Best AI Prompts for SQL Query Generation with ChatGPT

Every data analyst has spent hours staring at a blank query editor, knowing what they want to find but struggling to translate the business question into the right SQL syntax. ChatGPT bridges that gap. With the right prompting approach, it can take a description of what you want to analyze and produce the query you need.

This guide covers the specific prompting strategies that make ChatGPT genuinely useful for SQL query generation, from simple SELECT statements to complex window functions and optimization.

TL;DR

• ChatGPT accelerates SQL query writing by translating business questions into syntax
• Providing your database schema dramatically improves query accuracy
• The most effective prompts describe the desired outcome rather than the technical implementation
• Always validate AI-generated queries against your schema before running them
• Complex queries (CTEs, window functions) benefit most from AI assistance
• Building a query prompt library for common analysis types speeds up recurring work
• ChatGPT cannot see your data; it can only work from your descriptions

Introduction

ChatGPT is a powerful SQL assistant when used correctly. It knows SQL syntax across all major dialects. It can reason about data relationships. It can explain query logic in plain language. But it has a critical limitation: it cannot see your database schema or your actual data.

This means the quality of your prompts determines the quality of your queries. Vague prompts produce vague or incorrect queries. Specific prompts with schema context produce production-ready SQL.

This guide teaches you how to write prompts that get accurate, efficient, and useful SQL from ChatGPT.

Table of Contents

1. Why Use ChatGPT for SQL?
2. Providing Schema Context
3. Basic Query Prompts
4. Intermediate Query Prompts
5. Complex Query Prompts
6. Query Review and Optimization Prompts
7. Building a SQL Prompt Library
8. FAQ

Why Use ChatGPT for SQL?

ChatGPT provides several distinct advantages for SQL query writing:

Syntax translation: ChatGPT handles the mechanics of SQL syntax, letting you focus on the analytical logic. If you know what you want to find but struggle with the JOIN syntax or window function mechanics, ChatGPT fills that gap.

Query explanation: ChatGPT can explain what a query does in plain language, which helps when reviewing queries from others or trying to understand a complex legacy query.

Debugging assistance: When a query returns an error, ChatGPT can often identify the issue from the error message and suggest corrections.

Learning aid: By generating queries and asking for explanations, you can improve your own SQL skills over time.

The key is knowing how to describe what you want in a way that produces accurate SQL.

Providing Schema Context

The most important factor in query accuracy is the schema context you provide. The more complete your schema description, the better ChatGPT’s output.

Schema Description Prompt

I need to write SQL queries against the following database schema.
Please confirm your understanding of the structure:

Tables:
[TABLE NAME 1]
- Column A (TYPE, description)
- Column B (TYPE, description)
- Primary key: [COLUMN]
- Foreign keys: [COLUMN] references [TABLE.COLUMN]

[TABLE NAME 2]
[REPEAT FOR ALL TABLES]

Please confirm:
1. I have correctly identified the primary and foreign keys
2. Any assumptions you are making that might not be correct
3. Whether [TABLE/COLUMN NAMES] naming suggests any conventions
   I should be aware of

Schema Summary Prompt

Provide a concise schema summary for pasting into SQL query prompts.
Focus on the tables and columns most relevant for [TYPE OF ANALYSIS].

Include:
- Table name and purpose
- Key columns (primary key, foreign keys, commonly filtered columns)
- Important columns for [ANALYSIS TYPE]
- Any relationships between tables that matter for [ANALYSIS TYPE]

Format as a clean reference table for quick prompt inclusion.

Basic Query Prompts

Start with simple queries and build complexity as you validate accuracy.

Simple SELECT Prompt

Write a SQL query that retrieves [WHAT YOU WANT] from [TABLE(S)].

The database is [DIALECT - E.G., PostgreSQL, MySQL, SQL Server].

Here is the relevant schema:
[SCHEMA]

Filters:
- Include only [CONDITION]
- Date range: [IF APPLICABLE]
- Sort by: [COLUMN AND DIRECTION]

Requirements:
- Only return columns: [LIST]
- Limit to [NUMBER] results
- Exclude [ANY EXCLUSIONS]

WHERE Clause Generation Prompt

I have a table: [TABLE NAME]
Columns: [COLUMN LIST]

Write a WHERE clause that filters for:
1. [CONDITION 1]
2. [CONDITION 2]
3. [CONDITION 3]

Combine these conditions with AND/OR as appropriate for [DESIRED LOGIC].

Also provide:
- Alternative phrasings of the conditions if my description
  might be ambiguous
- Any potential issues with my condition logic that I should verify

Intermediate Query Prompts

Once basic queries work well, use these prompts for JOINs and aggregations.

JOIN Query Prompt

Write a SQL query that joins [TABLE A] with [TABLE B] to get [RESULT].

Schema:
[TABLE A SCHEMA]
[TABLE B SCHEMA]

Join logic:
- Connect on: [JOIN CONDITION]
- Join type: [INNER/LEFT/RIGHT/FULL]

Aggregation needed:
[SUM/COUNT/AVG/etc.] of [COLUMN] grouped by [DIMENSION]

Filters: [CONDITIONS]
Sort: [ORDER]

Aggregation Query Prompt

I need to calculate [METRIC] for [DIMENSION].

Database: [DIALECT]
Tables: [TABLES]

Please write a query that:
1. Calculates [METRIC - E.G., total revenue, average order value]
2. Groups by [GROUPING COLUMN - E.G., customer segment, month]
3. Filters to [DATE RANGE OR OTHER FILTER]
4. Includes only [ANY OTHER CONDITIONS]

Also include:
- A version using window functions if that would be more efficient
- A version with rollup for subtotals if applicable
- An explanation of which version is better for which use case

Complex Query Prompts

Complex queries with CTEs, window functions, and subqueries are where ChatGPT provides the most value.

Window Function Query Prompt

Write a SQL query using window functions to [DESIRED OUTCOME].

Database: [DIALECT]
Tables: [TABLES AND SCHEMA]

Window functions needed:
1. [FUNCTION TYPE - E.G., RANK, LAG, LEAD, running total]
   over [PARTITION BY] order by [ORDER BY]
   for [PURPOSE]

2. [FUNCTION TYPE] over [PARTITION BY] order by [ORDER BY]
   for [PURPOSE]

Full requirements:
[DETAILED DESCRIPTION OF DESIRED OUTPUT]

Also explain:
- What the window function is doing at each step
- Why this approach vs. alternatives (e.g., self-join vs. window function)
- Any performance considerations

CTE Query Prompt

Write a SQL query using CTEs (WITH clauses) to [DESIRED OUTCOME].

Database: [DIALECT]
Tables: [TABLES AND SCHEMA]

CTE structure (conceptual):
1. [CTE NAME]: [WHAT IT CALCULATES]
2. [CTE NAME]: [WHAT IT CALCULATES, referencing prior CTE(s)]
3. Final SELECT: [HOW THE CTEs COMBINE TO PRODUCE THE RESULT]

Final output: [DESCRIPTION]
Filters: [CONDITIONS]

Cohort Analysis Prompt

Write a SQL query to perform a cohort analysis for [WHAT YOU ARE
ANALYZING - E.G., user retention, subscription revenue].

Database: [DIALECT]
Tables: [RELEVANT TABLES AND SCHEMA]

Cohort definition: [HOW TO DEFINE A COHORT - E.G., first purchase month]
Cohort period: [TIME GRANULARITY - monthly, weekly]
Metric to track: [WHAT YOU ARE MEASURING - E.G., retention rate, revenue]

Please provide:
1. The CTE structure defining cohorts
2. The calculation of the cohort metric
3. A pivot or matrix format showing cohort evolution over time
4. An explanation of what each CTE step contributes

Query Review and Optimization Prompts

ChatGPT can review and optimize existing queries.

Query Review Prompt

Please review the following SQL query for correctness, efficiency,
and best practices:

[QUERY]

Database: [DIALECT]
Table sizes (approximate): [IF KNOWN - HELPFUL FOR OPTIMIZATION]

Please identify:
1. Any syntax errors or logic bugs
2. Potential performance issues (missing indexes, inefficient joins)
3. Deviations from SQL best practices
4. Ways to improve readability
5. Alternative approaches that might be more efficient

Provide the corrected/optimized version with notes on each change.

Query Optimization Prompt

I have a SQL query that currently runs slowly. Here is the query:

[QUERY]

Database: [DIALECT]
The [SLOW CLAUSE - E.G., WHERE, JOIN, GROUP BY] takes [APPROXIMATE TIME]
when it should take [EXPECTED TIME].

Please suggest:
1. Specific optimizations for this query structure
2. Index recommendations (note: I will verify these against my schema)
3. Alternative approaches if the current approach is fundamentally inefficient
4. How to verify the optimization actually helped

Provide the optimized query with explanations.

Building a SQL Prompt Library

The fastest way to use ChatGPT for SQL is building a reusable library of prompts for your most common analysis types.

Prompt Library Entry Format

ANALYSIS TYPE: [NAME]
Common database: [YOUR MAIN DB DIALECT]
Frequency: [HOW OFTEN YOU RUN THIS TYPE OF QUERY]

Standard prompt template:
[YOUR REUSABLE PROMPT WITH [VARIABLE] PLACEHOLDERS]

Schema elements typically needed:
- Tables: [LIST]
- Common filters: [LIST]
- Standard aggregations: [LIST]

Validation checklist:
- Verify table names against actual schema
- Check column names are exact matches
- Confirm date/time handling matches your DB's requirements

FAQ

What SQL dialects does ChatGPT support? ChatGPT has training data covering most major SQL dialects including PostgreSQL, MySQL, SQL Server, Oracle, SQLite, and BigQuery. When prompting, specify your dialect for the most accurate syntax.

How do I handle queries that need data ChatGPT cannot see? Include sample values or value ranges in your prompt. If a column contains status values, list them. If date ranges matter, specify them. The more data context you provide, the more accurate the query logic.

What if my schema is too large to include in a prompt? Focus on the relevant tables for your specific query. You do not need to include the entire schema, just the tables and columns involved in the query. Use the schema summary prompt to create a concise reference.

Can ChatGPT write queries that join my data with third-party data? No. ChatGPT can only work with the schemas and structures you describe. Any third-party data must be incorporated through descriptions of its structure, not through direct access.

How do I validate AI-generated SQL before running it in production? Always run AI-generated queries in a development or staging environment first. Review the output against your expectations. If the results seem wrong, provide the incorrect output to ChatGPT and ask for debugging help. Never run untested queries directly in production.

Can ChatGPT help with SQL injection prevention? Yes. When you share query logic with ChatGPT, ask it to identify potential injection vulnerabilities and suggest parameterized query alternatives. This is a valuable use of AI review even if you did not use AI to generate the original query.

Conclusion

ChatGPT is a capable SQL assistant when you provide adequate schema context and clear outcome descriptions. The prompts in this guide range from basic SELECT statements to complex window function queries, giving you a complete framework for SQL query generation.

Build your analysis-specific prompt library, validate queries against your actual schema, and always test in a development environment before production deployment. Used this way, ChatGPT meaningfully accelerates the data analysis workflow.

Your next step: Pick one recurring analysis you do weekly. Use the prompt templates above to draft the query, include your schema context, and generate the SQL. Validate it against your database and save the working prompt in your library.