ChatGPT, Gemini and Claude differ in reasoning styles on identical exam question

A comparative study by the RADTE research group at UNED tested three widely used large‑language models—ChatGPT, Gemini and Claude—by giving each the same university‑level exam prompt in Spanish. While all three produced fluent, well‑structured answers, the analysis showed distinct ways of organizing ideas. ChatGPT tended to list actions in a cumulative series, Gemini clarified and delimited the concept more precisely, and Claude added nuanced contrasts and conditional statements. The researchers note that these stylistic differences can influence how students think and write, suggesting that the choice of AI tool matters beyond mere correctness.

The study examined 90 academic‑style texts (30 per model) and highlighted that the way AI constructs arguments may shape learning outcomes, underscoring the need for educators to consider the specific reasoning patterns each system promotes.