How do you know that they were LLM scrapers? The reason I ask is because user agents could easily be spoofed?