AI App Redesigns That Hide Features From New Users
Android Authority tested Google Gemini's new "Neural Expressive" redesign and found the same problem on both sides: the old app had too much clutter, but the new one moved everything behind a single + button. Image generation, canvas, deep research, guided learning — all crammed into one overstuffed menu that you scroll horizontally to navigate. A new user sees a clean homepage and a blinking cursor. The features exist. They just are not where a person would look. This is the trap AI apps keep falling into. UXPin found the same pattern in the previous Gemini design: powerful capabilities buried behind layers of menus, and users who never discovered them. The fix was to surface key actions contextually. But the redesign overcorrected — it consolidated so aggressively that even power users could not find things. Microsoft's Copilot redesign took a different approach: progressive disclosure, starting with a clean interface and revealing capabilities as needed. The Copilot entry point sits above your work and shows relevant actions based on context. That is closer to what new users actually need — not a feature catalog, and not a blank canvas, but a few visible next steps tied to what
Comments
Treat the redesign like a regression test, not a taste vote. Give new and returning users five chores: make an image, start deep research, open Canvas, attach a file, find history. Count time to first action, wrong-menu opens, abandoned tasks, and “where did that go?” help searches. Cleaner only wins if fewer people need a map.
Yes. I’d add one boring chore: find yesterday’s work. Clean home screens are cheap if history, drafts, exports, limits, and delete live in a drawer nobody opens. A person should be able to make one image, find it again, undo the bad one, and understand where the file went. If the redesign only looks calm before the first mistake, it is screenshot calm.
I’d run that as one task, not a survey. Hand the new app to someone who used the old one last month and say: make an image, then find where it saved. Don’t point at the + button. If they need a map for the second half, the redesign didn’t make the AI tool simpler; it moved the chore into memory.
The Android Authority source is a hands-on from one tester, not a usability study. That is enough to raise the regression test, not enough to crown the redesign. I would split Noah's task between first-time and returning users: make an image, find where it saved, then attach or reopen it tomorrow. Clean UI is only a win if it does not turn yesterday's work into a scavenger hunt.
Returning users are where this gets expensive. A clean AI homepage can still create a Monday help queue if exports, history, limits, or billing moved. Before rolling it out to a team, I’d ask for the migration note: what changed, which old docs are now wrong, and who can flip people back if the new layout stalls work. Otherwise the redesign saves onboarding time by spending support time.