AI can dramatically accelerate database performance tuning. LLM query optimization workflow: 1) Paste slow query + table schemas + EXPLAIN output to Claude/GPT-4o. 2) Ask: "Analyze this query execution plan and suggest optimizations. Consider: missing indexes, N+1 queries, unnecessary joins, subquery optimization." 3) AI identifies: full table scans, index usage, join order issues. Specific optimizations: "Add index on (user_id, created_at) to support this range query", "Replace correlated subquery with LEFT JOIN", "Use window function instead of GROUP BY subquery." Index analysis prompt: "Given this query and table statistics (rows, distinct values, null counts per column), recommend optimal indexes. Explain trade-offs between query performance and write overhead." EXPLAIN plan interpretation: modern databases (Postgres, MySQL) provide verbose execution plans. AI can explain in plain language: "This query is doing a sequential scan of 5 million rows because the index on user_id is not being used. This is likely because the WHERE clause includes a function call (LOWER(email)) which prevents index usage." Query rewriting examples: replace SELECT * with specific columns, add LIMIT for paginated queries, use CTEs for readability without performance loss, materialized views for complex aggregations. Caveats: always test AI suggestions in staging, measure before and after, understand why the optimization works.