A token is one of the small chunks of text an AI reads and generates — roughly a word, or sometimes a piece of a word. AI models break all text into tokens to process it, and usage, limits (including the context window) and, on paid plans, cost are usually measured in tokens rather than words or characters.
You don't need to think in tokens to use AI well, but the word comes up often enough that it's worth knowing. When an AI model processes text, it doesn't work with whole words the way you might expect — it breaks everything down into smaller units called tokens, which are typically a bit shorter than an average English word.
A short, common word might be one token; a longer or less common word might be split into two or three. Punctuation and spacing count too. As a rough rule of thumb, a page of everyday English text works out to a few hundred tokens — enough to give a rough sense of scale without needing to be precise.
Tokens matter practically in two places: they're what fills up a model's context window (its working memory limit), and on paid AI plans, usage is frequently billed per token, so longer documents and longer conversations use more of your allowance.
ExampleA short customer email might be a few hundred tokens. A long report or contract could run into many thousands — which is part of why very long documents can bump into a model's context window limit, or add up faster on a metered plan.
If you're on a metered AI plan and want to manage cost, keep an eye on how much text you're pasting in and how long your conversations run — both add up in tokens, even if you never see the number directly.
See also
Learn to use this on your own work
Our AI Training & Enablement program takes non-technical Australian teams from the terminology to real, working AI builds on the job they already do.
Book a call →