Deep Research Just Got a Serious Upgrade !

OpenAI has quietly rolled out substantial upgrades to its Deep Research agent, significantly enhancing its capabilities for comprehensive investigation and analysis. The improvements, first noticed by users after a period of apparent development, include two major enhancements: automatic generation of tables of contents for research outputs and a dramatic increase in the number of sources searched per query. In recent tests, the agent processed 337 sources for a single research request, representing what users estimate to be a 2-3x increase over previous versions. This expansion in source depth suggests OpenAI is prioritizing more thorough investigation capabilities, potentially using improved web search integration or more sophisticated query decomposition.

The technical improvements address key limitations in AI research tools—specifically source breadth and output organization. By automatically generating tables of contents, Deep Research now provides better-structured overviews of complex topics, making lengthy research outputs more navigable. The source count increase from approximately 100-150 to 337 sources per query represents a substantial leap in research thoroughness, though it may come with increased computational costs and processing time. These enhancements position Deep Research as a more serious tool for professionals needing comprehensive literature reviews, competitive analysis, or deep-dive investigations, potentially reducing the need for manual source gathering and organization.