# Synthetic Data

> Data made by a computer to train a model. It is not taken from real life.

**Field:** Artificial Intelligence  
**Category:** LLMs  
**Also known as:** AI-Generated Training Data, Generated Data

## Think of it like

Think of it like you make up quiz cards for a friend. You check each card first.

## Definition

Synthetic data is training data made by a computer, not collected from the real world. For language models, it is often text written by another model: questions, answers, worked steps or code. It is cheap and can fill gaps where real data is rare. But it must be checked. A model trained on its own mistakes can get worse over time.

## Example

Jack needs 10,000 maths word problems with worked answers. A model writes 30,000, a checker keeps the 12,000 whose answers are right, and those go into training.

## Why it matters

Good real data is running short for the biggest models. Made data is how labs keep feeding them.

## Common mistake

Using it unchecked. Errors and odd habits in the made data get baked into the new model.

## Three principles

- **Made, not collected.** Often by another model.
- **Fills gaps.** Where real data is rare.
- **Must be checked.** Or mistakes compound.

## Related terms

- [Training Data](https://learn.benibix.com/term/ai/training-data)
- [Knowledge Distillation](https://learn.benibix.com/term/ai/knowledge-distillation)
- [Fine-Tuning](https://learn.benibix.com/term/ai/fine-tuning)
- [Data Augmentation](https://learn.benibix.com/term/ai/data-augmentation)
- [Model Collapse](https://learn.benibix.com/term/ai/model-collapse)

## Told apart from

- [Data Augmentation vs Synthetic Data](https://learn.benibix.com/compare/data-augmentation-vs-synthetic-data): People think data augmentation makes synthetic data, and it does not. Augmentation takes a real example and changes it a bit, like a flipped photo. Synthetic data is made from scratch, often by a model or a program. One stretches the data you have. The other makes new data you never collected.

## Taught in

- [Knowledge Distillation](https://learn.benibix.com/learn/ai-llms/knowledge-distillation): LLMs, 10/15

![A model writes many new examples. A filter throws out the bad ones. The ones that pass join the training set.](https://learn.benibix.com/diagrams/llm-synthetic-data.svg)

---

Source: https://learn.benibix.com/term/ai/synthetic-data
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
