Deep Research Is Now GPT-5.2 And Here's Everything That Changed

Deep Research was always the feature you'd kick off and then walk away from, hoping the report it handed back wasn't off in the weeds. That just got better.

ChatGPT's Deep Research now runs on GPT-5.2, and the headline change isn't raw smarts. It's control.

Why "controllable" is the real upgrade

A research agent that's smart but ignores your direction is a fancy way to waste twenty minutes. The win with the GPT-5.2 swap is that the process feels more steerable. You point it, it stays pointed.

That matters more than another benchmark point. (A report you can actually direct beats a slightly smarter one that wanders off topic and comes back with the wrong continent.)

A research tool you can steer is worth more than a smart one you can't.

If you use Deep Research for anything real, give it another run now that it's on GPT-5.2. The difference shows up in how often the output is actually what you asked for. Go test it.