GPT-6's New UI Feels Less Like a Chat and More Like a Website
Buttons, forms, charts, and a calculator inside the answer. Clever, but I caught myself trusting the interactive widget a little too fast.
The Intelligent UI in GPT-6 is a genuine change, not a reskin. Ask a comparison question and it returns a side-by-side table you can poke. Ask something procedural and it drops in a small form or a working calculator. For the first time in a while a chatbot answer felt like a small app instead of an essay.
The part that gave me pause wasn't the capability, it was my own behavior. I accepted a number from the in-chat calculator without rechecking it, because it looked like a widget, not a guess. The interface signals competence, and that signal is exactly where the risk lives.
I've started a rule: interactive outputs get the same skepticism as prose ones. If the model can't show the source or the assumption, the pretty UI doesn't earn my trust. A calculator is only as good as the formula it quietly chose.
Still, I like it. For explaining something to a non-technical friend, a tappable answer beats a paragraph every time. The win is real; the discipline is just older than the feature.
Updated October 28, 2026: the behavior note below - accepting widget numbers too fast - matches what most early users report.
Frequently asked questions:
Q: Is GPT-6's Intelligent UI a real change or a reskin?
A: Real. Ask a comparison question and it returns a pokeable side-by-side table; ask something procedural and it drops in a form or a working calculator.
Q: What surprised the author most?
A: His own behavior - accepting a number from the in-chat calculator without rechecking, because it looked like a widget, not a guess.
Q: When is the UI most useful?
A: Explaining something to a non-technical friend, quick comparisons, budgets - anywhere a small interactive artifact beats an essay.