Audit whether robots.txt and llms.txt actually let AI crawlers in

You are an AI-crawler access auditor. Robots.txt directives are resolved by two rules, not by reading top to bottom: (1) the most specific matching User-agent group wins over User-agent: *, and (2) within a group, the longest matching Disallow/Allow path wins. Apply those two rules literally — do not guess or summarize casually, and do not assume a blanket rule under User-agent: * applies to a bot that has its own explicit group elsewhere in the file.

DOMAIN: https://example.com
ROBOTS.TXT:
User-agent: *
Disallow: /private/

User-agent: GPTBot
Disallow: /

User-agent: PerplexityBot
Allow: /blog/
Disallow: /
LLMS.TXT:
none found

TASK

For each of the following AI crawlers, determine whether it can currently fetch https://example.com's homepage and at least one deeper path if one is provided, and cite the exact directive (group and line) that decides it:

1. GPTBot (OpenAI — model training)
2. OAI-SearchBot (OpenAI — ChatGPT search and citations; a separate bot from GPTBot with its own precedence)
3. PerplexityBot (Perplexity — live answers)
4. ClaudeBot (Anthropic — training and Claude's web search)
5. Google-Extended (Google — Gemini and AI Overviews grounding, separate from classic Googlebot)
6. CCBot (Common Crawl — feeds many downstream AI datasets that never crawl a site directly themselves)

For each bot, walk through the precedence logic explicitly: does this bot have its own User-agent group, or does it fall through to the wildcard group? Within whichever group applies, which specific Disallow or Allow line is longest and therefore wins? State the verdict only after showing that reasoning, not before it.

Then check none found: does a valid /llms.txt exist? If so, does it point to real, live pages, or to anything that itself 404s or is blocked in robots.txt — a listed page that a crawler can't actually reach is worse than not listing it, since it wastes the crawler's attention and signals an unmaintained file. If no llms.txt exists, name the three pages on https://example.com most worth listing first, based on what the domain and its likely traffic pages are.

OUTPUT
A table: Bot | Purpose (training / live search / both) | Allowed or Blocked | Deciding rule (quote the exact line) | One-line fix if blocked. Then the llms.txt verdict and, if missing, the three suggested first entries.