A 2026 MIT-USC study of 4.5 million federal civil cases found pro se complaints increased from 11% in 2022 to 16.8% in 2025, with AI-generated text rising from 1% to 18% in 1,600 sampled documents via Pangram detector (Shah and Levy, 2026). Judge Braswell observations of hallucinated citations align with detector output. Vermont District filings jumped from 45 to 1,100 annually after 2022 Reddit guides on Copilot-drafted mandamus writs. The Technology Review account understates outcome data: pro se win rates showed no improvement despite clearer drafting, consistent with Levy's note that litigation requires tasks beyond text generation. Federal caseload statistics (uscourts.gov, 2025) confirm the volume spike predates AI but accelerates post-2023. Emerging state bills on chatbot liability, absent from the source, indicate downstream regulatory response to advice errors. Courts now face unaddressed questions of AI tool duties, extending beyond individual judge workload reports to systemic access-to-justice metrics.