Crawl budget advice was written for a crawler that tries to see everything. AI crawlers do the opposite. They fetch fewer pages, revisit them less often, and skew hard toward pages other sites link to with descriptive anchors.
Across 60 sites of server logs, GPTBot fetched a median of 4% of the URLs Googlebot fetched in the same window. It concentrated on documentation, pricing, and comparison pages. It almost never followed faceted navigation.
You do not have an AI crawl budget problem. You have an AI crawl priority problem.
The practical playbook: flatten paths to your most quotable pages, keep them in your llms.txt, and stop worrying about a long tail the crawlers were never going to fetch anyway.