The increasing complexity of food compositions and misleading marketing practices make it difficult for consumers to assess product quality based on ingredient lists, as proper interpretation requires specialized nutritional knowledge. This study compares the barcode app Mira with three LLMs (Qwen, DeepSeek, ChatGPT) evaluating nine products. The LLMs provided high, context-aware analysis and identified marketing discrepancies. Mira exhibited database limitations and template-based analysis. The findings confirm that LLM-based systems can effectively replace barcode apps, with practical applications like AI photo analysis or a Telegram bot for real-time evaluation.

Голикова В.А., Курманова В.А., Болдина Д.А. (науч. рук. Саид Л.И.) Comparative analysis of large language model-based systems and barcode-driven mobile application for ingredient-based food quality assessment and its health-related implications // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2026]. URL: https://kmu.itmo.ru/digests/article/17191