How to Use ChatGPT to Write SQL Queries

SQL is one of the things ChatGPT is genuinely good at. Queries follow clear rules, there is a mountain of public SQL to learn from, and the answer is easy to check by running it. So rather than describe how it does, I gave it real schemas, asked the kind of things I actually ask, and judged the SQL as a developer. This was ChatGPT on the free tier, August 2026.

Writing a Query from a Description

Give it the tables, then ask in plain language. Here is the prompt, and the query it wrote.

I have two tables: orders(id, customer_id, total, created_at) and
customers(id, name, country). Write a SQL query that returns the top 5
customers by total spend in 2025, showing name, country and total spent,
ordered by total spent descending.

SELECT
    c.name,
    c.country,
    SUM(o.total) AS total_spent
FROM customers c
JOIN orders o ON o.customer_id = c.id
WHERE o.created_at >= '2025-01-01'
  AND o.created_at <  '2026-01-01'
GROUP BY c.id, c.name, c.country
ORDER BY total_spent DESC
LIMIT 5;

It was correct and readable, and it did one careful thing I had not asked for. Instead of filtering the year with something brittle, it used a half-open date range, from the first of January up to but not including the next January.

WHERE o.created_at >= '2025-01-01' AND o.created_at < '2026-01-01'

That quietly avoids the classic bug where an order timestamped late on the 31st of December falls outside a naive range. It is the instinct of someone who has been bitten before, and it is the kind of thing that makes ChatGPT a useful pair for SQL rather than just an autocomplete.

When Your Request Is Vague

Then I under-specified on purpose, to see what it would do with an ambiguous ask.

Using those same two tables, give me my best customers.

It did not stop to ask me what "best" meant. It picked a definition and, to its credit, said so out loud before writing a line of SQL.

That is the pattern to watch. ChatGPT will rarely pause to ask a clarifying question, it will assume and move on. Here the assumption was reasonable, but if "best" had meant most orders, most recent, or highest margin, I would have got a confident, correct-looking answer to the wrong question. Either spell out what you mean, or read its stated assumption before you trust the result.

Switching Between Databases

SQL varies between databases, and this is where it quietly saves time. I asked it to move a MySQL query to PostgreSQL and flag the differences.

Rewrite this so it works in PostgreSQL instead of MySQL, and point out any
differences I should know about:

SELECT id, DATE_FORMAT(created_at, '%Y-%m') AS month
FROM orders WHERE total > 100 LIMIT 10;

SELECT id, TO_CHAR(created_at, 'YYYY-MM') AS month
FROM orders
WHERE total > 100
LIMIT 10;

It correctly swapped MySQL’s DATE_FORMAT for Postgres’s TO_CHAR, translated the format codes from %Y-%m to YYYY-MM, and confirmed that LIMIT, the WHERE clause and the alias all work unchanged. For anyone moving between MySQL and Postgres, that is a genuinely useful few seconds, and it explains the change rather than just handing you a new query to paste blind.

For turning query results into insight, see how to use ChatGPT for data analysis. For getting sharper answers out of any prompt, see how to write better AI prompts.