In a randomized experiment involving more than 1,000 first-year undergraduate students at Bocconi University, researchers found that access to ChatGPT and training in critical thinking improved student work in different ways, according to a study published by OpenAI on August 27, 2026, conducted in collaboration with OpenAI Economic Research. Students worked on a real-world business case developing marketing recommendations for the university's merchandise store. They were randomly assigned by class period into four groups: access to ChatGPT (GPT-4o), training in causal reasoning, both, or neither. Submissions were evaluated by trained human graders using a five-point rubric, and researchers also used automated text analysis to measure the number and variety of ideas, signs of causal reasoning, and similarity to expert submissions. According to the study, students with ChatGPT access scored almost a full point higher on the five-point scale, producing more ideas, clearer logic, and answers more similar to expert recommendations. Students who completed the critical-thinking exercise did not score higher on the rubric but produced a wider range of more distinct ideas. Students who received both showed gains across the widest range of measures. OpenAI stated the findings suggest assignments may need to adapt to measure qualities such as originality.
- More than 1,000 first-year Bocconi University students participated in a randomized experiment using GPT-4o.
- Students with ChatGPT access scored almost a full point higher on a five-point rubric.
- Critical-thinking training increased idea variety and originality but did not raise rubric scores.
- The study was conducted by Bocconi University researchers with OpenAI Economic Research and published August 27, 2026.
What it means for you
This is a study OpenAI ran (partly on its own product) showing AI helps people produce more polished, expert-looking work, while separate reasoning training helps people generate more original ideas. In plain terms: AI raises the floor of your output quality, but it doesn't make you think more originally on its own. The two skills do different jobs.
Try this
Next time you use ChatGPT for a work task, do the thinking first: jot down your own ideas and reasoning before you open the tool, then use AI to polish and pressure-test them. That combination is what the study found produced the best all-round results.
Who should care
Teachers, trainers, and anyone who evaluates other people's written work, plus small-business owners using AI to lift the quality of drafts, proposals, or marketing copy.
Skip this if
You don't teach, hire, or grade written work, and you already treat AI as a drafting assistant rather than a replacement for your own thinking.
Sources: OpenAI — read the original