Plain definition: A token is a small chunk of text—roughly three to four characters or about three-quarters of a word—that an AI model reads and generates as its basic unit of processing. Tokens are also how AI providers measure and charge for usage.

In plain terms

Think of tokens like postage. Every word you send an AI and every word it sends back costs a little postage. The longer the conversation or document, the more tokens used and (if you’re on a paid plan) the higher the cost. A short email might be 100 tokens; a long contract could be 5,000.

Why it matters for operators

Tokens affect two things: cost and limits. If you’re using the API directly or building an AI tool for your business, token count determines your bill. It also determines how much text the AI can handle at once—there’s a ceiling called the context window. Knowing this helps you avoid wasting money sending the AI information it doesn’t need.

Example

A law firm using an AI tool to summarize depositions learns they can cut their monthly AI bill by 30% simply by trimming the boilerplate header from every document before sending it—saving thousands of tokens per run.