SEO for AI-built websites

Millions of sites are now written by AI: prompted in Lovable or Bolt, generated by v0, agent-built in Cursor, Claude Code or Windsurf, shipped from Replit. They work, they look good, and the overwhelming majority are invisible on Google and unreadable to ChatGPT. Not because AI writes bad code, but because nobody asked it for the one thing that gets sites found.

The six problems almost every AI-built site shares

1. Placeholder metadata. Titles like "My App", empty descriptions, identical tags across pages. Search listings are built from these fields, and defaults rank for nothing.

2. JavaScript-only content. Many AI tools scaffold single-page apps: the HTML arrives nearly empty and JavaScript paints the content. Google renders JS slowly and expensively. AI crawlers (GPTBot, ClaudeBot, PerplexityBot) do not render it at all, so AI search reads a blank page.

3. No crawl infrastructure. Missing sitemap.xml, missing robots.txt, missing canonical tags. Crawlers are left guessing what exists and which version counts.

4. Nothing for AI engines to quote. No llms.txt, no structured data, no clear plain-language statement of what the product is, who it is for and what it costs. Engines cite sites they can confidently summarise.

5. Zero authority. New domain, no links, no mentions anywhere AI engines read. Rankings and AI citations both lean on evidence from the wider web.

6. Regression by regeneration. The same tools that built the site can rebuild it in minutes, and rebuilds silently wipe earlier fixes. SEO for AI-built sites is not a one-time task, it is monitoring.

Fix order that actually works

• Readability first: get real content into the initial HTML so every crawler can read you
• Metadata second: unique buyer-intent titles and descriptions on every page
• Infrastructure third: sitemap.xml, robots.txt, canonical tags, llms.txt
• Machine context fourth: JSON-LD structured data and Open Graph tags
• Then measure: Google Search Console + an AI visibility check (does ChatGPT mention you?)
• Then keep watch: re-scan continuously, because rebuilds eat fixes

Tool-specific guides

The pattern is shared, the details differ by tool. We keep a dedicated guide for each:

SEO for Lovable sites · SEO for Bolt sites · SEO for v0 sites · SEO for Cursor projects · SEO for Replit apps · SEO for Claude Code sites · SEO for Windsurf projects

Or hand the whole thing to an agent

This checklist is exactly what Drazen automates. It scans your site the way crawlers see it, finds what is hurting you, writes the code fixes itself and ships them as GitHub pull requests you approve with one click (or as paste-ready prompts for your AI editor). Then it monitors your site so the fixes survive the next rebuild. The scan is free and takes about 30 seconds.