In AI, a token means a small chunk of text that the AI reads and writes, usually a short word or a piece of a longer word. AI tools count everything in tokens: how much you can paste in, how long the reply can be, and how usage is measured.
Think of LEGO. You see a castle, but the AI sees the individual bricks. A short word like “cat” is one brick. A long word like “unbelievable” may be built from a few bricks snapped together. Tokens are those bricks.
In a sentence: “The chatbot stopped reading my document halfway through because I had gone over its token limit.”
What the word means
The plain point: an AI does not read letters or whole words the way you do. It reads tokens, which sit somewhere in between.
Google’s own documentation for its Gemini models puts it simply: a token can be one character or a whole word, and long words get split.1
Outside AI, “token” has other meanings: a coupon, a coin for an arcade, a crypto asset, or a security code. This page is only about the AI meaning.
A simple example
The plain point: as a rough rule, a token is a bit less than a word in English.
- Google’s rule of thumb: about 4 characters per token, and 100 tokens is about 60 to 80 English words.1
- OpenAI’s rule of thumb: a token is roughly three-quarters of a word.2
These are estimates. Each company splits text its own way, so the same sentence can be a different number of tokens in different tools.
Pictures, sound and video are turned into tokens too. Google, for example, counts audio by the second.1
Why it matters to you
The plain point: tokens are the unit behind the limits you bump into.
- How much the AI can handle at once is measured in tokens. That limit is called the context window.1
- Developer billing depends on tokens, both what you send and what comes back. What you are actually burning explains where the tokens go; this page gives no prices.1
Shorter, focused requests use fewer tokens. Pasting a whole book to ask one question uses a lot.
Often confused with
- A word: close, but not the same. Many words are one token; long or unusual words are several.
- A crypto token: unrelated. That is a digital asset on a blockchain.
- A login token or API key: unrelated. That is a secret code that proves who you are. Never paste one into a chat.
Questions people ask
How many words is a token?
A bit less than one word, in English. OpenAI’s rule of thumb is that one token is about three-quarters of a word,2 and Google says 100 tokens is about 60 to 80 English words.1
What is a token limit?
It is the most text a model can handle at once, counting both what you send and what it writes back. Google calls this the context window, “the combined limit of input and output tokens”.1
Learn more on this site
- Common mix-ups: AI terms people get wrong, including tokens.
- The limit tokens fill: context window meaning.
- How costs add up: What you are actually burning and prompt caching.
- The basics: the AI glossary and AI explained for complete beginners.
- Related: How Does ChatGPT Work? What Happens When You Press Send.
- Related: Latency meaning.
- Related: Free Word and Token Counter for ChatGPT, Claude and Gemini.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is a token in AI?
- A secret code that proves who you are
- A digital coin kept on a blockchain
- A small chunk of text the AI reads and writes
Show the answer
Answer: A small chunk of text the AI reads and writes. The page says a token is usually a short word or a piece of a longer word; login codes and crypto tokens are unrelated meanings.
2. Roughly how much text is one token in English?
- A bit less than one word
- About one whole sentence
- Always exactly one letter
Show the answer
Answer: A bit less than one word. The page gives rules of thumb from Google and OpenAI: a token is a bit less than a word, and each company splits text its own way.
3. Why do shorter, focused requests help?
- They make the AI more truthful
- They use fewer tokens
- They switch off the token limit
Show the answer
Answer: They use fewer tokens. The page says limits are measured in tokens, so a focused request uses fewer than pasting a whole book to ask one question.
Related terms: Agent · AI model · Algorithm (in AI) · Natural language processing (NLP) · Neural network · Prompt
- 1Google, Understand and count tokens, updated 23 Sep 2026, read at source 1 Oct 2026“Tokens can be single characters like z or whole words like cat. Long words are broken up into several tokens.” ↑a“For Gemini models, a token is equivalent to about 4 characters. 100 tokens is equal to about 60-80 English words.” ↑b“32 tokens per second”. ↑c“The context window defines the combined limit of input and output tokens.” ↑d“the cost of a call to the Gemini API is determined in part by the number of input and output tokens”. ↑e“For Gemini models, a token is equivalent to about 4 characters. 100 tokens is equal to about 60-80 English words.” ↑f“The context window defines the combined limit of input and output tokens.” ↑g
- 2OpenAI, Understanding and counting tokens, read at source 22 Sep 2026“1 token is approximately three-quarters of a word.” ↑a“1 token is approximately three-quarters of a word.” ↑b
OpenAI, Understanding and counting tokens, read at source 22 Sep 2026 (as cited on AI terms people get wrong; the page refused an automated re-read on 1 Oct 2026).
The comparison on this page (LEGO bricks) is this site’s own way of explaining, not a claim from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK