📎 This article includes 1 downloadable practice file ↓
Chat AI tools are large language models: they predict the next piece of text, one token at a time, based on everything in the conversation so far. Understanding that explains most “why did it do that?” moments.
What the model sees
- Tokens: text is split into pieces (roughly ¾ of a word each). Limits and costs are counted in tokens.
- Context window: the conversation it can “see” at once, from a few thousand to hundreds of thousands of tokens. Older parts of very long chats can fall out.
- Training data: knowledge up to a cut-off date. Newer facts only come from tools like web search or files you provide.
Why wording matters
The model continues the most likely text given your prompt. A vague prompt (“Explain GST”) gets a generic answer; a specific one (who it’s for, length, format, example) gets a specific answer. You are setting up the situation it continues from.
Why answers vary
Most tools sample with some randomness, so the same prompt can give different answers. Regenerate a few times for creative work; for facts, ask for sources and check them.
Practice
Work through the four experiments in the worksheet and note the differences.
📎 Practice files for this article
- 📄Lesson 1 prompt worksheetCopy-paste prompts and exercises to try in ChatGPT, Copilot, Gemini or Claude.⬇ TXT · 366 B
Free to use for learning. Files with macros (.bas) are plain text — import them with Alt+F11 → File → Import File, and always test on a copy.
Stuck on a step? Ask a question and the AI answers using this article.