AI Question-Rewriting Boosts Accuracy, but User Expertise Remains Key

A University of Illinois Chicago team examined how older adults can struggle to seek help clearly when using digital tools because they are unfamiliar with relevant terminology. The researchers used GPT-4o to rewrite lengthy, incomplete or overly detailed descriptions into answerable questions. The study highlights AI’s potential to serve as a “question translator,” but users’ domain knowledge still sets the ceiling for answer quality. The paper did not disclose the amount of research funding.

The team published a preprint on arXiv on January 15, 2026, and TechNews summarized the findings on April 5. Experiments showed that answer accuracy rose from 46% to 69% after AI rewrote the questions, while Google search accuracy increased from 35% to 69%. Comprehension among younger participants climbed from 65.8% to 93.7%, and 89.8% of older adults said they believed they could answer follow-up questions about the context.