SEO for sites built with Claude Code
Claude Code will happily build your entire site: architecture, code, deploy scripts, the lot. What it builds is exactly what you asked for, and that is the catch. SEO was probably not in the prompt, so it is probably not in the site.
The 'nobody asked' problem
Agent-built sites have a distinctive failure pattern: they are technically clean but commercially invisible. The agent chose sensible defaults for everything you specified and skipped everything you did not. In practice that means sites launch without buyer-intent titles and descriptions, without a sitemap or robots.txt, without llms.txt, without structured data, and sometimes with client-rendered content that AI crawlers, which do not execute JavaScript, cannot read at all.
The irony is sharp: a site built by AI that cannot be read by AI search engines. ChatGPT, Perplexity and Gemini answer buyer questions by reading sites. If yours is unreadable or undescribed, the answer cites a competitor.
You could prompt it yourself, if you knew what to prompt
Claude Code can implement any of these fixes brilliantly, once someone tells it precisely what is broken on the live site. That diagnosis is the hard part: what crawlers actually see, which tags are missing or duplicated, whether your robots.txt blocks something important, whether AI engines currently mention you at all. An audit from the outside is the missing input.
- Scan the live site the way Google and GPTBot see it
- Fix metadata, canonicals, sitemap, robots.txt and llms.txt
- Ensure key pages ship real HTML, not JavaScript-only content
- Add JSON-LD structured data for your product, pricing and FAQs
- Verify every fix with a re-scan, then monitor for regressions
Drazen plus Claude Code is the natural pairing
Drazen audits from the outside, writes the fixes, and ships them as GitHub pull requests you approve. And if you prefer to keep all changes flowing through your own agent, Drazen's fix prompts are written to be pasted directly into Claude Code: your agent applies them, Drazen re-scans and confirms. Diagnosis and verification from one agent, implementation by the other, and continuous monitoring so later refactors do not quietly undo the work.
Questions
Why not just ask Claude Code to 'do the SEO'?
Because without an external audit it is guessing at what is broken. It cannot see your site the way crawlers see it, check what AI engines currently say about you, or verify afterwards that the live site changed. Drazen supplies exactly those three things.
Will Drazen's pull requests fight my agent's code style?
PRs touch metadata, configuration and markup, and you review every diff before merging. If you would rather your own agent write the code, use Drazen's fix prompts instead and let it verify the result.
What is GEO and do I need it?
GEO (Generative Engine Optimization) is making sure AI engines read, understand and cite your site. If your buyers ask ChatGPT for recommendations, and increasingly they do, then yes.
See what your site scores
Drazen scans your site the way Google and ChatGPT see it, shows you everything that is hurting you, and writes the fixes itself. Free, about 30 seconds.
Run the free scan →