Study finds raw HTML wastes most tokens in web content processing for AI

A study analyzing 20 web pages found that raw HTML contains a median of 128,000 tokens, mostly from scripts and markup. Converting HTML to text retains unnecessary elements like menus and footers, tripling token counts versus actual content. Extracting main content reduces tokens by a median of 51% compared to full-page text, with additional cleanup saving about 9%. The savings vary by page type, with portal pages benefiting most and long articles like Wikipedia gaining less.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in