do the “deep research”, “agentic” AI tools know how to avoid AI slop?