Gemini's Language Divide: AI Search Exposes Disparities in Professional Identity

A recent experiment conducted by an individual on Google's Gemini AI search revealed stark differences in how it portrays professional identity across languages.

The test involved querying the same name, 'Vladimir Sinitsyn', in both Russian and English, with the latter including the keyword 'seo'.

Despite being the same person, the responses were vastly different. In Russian, Gemini provided a confident, well-structured biography, complete with six citations, real dates, and bullet points highlighting areas of expertise.

The response accurately linked to four sources, including articles written by the individual, but also contained two fabricated attributions. In contrast, the English response denied the existence of an SEO expert with that name, instead recommending three well-known Russian SEO experts as possible matches.

Analysis suggests that this disparity is due to RAG (Retrieval-Augmented Generation) behavior and poor cross-lingual entity resolution. The Russian-language corpus contains sufficient recent content connecting the individual's name to GEO topics, leading Gemini to attribute adjacent chunks of content to the identified entity.