ChatGPT’s New Interactive Quiz Feature Exposes Knowledge Gaps in Five-Subject Test

In a hands-on report, the writer requested a college-level quiz on the history of AI with 10 multiple-choice questions, asking ChatGPT not to reveal answers until each response and to explain anything wrong before giving a final score. The one-question-at-a-time format made a bigger difference than expected, as the chatbot reacted after every answer with a brief explanation before moving on. The writer admitted to being surprised at the difficulty but said the experience taught them new facts.

The next subject was astronomy, which the writer felt confident about after years of reading about space. After telling ChatGPT that ordinary solar system trivia was too easy, the questions turned to orbital mechanics and astronomical measurements, and the writer barely got half right. The test showed the ability to adjust difficulty on the fly, something a fixed quiz cannot offer. The writer also found they could tell ChatGPT to make all four multiple-choice options plausible rather than including obviously wrong answers.

A 1990s pop culture quiz, with a request to mix television, movies, music, video games, and technology questions and require more than just remembering titles, produced a perfect score. The writer had to guess on some late-1990s boy band composition questions but did well. A Shakespeare quiz, which explicitly avoided extremely famous quotes and obvious plot questions, focused on secondary characters and plot details, and the writer noted that ChatGPT could be told to track weak areas and revisit them, or upload class notes for targeted testing.

The final category was world geography. At ChatGPT’s suggestion, the writer asked for the difficulty to increase every time they got three answers in a row correct. After breezing through six questions, the writer was left struggling with obscure mountains and deep-sea trench locations. Once the quiz was over, the writer asked ChatGPT to analyze the answers, and the AI said it could distinguish careless mistakes from repeated knowledge gaps and offered to create a new round of questions on those weak areas.

TechRadar’s review also carried a caution: ChatGPT can still get things wrong and hallucinate answers even in a quiz, so users should independently check any answer that seems off. For casual learning, the feature worked well, but the writer stressed that users still have to do the studying themselves and can use ChatGPT to keep practicing before a classroom quiz.