Glossary
These are the terms introduced in Your First AI Skill, the free starter module, in plain language. Lesson numbers show where each term first appears. Later modules will add more words. Dated: November 2026. Product names and settings change; lessons link to each company's current pages.
AI and chatbot basics
AI chatbot (or chatbot). A program you use by typing or speaking a message and reading a reply. It isn't a person. Popular examples in November 2026 include ChatGPT, Claude, Gemini, and Grok. (Lesson 1)
Artificial intelligence (AI). In everyday talk, computer systems that can do things that used to take human intelligence, such as writing, summarizing, or answering questions in ordinary language. (Lesson 1)
Fluency. How smooth and natural writing sounds. Fluency isn't the same as truth. (Lessons 1 and 8)
Hallucination. A chatbot reply that isn't factually accurate, even if it sounds plausible and confident. Examples: wrong facts or dates; made-up quotes, studies, citations, or sources. (Lessons 1 and 8)
Language model (also large language model, or model). The program underneath a chatbot, trained to work with language. "Large" only means it was built at a very large scale. In this course, model is the short word for it. (Lesson 1)
Output / response. The chatbot's reply to your prompt. (Lesson 1)
Search engine. A tool built to find pages and show you links. A chatbot is built to write a reply. Some chatbots can search the web while they work, but they still compose the answer. (Lesson 1)
Token. A word or a piece of a word. A language model writes a reply by predicting one token at a time. (Lesson 1)
Training (of a model). Teaching a language model by having it practice, on a huge amount of text, the job of guessing what comes next. (Lesson 1)
Prompting
Prompt. The message you send to a chatbot. (Lesson 1)
Prompting. The skill of telling an AI tool what you need clearly enough that it can actually help. Some companies call the more technical version prompt engineering. (Lesson 1)
Clear ask. A request that says what you want done and what you want back, for example "List eight ideas, one line each." Prompting skill P1. (Lesson 1)
Context. The background a chatbot needs so it doesn't have to guess: the situation, the reader, the facts, and your limits. Prompting skill P2. (Lesson 2)
Reader. The person the output is for: a sister, a landlord, a teacher, you. Naming the reader changes the words. (Lesson 2)
Tone. The manner of the writing, for example warm, formal, brief, or straight to the point. (Lesson 2)
Constraint (or limit). A limit you set in a prompt, such as a length, a format, something to include, or something to avoid. Prompting skill P4, "shape and limits." (Lesson 2)
Draft. A starting version, not the finished product. Treat a chatbot's first reply as a draft. (Lessons 2 and 5)
Follow-up. A second (or later) prompt in the same chat, used to revise or add to the previous reply. (Lesson 5)
Iterate (iteration). Improving something in steps, using what you already have, instead of accepting the first draft or starting over. Prompting skill P5. (Lesson 5)
Ask for a check. Adding a request such as "say if you're not sure," "name the source I should check," or "list the claims I should verify," so the reply is easier to check. It doesn't replace checking outside the chat. Prompting skill P6. (Lesson 8)
Safe prompting. Giving the chatbot the situation it needs while leaving out private details, by describing the kind of information and using placeholders. Prompting skill P7. (Lesson 9)
Placeholder. A clearly marked stand-in in a draft, such as [ACCOUNT NUMBER], that you fill in later yourself through a trusted channel, never a real secret pasted into the chatbot. (Lesson 9)
The loop. The four-part habit this module teaches: ask clearly; improve the answer; check anything that matters; protect privacy, and keep the decision yours. (Lesson 10)
Checking and privacy
Reliable source. A source you can reasonably trust for a particular claim, often the organization that publishes the rule or figure, such as a government agency's official page. The chatbot itself isn't the source. (Lesson 8)
Verify (check, fact-check). To confirm a claim against an independent, reliable source before you repeat it or act on it. (Lesson 8)
Sensitive information (private details). Information you shouldn't paste into a personal chatbot: passwords, government ID numbers, bank or card numbers, detailed health records, other people's private information, and confidential work material. Lesson 9 has the full "don't paste" list. (Lesson 9)
Model training (on your chats). Using conversations, and sometimes other inputs, to help improve future models. On many consumer chatbots this depends on a privacy setting you can change. (Lesson 9)
Opt out. Turning off a setting that would otherwise let your new chats be used for model training. Opting out usually doesn't mean a company keeps nothing, and it may not cover feedback you choose to send or safety reviews. (Lesson 9)
Feedback (thumbs-up or thumbs-down). A signal you choose to send about a reply. Several companies say that when you send feedback, the related conversation may be used to improve models even if you've opted out of training. (Lesson 9)
Incognito chat / Private Chat / Temporary Chat. Product modes that typically keep a conversation out of your visible history and out of model training, while still keeping it for a short period for safety or service reasons. The name and rules depend on the tool; check the company's current help page. (Lesson 9)
How this course works (learning words)
The Prompt Check (rubric). The scoring guide you use to grade your own prompts. A rubric describes what weak and strong look like for each part. The Prompt Check has six rows (Task; Context; Reader and tone; Shape and limits; Safe to send; Check built in) plus three follow-up rows (Points at the problem; Says what to keep; One main change). Rows unlock lesson by lesson. Row 5, Safe to send, must always score 2. (Lessons 1–11)
Worked example. A fully solved example, here a weak prompt and a stronger one, with notes explaining each change. Studying these first makes practice easier. (Every lesson)
Fix-it drill. A short exercise where you rewrite a weak prompt or follow-up into a strong one, then compare with a model answer. (Every lesson)
Model answer. One strong version of an answer, shown after you try. It's never the only right answer. (Every lesson)
Retrieval check. Questions you answer from memory before looking at the answers. Pulling information out of memory helps you keep it. (Every lesson)
Warm-up (spaced review). Two or three quick questions from earlier lessons at the start of a lesson. Coming back to skills over time helps them stick. (Lessons 2–11)
Prompt Practice Lab. A practice session at the end of each module with eight stations: warm-up, before and after, spot the weak prompt, fix-it drills, run and compare, which skill, make it yours, and a mastery check. (Lesson 11)
Mastery check. A short self-check (recall questions plus one real-feeling task) that shows whether you can do the module's skills on your own. "Not yet" points you to exactly what to revisit. (Lesson 11)
Voice coach. An optional AI coach (called River) in the "Talk to your coach" panel on each lesson page. It can warm you up, quiz you, and coach your prompts by asking questions, by voice or by typing. It's an AI, not a person. Everything it does is also available in the lesson text. (Lessons 1–11)
Prompting Lab terms (lessons 3, 4, 6, 7, and 11)
Prompt Card. The checklist for building a prompt on purpose. Its rows match the Prompt Check: 1 Task, 2 Context, 3 Reader and tone, 4 Shape and limits, 5 Safe to send, 6 Check built in, plus two optional extras, Role and Example. Memory line: "Say what you want, about what, for whom, in what shape — and, when it helps, as whom and like what. Then ask: is it safe to send, and how will I check?" (Lesson 3)
Task. The job you want done, usually starting with a verb: write, explain, list, compare. Row 1 of the Prompt Card. (Lessons 1 and 3)
Shape and limits. The format and length of the reply (list, table, steps, short email) plus what to include or avoid. Row 4. (Lessons 2 and 3)
Role. A part you ask the chatbot to play, such as "a patient cooking teacher." It shapes tone and focus. It doesn't add real expertise or make facts more reliable. Previewed in the starter; taught fully in Module 4. Prompting skill P8. (Lesson 3)
Example (in a prompt). A short sample of what you want, so the chatbot can match the pattern. Technical guides call prompts with examples "few-shot" prompts. Previewed in the starter; taught fully in Module 4. Prompting skill P9. (Lessons 3 and 4)
Before and after. A pair showing a weak prompt and a stronger version, with notes on which rows changed. (Lesson 4)
Fading. Taking away help step by step — fully worked examples, then partly worked, then hints, then on your own — as your skill grows. (Lesson 4)
Conflicting limits. Limits that can't both be met, such as "detailed and thorough" and "under 30 words." (Lesson 6)
Loaded question. A question that assumes something is true, such as "Why is X essential for everyone?" Ask openly instead, and check a reliable source. (Lesson 6)
Interleaving. Mixing different kinds of practice problems together, so you learn to tell which fix each one needs. (Lesson 6)
Ask me questions first. A coaching move: ask the chatbot to ask you a few questions before it answers, so you can see what context it's missing. (Lesson 6)
Coach my prompt. A coaching move: ask the chatbot to critique your prompt against the Prompt Card without answering it. (Lesson 6)
Prompt template. A reusable prompt with labeled blanks for the parts that change each time. Your personal template is built from the Prompt Card. Module 11 builds it into a tested library. Prompting skill P12. (Lesson 7)
Check line. The last lines of your template, asking the chatbot to ask you questions if something important is missing and to say when it's unsure. It helps, but it doesn't guarantee every gap or error is caught. Row 6. (Lesson 7)