

Lovable can take you from prompt to a live product in an afternoon. What it does not do, unless you specifically ask, is make that product findable. Most Lovable sites launch with no meta descriptions, no sitemap, and content that AI search engines cannot even read. Here is what is going wrong and how to fix all of it.

Why Lovable sites struggle to rank

Lovable generates modern React apps, and by default a lot of them render client-side: the HTML that gets sent to a visitor is close to empty, and the actual content only appears after JavaScript runs in the browser. Human visitors never notice. Crawlers do.

Google can execute JavaScript, though it costs extra crawl budget and delays indexing. The AI crawlers are stricter: GPTBot (ChatGPT), ClaudeBot and PerplexityBot generally do not execute JavaScript at all. If your content only exists after JS runs, those engines see a blank page. You cannot be the answer if the engine never read the question you solve.

On top of the rendering problem, prompt-built sites usually skip the unglamorous plumbing: page titles that describe the product, meta descriptions, canonical tags, a sitemap.xml, a robots.txt, structured data, Open Graph tags for link sharing. None of this is Lovable's fault, you simply never asked for it, and neither did the prompt.

The checklist a Lovable site needs

• A real title and meta description on every page, written around what your buyers search for, not your project name
• Content present in the initial HTML (pre-rendering or server rendering), so AI crawlers can read it without executing JavaScript
• sitemap.xml and robots.txt, so crawlers know what exists and what to index
• llms.txt, a plain-language summary file AI engines use to understand what your site is
• Canonical tags, so www vs non-www and trailing-slash duplicates do not split your ranking signal
• Structured data (JSON-LD) describing your product, pricing and FAQs
• Open Graph tags, so shared links unfurl with a real title and image

How Drazen fixes a Lovable site

Drazen scans your site the way crawlers see it, finds every issue above, and then writes the actual fix: real code, not a PDF report. If your Lovable project is connected to GitHub, Drazen opens a pull request and you approve it with one click. If you would rather stay inside Lovable, Drazen writes a precise fix prompt you paste straight into the Lovable editor, then re-scans your site to confirm the fix went live.

After the first round of fixes, Drazen keeps watching. Lovable makes it very easy to regenerate your site, and regenerating tends to wipe out earlier fixes. When that happens, Drazen catches the regression and writes the fixes again.