Database World Trying To Build Natural Language Query Systems Again - This Time With LLMs
The Register, Wednesday, April 22nd, 2026
Database vendors are using LLMs to build text-to-SQL systems that translate natural language into database queries, but accuracy limitations and semantic risks require caution.
Database and analytics vendors are leveraging large language models to create natural language query systems that convert business questions into SQL without requiring users to know the language. While these text-to-SQL tools can help database developers and analysts save time, experts warn that current systems achieve only about 80% accuracy and risk generating syntactically correct but semantically wrong queries that could produce misleading results for non-technical users.