HTML carries browser baggage
HTML contains navigation shells, wrappers, tracking-friendly attributes and presentational fragments that do not improve semantic understanding. Models pay for those tokens anyway.
Technical blogTechnical article
HTML vs Markdown for LLM pipelines
Raw HTML preserves browser detail that models rarely need. Markdown keeps the document shape while dropping most layout overhead, which makes it a better default boundary format for LLM systems.
Primary use
Why Markdown is usually a better interchange format than raw HTML for retrieval, prompting and agent memory.
Recommended flow
Fetch, clean, measure tokens, then hand consistent Markdown to agents or retrieval systems.
Next step
Use the Playground to compare raw HTML against optimized output before integrating the API.
Markdown keeps the useful structure
Headings, lists, tables, links and code blocks survive in Markdown with much lower overhead. That makes chunking, diffing and cache invalidation simpler.
When HTML still matters
Keep the original HTML if you need precise rendering, DOM-targeted actions or post-processing. But for retrieval and prompt context, a cleaned Markdown derivative is usually the better working format.
Internal links
Related technical paths
Preparation layer
LLM-ready content from messy HTML
Prepare documents for models with clean headings, preserved code and lower-noise context windows.
Cost control
Reduce HTML token usage before RAG or agents
Shrink token spend by converting bloated HTML into compact Markdown before chunking, prompting or embedding.
Reader API
URL to Markdown API for AI ingestion
Convert live pages into clean Markdown with stable token metrics and predictable output for AI systems.