I audit whether AI crawlers can actually read your site, and hand you a prioritized list of fixes with the numbers I measured — not promises about ChatGPT.

What I check and measure:

• Server-side rendering: how many words a crawler that does not run JavaScript actually sees. On my own site that went from 6 to 152.

• robots.txt for the AI user-agent set — GPTBot, ClaudeBot, anthropic-ai, PerplexityBot, OAI-SearchBot, CCBot, Applebot-Extended, Bytespider, Google-Extended — validated against what your server really returns, not just a 200.

• JSON-LD: typed facts a parser can extract, and entity resolution via sameAs to canonical IDs.

• Content structure: answer-first, self-contained paragraphs, H2s written as real questions — what survives being chunked by a retrieval system.

• llms.txt and llms-full.txt if they fit you, and I will tell you straight that they are hygiene for agents, not a growth lever.

What I will not sell you: a guaranteed citation rate, a "+40%" figure (that number comes from a simulated benchmark and did not survive independent evaluation on real engines), or control over Google's AI Overviews. Extractability is a precondition for being cited, never a guarantee of it. Everything I measure comes with its error bar.

You get the before and after measurements where they apply, a prioritized fix list (critical first) with an effort estimate, and the exact snippets to paste.

De-risk: I run the crawler-visibility check on your homepage and send you that number before you commit. Fixed price, 100% async. Implementation quoted separately.