Every wrong SQL query fired at a production database costs someone an hour of debugging they did not budget for.

Writing SQL from natural language has been a liability, not a feature

Business analysts and data engineers waste measurable time translating stakeholder questions into queries, then fixing the ones that run but return wrong results. The problem is not syntax, it is business context, dirty values, and domain-specific logic that older tools cannot interpret.

Google now ships a system that returns correct results, not just valid queries

Google Releases Gemini takes a natural language question as input and outputs what Google calls execution-ready SQL, meaning the query runs and the result matches the intended answer. It is built on Gemini 3.1 Pro and evaluated on execution accuracy, not surface-level syntax checks. That distinction is the entire point: a query can look right and still return garbage.

Data teams are the first to feel this gap close

• Data analysts who spend Friday afternoons rewriting dashboard queries that returned mismatched totals
• Database engineers who review natural language SQL generated by junior staff before it touches production
• Business intelligence leads who need non-technical stakeholders to self-serve accurate reports without filing tickets

The use case is narrow but the blast radius is wide, because bad SQL touches every reporting workflow downstream.

80% execution accuracy on a benchmark designed to fail most tools

BIRD contains 12,751 question-SQL pairs across 95 databases in 37 professional domains, with dirty data and required external knowledge grounding, conditions that exposed the limits of every prior benchmark leader including Google’s own earlier Gemini-SQL entry. If Gemini-SQL2 integrates into BigQuery Studio, AlloyDB AI, and Cloud SQL Studio as the announcement implies, the competitive pressure on standalone text-to-SQL vendors becomes significant fast.

What you can do with it today

• Query large databases using plain English questions without writing SQL
• Generate execution-ready SQL across multi-domain enterprise databases
• Replace manual query review cycles on business context-heavy reports
• Test natural language inputs against databases with inconsistent or dirty values

Gemini-SQL2 is a capability layer, not a standalone product, so access depends on which Google Cloud products receive the update first.

Pricing

Pricing not listed, check our directory.

The honest limitation here

Google has not confirmed which products will ship Gemini-SQL2, so there is no guaranteed access path yet for teams outside Google Cloud.

What else is competing for this workflow

Defog and Text2SQL.ai both target enterprise text-to-SQL with their own BIRD submissions, neither has posted above 79% on the single-model track. The gap is narrow enough that product integration, not benchmark rank, will likely decide adoption.

Text-to-SQL just became a serious product battleground

The tools that win here will be the ones embedded inside the databases teams already use, not the best standalone demos. We cover tools like this every Friday — subscribe here and we’ll send the best ones straight to you.