Attention
Attention is the part of an AI language model that works out which words in a text matter most to each other while it reads and writes.
Attention is the part of an AI language model that decides which words to focus on. As the model reads a text and writes its reply, attention works out, for each word, which other words in the text matter most to it.
It's the idea behind the transformer, the design that most of today's language models use, including the ones behind ChatGPT, Gemini and Claude.
How it works, roughly
Take the sentence: "Fernhill sent the beans late, but they were still fresh." To understand the word they, the model needs to link it to the beans and not to Fernhill. Attention is how it makes that link. It weighs every word against every other word and gives more weight to the ones that matter for the meaning.
The same thing happens across longer texts. When a model reads a web page to answer a question, attention helps it pick out the sentences that relate to the question.
Why a marketer might care
You don't need to understand the maths. One finding is worth knowing, though. Researchers found that models given a long text tend to use information at the start and end more reliably than information buried in the middle. The paper is called Lost in the Middle.
For example, if the one sentence that says Harrow & Pine delivers to Scotland sits halfway down a long page, a model may miss it. Put the key facts early and state them plainly, and they're more likely to be used.