PDFs are designed for human eyes, not machine reading. AI struggles with them because of, 1. Text in columns, sidebars, and tables breaks logical flow. 2. Many PDFs are just images of text, requiring error-prone OCR to convert. 3. No semantic HTML/metadata exists to tell the AI what is a header, table, or paragraph. 4. Page numbers, footers, and images confuse context, causing "hallucinations" or skipped information.