Token
A token is the small piece of text an AI model reads and writes, a word or part of one. Models count length, limits and cost in tokens.
A token is the unit of text an AI model works with. It might be a whole word, part of a word, a number or a punctuation mark. When a model reads your page, it reads it as a string of tokens. When it writes an answer, it writes one token at a time.
OpenAI's help article on tokens gives a rough rule for English: a token is about four characters, or three quarters of a word. Other languages and other models split text differently.
Why a marketer might care
Most of the time you won't think about tokens at all. They come up in three places.
How much a model can read at once is measured in tokens, through its context window. A platform that has read ten long pages has to fit them all in, and it may only use part of each one. That's one reason a page that gets to the point early tends to do better.
The cost of running a model is counted in tokens, which is part of why platforms keep answers short and summarise what they read.
And unusual names get split up. A common word like "coffee" is likely to be a single token. A made-up brand name may be broken into several pieces. Models still handle these names, but a name that's spelled several ways across the web gives a model less to go on.
An example
For example, a model asked about Harrow & Pine reads a 4,000-word buying guide that mentions the brand once, near the end. If the platform keeps only the first part of each page it reads, the mention may never make it into the answer. A short paragraph near the top that names Harrow & Pine and says what it sells is much more likely to survive.