Nielsen's UX Roundup Revisits 2025 Predictions

Jakob Nielsen's latest UX Roundup returns to the predictions he made for 2025 and measures them against what has actually happened in the industry. The subtitle names the remaining agenda plainly: AI paradigm shifts, AI and doctor burnout, and the observation that useful AI sells.

The roundup format itself is the story here. A practitioner of Nielsen's standing publishing a retrospective — rather than another forward-looking forecast — signals a shift in how the UX community treats AI claims. The cycle of breathless prediction is giving way to accounting: which interface patterns survived contact with users, which AI paradigms shifted, and which did not.

The first theme, revisited predictions, implies a scorecard approach. For design teams, that matters more than fresh speculation. A prediction that fails instructs; one that lands confirms where to invest research and prototyping effort. Nielsen's willingness to publish the audit, not just the original bet, is a methodological cue for any team running its own annual foresight exercises.

The second item, AI paradigm shifts, sits at the center of current interface debates. The question for designers is no longer whether conversational and generative systems enter the product surface, but which interaction model users actually adopt once novelty fades. Roundups of this kind typically separate durable shifts — changed user expectations, new error states, revised trust patterns — from temporary hype.

Third, AI and doctor burnout addresses the sharpest edge of the technology's deployment: clinicians. This is a usability story with stakes. When AI tools reach physicians, interface quality translates directly into cognitive load, documentation time, and ultimately patient interaction. The inclusion of burnout in a UX roundup frames it as a design failure mode, not merely an HR problem.

Finally, the observation that useful AI sells compresses a year of market behavior into one clause. The framing echoes a longstanding usability truth: features do not move products; task completion does. Users pay for tools that do work they recognize as theirs, faster or more reliably than before.

Read together, the four items sketch a coherent position. Prediction without follow-up is noise; paradigm talk without adoption data is marketing; and AI deployed into high-stakes clinical work is a usability liability before it is anything else. The selling point, as the headline puts it, is usefulness.

Nielsen's roundups continue on his Substack, where he applies evidence from user research to the AI transition as it unfolds.