Student reading, from the activity pack
How Does AI Write an Answer?
When you type a question into a chatbot, it seems like the AI is thinking about your question. What is really happening is closer to your phone keyboard suggesting the next word. The chatbot is powered by a large language model (LLM), an AI trained on huge amounts of text so it can predict what words are likely to come next.
Everything starts with data, the information an AI learns from. The text you type is the input, and the answer the AI gives back is the output. A computer cannot read words the way you do, so first it uses tokenization: it breaks the input into small pieces called tokens. A token can be a whole word, part of a word, or a punctuation mark. For example, "unbelievable" might be split into "un", "believ", and "able".
Next, each token goes through embedding, which turns it into a list of numbers. Words that are used in similar ways get similar numbers, so "dog" and "puppy" end up close together. Now the model can do math with language. Using patterns it learned in training, it makes a prediction about which token should come next, adds it, and repeats, one token at a time, until the answer is finished. It is not looking facts up. It is choosing what usually comes next.
To learn those patterns, models need enormous amounts of text. A lot of it comes from web scraping: using automatic programs to copy information from websites, such as articles, forums, and posts. More data usually makes a model better at sounding natural, so AI companies have a strong reason to collect as much as they can. This raises questions. Did the people who wrote that text agree to have it used? Was private or copyrighted material included?
Because an LLM predicts likely words instead of checking truth, it can produce answers that sound confident and polished but are wrong. A confident tone is not proof. The model also only knows what was in its data. If most of the text came from one language, one country, or one point of view, the model may be weaker or unfair when it deals with other languages, cultures, or communities. It can also repeat the stereotypes and opinions found in human writing.
Knowing how AI writes helps you use it wisely: treat its answers as a starting point, double-check important facts, and notice whose voices might be missing from its data.