A token is the unit an AI model reads and writes in. A token is a word or part of a word, roughly three quarters of an English word on average. Usage limits, context windows and API prices are all counted in tokens rather than words or pages.
AItopic
2 minread
1related terms
In one line
The unit an AI model reads and writes in. A token is a word or part of a word, roughly three quarters of an English word on average. Usage limits and API prices are counted in tokens.
Models do not read letters or whole words. They break text into chunks called tokens. Common short words are often one token, while long or rare words split into several. OpenAI's own guidance gives a rule of thumb of about four characters, or three quarters of a word, per token in English, so 1,000 tokens is roughly 750 words. Other languages, code and odd formatting usually use more tokens per word.
Tokens matter in two places. First, every model has a context window, a maximum number of tokens it can hold at once, including your prompt, any pasted files and its own reply. Second, paid API use and many plan limits are counted in tokens, with output tokens usually priced higher than input tokens. Long resumes and full job descriptions add up fast.
Why it matters to recruiters
When you paste a stack of resumes or a long call transcript into an AI tool, you are spending tokens. Knowing that helps you understand why a tool cuts off, forgets earlier parts of a chat or hits a usage cap mid afternoon. It also helps you judge vendor pricing for AI screening or note taking tools that bill by usage.
An example from a recruiting desk
Forty resumes go into one chat, and a sourcer asks for a ranked shortlist. The answer ignores the first dozen. The resumes ran past the useful working space of the model. She splits them into four batches of ten, asks for a short summary of each batch, then compares the summaries. The ranking now covers every candidate.
Tips that help
Strip headers, footers and boilerplate from documents before you paste them.
Ask for short, structured output when you only need a shortlist or a table.
Start a fresh chat for a new task instead of continuing a very long one.
Use a free tokenizer tool to check how big a long document really is.
Mistakes to avoid
Treating tokens and words as the same thing when estimating cost or limits.
Assuming the model read every page of a huge upload just because it accepted the file.
Frequently asked questions
How many words are in 1,000 tokens?
In ordinary English, about 750 words, using the common rule of thumb of three quarters of a word per token. The real number shifts with the text: names, technical terms, other languages and tables use more tokens per word.
Do input and output tokens cost the same?
Usually not. Most AI APIs price output tokens higher than input tokens, and prices differ by model. Chat subscriptions hide this behind message or usage limits, but the limits are still driven by how many tokens you use.
Why does the AI forget what I said earlier?
Each chat has a context window measured in tokens. Once a conversation runs past it, older parts fall out or get summarized. Long pasted documents fill the window faster, so restate key facts or start a new chat for a new task.
