Your Brain on ChatGPT: Is Bilingualism the Solution?
Koko Kusano
17/02/2026
The widespread adoption of Artificial Intelligence (AI) Large Language Models (LLMs) such as ChatGPT has raised concerns about their cognitive impact. Research suggests that reliance on LLMs may impair not only users’ perceived but also palpable ownership over their thinking, surpassing cognitive-offloading effects previously observed with other technologies. Recent studies found diminished cognitive activity in individuals using ChatGPT to write essays, compared to those using Google Search or internal cognition. These findings point to potential declines in executive functions inclusive of but not limited to working memory, attention, and decision-making.
In contrast, bilingualism has been linked to cognitive resilience. According to Abutalebi and Green (2007), bilinguals routinely manage dual linguistic systems, strengthening executive functions that extend beyond linguistic into non-linguistic domains.
Therefore, this study investigated how language background influences LLM engagement and cognitive performance. Focusing on individuals aged 18-25, it examined differences among trilinguals, bilinguals and monolinguals in their use of LLMs and their performance on either cognitive task testing executive functioning: the Simon or AX Continuous Performance Task (AX-CPT). Trilinguals were included to examine the scalability of the bilingual advantage to those managing three or more languages. The study found a hierarchy in LLM engagement, with trilinguals and bilinguals reporting higher levels of reliance, frequency of use, and a greater number of cognitive purposes (e.g. attention, cognitive flexibility) offloaded onto LLMs than monolinguals. LLM use was negatively correlated with performance on cognitive tasks. Bilinguals generally performed better than monolinguals on these tasks. However, the subsequent impact with language background remained statistically inconclusive. The findings did not support a definitive protective effect of bilingualism against LLM-induced cognitive decline.