GPT-6's Intelligent UI Makes Chat Feel Like a Tiny Website
The new buttons, forms and charts inside ChatGPT are useful -- until they hide the source.
OpenAI pushed GPT-6 to everyone in early October, and the headline feature isn't the model, it's the Intelligent UI: answers can now contain graphics, buttons, forms and little calculators. For a quick comparison or a budget, that's genuinely faster than copying numbers into a spreadsheet.
The risk is the same as every wizard: the interface answers the question you asked, not the one you meant. A side-by-side chart looks finished, but I still had to check which inputs it assumed. The UI invites you to stop thinking.
My rule now is to treat the interactive widget as a draft. It earns trust only after I've confirmed the numbers it used. The model got cheaper to run this month; that doesn't make its assumptions free.
Used that way, Intelligent UI is a real step forward. Used as a vending machine for conclusions, it's a nicer-looking way to be wrong. The button is not the answer.
Updated October 28, 2026: the trust rule below has aged well - treat every interactive widget as a draft until you've seen its assumptions.
Frequently asked questions:
Q: What is GPT-6's Intelligent UI?
A: Answers can contain graphics, buttons, forms and little calculators. For comparisons and budgets it's genuinely faster than copying numbers into a spreadsheet.
Q: What's the risk?
A: The interface answers the question you asked, not the one you meant. A widget signals competence, and that signal is exactly where over-trust lives.
Q: How should I use it safely?
A: Treat the interactive output as a draft. It earns trust only after you've confirmed the numbers and assumptions it used. The button is not the answer.