Gary Illyes: JavaScript Mess Harms AI Web Processing

Google analyst Gary Illyes states that AI agents and large language models would have had an easier time processing the web if sites had used clean HTML and avoided JavaScript. He made this observation on LinkedIn. What's interesting is that this statement comes as the industry debates solutions like llms.txt files and structured data layers for AI. Illyes suggests these efforts are remedial. He believes the problem stems from decades of web architecture choices. He notes that if sites used good HTML and no JavaScript, or server-side rendering, both base model training and web and agentic RAG would be much simpler. Modern web content often relies on JavaScript, meaning the raw HTML is often blank to crawlers. This makes Retrieval-Augmented Generation, or RAG, a complex engineering challenge for AI agents. The bottom line: Illyes highlights how past web development choices are now creating significant hurdles for AI systems trying to understand and use web content.