OpenAI upgraded its Deep Research feature in ChatGPT, shifting from a passive "run it and wait" model to an interactive guided process. Users can now focus their research on specific websites, pull in context from connected apps, and make adjustments during the session. This flexibility is especially useful for those who regularly engage in source-based tasks, like analysts and journalists. The new output format also enhances usability, allowing for easier navigation through lengthy, citation-heavy reports. The update addresses the challenge of broad searches by letting users limit research to trusted domains. Integration with personal workflows—like email and calendars—means ChatGPT can pull in relevant information that might otherwise be overlooked. The ability to interrupt ongoing research and redirect the focus without starting over is key for iterative projects. OpenAI is aligning this new feature with its latest model, GPT-5.2, which supports a more agent-like experience that combines browsing, synthesis, and tool access. Anticipation is building around the upcoming GPT-5.3, but its release timeline and impact on Deep Research remain uncertain. In addition to these changes, there are hints of a new "Skills" layer that would allow users to create editable workflow instructions. This could enable better standardization across tasks and teams, offering a straightforward way to manage workflows without extensive customization.