What is a token?

When people talk about ChatGPT, Claude, Gemini, or other AI tools, you sometimes hear about tokens.

So what are they?

What is it?

A token (in French you may see jeton) is a small piece of text the AI uses to read and generate content.

It is not always a full word.

A sentence like:

“My dog loves to run.”

will be split by the AI into several small pieces.

Those pieces are tokens.

Why split text at all?

Because an AI like ChatGPT does not handle a sentence exactly the way we do.

It works with these tokens to understand context and build its answer, piece by piece.

Why do we hear so much about tokens?

Because they measure how much text the AI has to process.

Your question uses tokens.

Documents you give it use tokens.

Its answer uses tokens too.

Tokens also drive cost

When you use an AI service directly — for example through an API — billing is often largely based on how many tokens you use.

There are mainly two kinds:

Input tokens
Everything you send to the AI: your question, instructions, documents, context, and so on.

Output tokens
Everything the AI generates in its reply.

For example:

You send 1,000 tokens to the AI.
It answers with 500 tokens.

Input and output tokens can have different prices depending on the model.

A simple analogy

Think of your mobile data plan.

You send and receive data.

With an AI, you send and receive tokens instead.

How much you use feeds into what the service costs.

Remember

A token is a small piece of text an AI uses to read and generate language.

Tokens measure how much information is processed and underpin billing for many AI services.

In very simple terms:

Tokens sent → 🤖 AI → Tokens received

The more the AI has to read and generate, the more tokens it uses.