What is synthetic data?
Synthetic data is generated by a computer instead of every example being measured directly in the real world.
What is it?
Synthetic data are generated by a computer instead of each example being measured in the real world. They are used when real data are scarce, expensive or sensitive. They are only as good as their likeness to reality.
An everyday example
A hospital wants to test a new program but may not share real records. So it generates artificial records with the same statistical properties and no real patients.
Why does it matter?
Synthetic data are increasingly used to train models. They can fill gaps, but they can also amplify the mistakes of the model that produced them.
A common misconception
That synthetic data are automatically anonymous and safe. Sometimes information about real people can still be inferred from them.
In more detail
Why is it used?
For simulations, developing models, adding extra examples and testing rare situations.
Is it always anonymous?
No. Synthetic data can imitate real records or reveal patterns. Privacy has to be checked separately.
What is the danger?
Unrealistic examples can teach a model the wrong things and hide real-world problems.
How is it evaluated?
By comparing its properties, its usefulness for the task, its variety and the risk that it reveals real data.
Remember: “Synthetic” does not necessarily mean “safe” or “real”.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.