The humble PDF is becoming a problem for AI

PDFs, designed 30 years ago to preserve visual document layouts, create significant challenges for AI systems that struggle with their fixed formatting and graphical coordinates. AI models trained on linear text often misread multi-column layouts, footers, and embedded graphics, leading to parsing errors and inaccurate interpretation. While startups like Factify are developing AI-friendly document formats and Adobe has added AI assistants to help interpret PDFs, the format's durability is evident in the estimated 2.5 trillion PDFs currently in circulation worldwide.