By Fintan Burke
Publication Date: 2025-12-29 07:36:00
From virtual assistants on our phones to chatbots on government websites, the large language models (LLMs) that power AI tools like ChatGPT are online almost everywhere.
However, there is increasing evidence that these LLMs judge dialect speakers harshly.
In 2024, researchers at the University of California, Berkeley tested ChatGPT’s responses to different varieties of English dialects from places like India, Ireland and Nigeria.
Compared to American or British English, dialect responses showed an increase in stereotypes (18% worse), derogatory content (25% worse), and condescending responses (15% worse).
Some models simply don’t understand dialects at all. In July 2025, an AI assistant deployed by Derby City Council had difficulty understanding a radio presenter’s Derbyshire dialect when she used words like mardy (complain) and duck (dear/dear) during a call she made on air to test the AI assistant.
Other dialect speakers have experienced much worse effects. As…
