What Is a Token in LLM Tools? A Plain-English Definition
A token is a unit of text, such as a word fragment, that a language model processes. Models break text into tokens before they read it. The number of tokens decides how much a model can consider at once, and many AI providers measure usage in tokens.
A token is a unit of text, such as a word fragment, that a large language model processes. The model does not read your message as whole sentences or whole words. It reads a sequence of tokens.
For a business owner, tokens matter for two practical reasons. Tokens set how much text a model can handle in one go, and many AI providers measure usage in tokens. If you use AI tools in your operations, tokens are the unit behind both the limits you hit and the usage you pay for.
What a Token Means
A token is the smallest piece of text a language model works with. A token can be a whole word, part of a word, or a fragment of text. The model turns your input into tokens, works on those tokens, and produces its answer as tokens.
The process of breaking text into tokens is called tokenization. Tokenization happens automatically every time you send text to a model. You never see the tokens directly in most tools, but every prompt, document and reply passes through this step.
How Tokens Work in Practice for SMEs
Tokens affect a small or mid-sized business in two places: limits and usage. Each model has a context window, which is the maximum number of tokens the model can consider at once. Your instructions, the documents you paste in, the conversation so far and the model's reply all share that window.
When the text goes past the context window, the model cannot consider all of it at once. In practice this means a very long contract, a full year of chat history or a large product catalogue may need to be split into parts or summarised first.
Usage is the second place tokens show up. Many AI providers measure usage in tokens, so longer prompts and longer answers use more of your allowance. A workflow that pastes the same long instructions into every request uses more tokens than a workflow that keeps instructions short and sends only the text the model needs.
An Everyday Example
Picture a clinic or a property management office that uses an AI assistant to draft replies to customer emails. Each request sends the assistant three things: a set of standing instructions, the customer's email, and maybe a few past messages for context. The assistant then writes a reply.
Every part of that request is tokenized. The instructions, the email, the past messages and the drafted reply all count as tokens. If the office pastes in an entire email thread going back months, the request uses more tokens and may come close to the context window. If the office sends only the latest message and a short summary of the history, the request stays smaller and the model has room to focus on what matters.
When Thinking in Tokens Is Not the Right Tool
Counting tokens is not useful for every decision. If your team sends a few short questions to an AI tool each day, tokens will rarely be your bottleneck. Focus on whether the answers are accurate and useful before you optimise length.
Tokens also do not measure quality. A short prompt is not automatically better, and a long document is not automatically worse. The right question is whether the model has the information it needs to do the task. Token awareness helps when you run high volumes, process long documents or build automated workflows, because that is where limits and usage add up.
Frequently Asked Questions
What is a token in LLM terms?
A token is a unit of text, such as a word fragment, that a language model processes. The model breaks your text into tokens before working on it, and the model writes its answer as tokens too.
What is tokenization?
Tokenization is the process of breaking text into tokens. Tokenization happens automatically whenever you send text to a language model, so you do not need to do it yourself.
What is a context window and how does it relate to tokens?
A context window is the maximum number of tokens a model can consider at once. Your instructions, any documents you include, the conversation history and the reply all count toward that limit. Long material may need to be split or summarised to fit.
Why do AI providers talk about tokens when describing usage?
Many AI providers measure usage in tokens. More text in your prompts and in the model's replies means more tokens used, so keeping requests focused helps you manage usage.