1. Objective
By the end of this lesson, you will be able to:
- Explain, in a sentence or two, what an AI chatbot is and why a smooth answer can still be wrong.
- Write a first clear request that says what you want done and what you want back.
2. Hook
Transcript
Welcome to My First AI Skill. You don't need any technical background. In this first lesson, you'll find out what you're actually talking to when you use an AI chatbot, and you'll send your first prompt.
Picture leaving a sticky note for a very well-read houseguest who has never met you. The note just says: "Eggs?"
They might cook you eggs. They might buy eggs. They might write you a poem about eggs. None of those would be wrong, exactly. They just had to guess.
An AI chatbot is a bit like that houseguest. It has read an enormous amount, it writes well, and it can't see what you meant. This lesson shows you what it actually is, and what that means for how you ask.
3. Warm-up: what you already know
You don't need any technical background. You already have useful experience. Take thirty seconds to think about these (no need to write anything):
- Think of a time you asked someone for help and got the wrong thing. What was missing from your request?
- When you use a search engine, what do you get back: an answer, or a list of links?
Hold on to those two thoughts. Both come back in this lesson.
4. The idea
Transcript
Here's the big idea, in plain words. A chatbot isn't a person, and it isn't a search engine. It's a program that writes a reply one small piece at a time, by predicting what's likely to come next after your words. That has three consequences. First, the reply can sound smooth and sure, and still be wrong. Second, the same prompt can get a slightly different reply next time, and that's normal. Third, what you write shapes what comes back. So your first prompting skill is the clear ask. Say what you want done, and say what you want back. For example: write, two short paragraphs. That one habit removes a lot of guessing.
A chatbot is a program you talk to
Artificial intelligence (AI), in everyday talk, means computer systems that can do things that used to take human intelligence, like writing, summarizing, or answering questions in ordinary language.
An AI chatbot is a program you use by typing (or sometimes speaking) a message and reading a reply. The message you send is called a prompt. The reply is called a response or output. Prompting is the skill of writing prompts that get you useful help. It's the main skill this course teaches.
These tools are not people. They don't have a life, feelings, or a private memory of you, unless a product feature is specifically set up to store things you've said. A friendly tone in the reply is a writing style, not a sign that someone is on the other side.
They are also not the same thing as a search engine. A search engine is built to find pages and show you links. A chatbot is built to write a reply. Some chatbots can now search the web while they work, which can help with current facts. Even then, the chatbot is still composing the sentences you read, and a link in the reply doesn't make those sentences trustworthy. You'll practice checking in lesson 8.
How a reply gets written
Today's chatbots run on a language model, a program trained to work with language. You'll sometimes see large language model, which just means one built at a very large scale. In this course, model is the short word for it.
The model was trained on a huge amount of text: books, websites, and other writing. During training, it practiced one simple-sounding job over and over: given the words so far, guess what comes next.
When you send a prompt, the model does that job for you. It predicts a likely next token (a word or a piece of a word), adds it, predicts the next one, and keeps going until the reply is done. Stephen Wolfram describes this as repeatedly asking, in effect, "given the text so far, what should the next word be?" (see Sources).
Three things follow, and all three matter when you prompt:
- The reply can sound smooth even when it's wrong. Fluency is what the model is good at. Fluency (how smooth and natural writing sounds) isn't the same thing as truth. OpenAI has written that models can produce "plausible but false statements," often called hallucinations, and may sound confident while guessing (see Sources, and lesson 8).
- The same prompt can get a slightly different reply next time. The model chooses among likely next words. It doesn't look up one stored answer the way a calculator returns
4for2 + 2. - What you write shapes what comes back. The model continues from your words. If your prompt is vague, it fills the gaps with whatever is typical. If your prompt is clear, it has much less to guess. That's why prompting matters.
You don't need the math. Just remember: this is a pattern-predicting program writing text one piece at a time, starting from what you wrote.
Your first prompting skill: the clear ask
A clear ask says two things:
- What you want done (write, explain, list, plan, compare, shorten…)
- What you want back (two paragraphs, a five-item list, one text message…)
That's it for today. Later lessons add more.
Tool notes: free chatbots (checked mid-November 2026)
You need one free chatbot in another browser tab. As of mid-November 2026, all of these offered a free way to chat. Plans, limits, and screens change, so check the product's own page before you sign up, including age requirements and terms of use.
- ChatGPT: chatgpt.com. OpenAI offered a free plan with no set time limit. Everyday text chats were described as unlimited, with separate limits for extras such as file uploads and image generation. A free account is the usual way to start. See ChatGPT's free plan page and the Free tier FAQ.
- Claude: claude.ai or claude.com. Anthropic offered a free plan in supported regions, with limited usage that resets on a five-hour and weekly schedule. See Claude plans and pricing.
- Gemini: gemini.google.com. Google offered free access with a personal Google account. Usage limits refresh on a schedule (including a five-hour window and a weekly cap). See Gemini plans and Gemini Apps limits.
- Grok: grok.com, the Grok mobile app, or the Grok tab on X. xAI offered a free plan with daily access and limited chat, without a published fixed message count. See grok.com and xAI's consumer FAQs.
You only need one. If a free plan asks you to create an account, that's normal. You don't need a paid plan. If you hit a usage limit, wait for it to reset or try a different free chatbot. That's a product limit, not a mistake on your part.
5. Worked example
Here's a weak ask and a clear ask for the same job. Read the notes. They show how an experienced user thinks.
Before:
Tell me about scrambled eggs.
After:
Write two short paragraphs about making scrambled eggs.
First paragraph: as a patient cooking teacher talking to a complete beginner.
Second paragraph: as a hurried roommate leaving a note on the fridge.
Keep each paragraph under 60 words.
What changed, and why it helps:
- "Write" instead of "tell me about": the task is now clear. (Rubric row 1: Task)
- "Two short paragraphs": you've said what you want back, so the model doesn't have to guess between a history lesson and a recipe.
- The two voices and the word limit are extras you'll learn properly in lesson 2. For now, just notice how much less guessing is left.
6. Guided practice
Open your chatbot and start a new chat. Copy the prompt below, paste it, and send it.
Prompt 1 (copy and paste):
Write two short paragraphs about making scrambled eggs.
First paragraph: as a patient cooking teacher talking to a complete beginner.
Second paragraph: as a hurried roommate leaving a note on the fridge.
Keep each paragraph under 60 words.
What to look for: two short paragraphs on the same topic, in two clearly different tones. The exact words will vary from tool to tool, and even from one try to the next. That's expected.
Notice what just happened. You didn't "look up scrambled eggs." You asked a program to write, and it followed your directions about tone and length closely enough to be useful. Drafting, rephrasing, and shaping text from your directions is where these tools often help most.
If it doesn't go as described: if the reply runs long or the tones feel similar, don't worry. In lesson 5 you'll learn to fix that with a follow-up.
Now, in the same chat, send this:
Prompt 2 (copy and paste):
In one short paragraph, explain how you produced those two versions. Use plain language, as if I have never used an AI chatbot before.
What to look for: the chatbot may mention following your instructions, predicting or generating text, or patterns from training. It may also skip important limits, or describe itself in a way that sounds more human than it is.
Treat this reply as a conversation, not as the textbook. The reliable idea is the one in section 4: the model predicts likely next pieces of text. That's why it can switch tones easily, and why a smooth answer can still be wrong.
7. Independent practice
Spot the weak prompt
Which of these three prompts is the weakest clear ask? Decide before you read the answer.
- A.
Write a five-item checklist for cleaning out a fridge. - B.
Fridge. - C.
Explain in three sentences why you should clean out a fridge regularly.
Answer
B. It names neither the task nor what you want back. A and C both say what to do (write a checklist; explain) and what shape to return (five items; three sentences).
Fix-it drill
Rewrite this weak prompt as a clear ask. Say what you want done and what you want back.
Birthday party.
Write your version on paper or in a notes app first. Then send it to your chatbot and see what comes back.
One strong version (not the only one)
List eight low-cost activity ideas for a backyard birthday party for a 7-year-old. One line each.
Why it's stronger: it says what to do (list ideas), what to return (eight, one line each), and enough detail that the model isn't guessing between a guest list, an invitation, and a cake recipe.
8. Rubric self-check
You'll grade your own prompts all through this course with a short rubric called the Prompt Check. A rubric is a scoring guide that describes what weak and strong look like. Today you only use row 1. The other rows unlock as you learn them.
Score your fix-it prompt:
| Row | Question | 0 | 1 | 2 |
|---|---|---|---|---|
| 1. Task | Does it say what I want done and what I want back? | Neither is clear ("Birthday party.") | One is clear ("Give me birthday party ideas.") | Both are clear ("List eight low-cost activity ideas… one line each.") |
Where to next: If you scored 0 or 1, add the missing half (the action or the shape) and send it again. Compare the two replies. That comparison is the lesson in miniature.
9. Retrieval check
Answer from memory first, then check.
- In one sentence, what is a prompt?
- Why can a chatbot's answer sound polished and still be wrong?
- Is an AI chatbot the same thing as a search engine? Why or why not?
- If you send the same prompt twice and get two slightly different replies, did something go wrong?
- What two things does a clear ask say?
Answers
- A prompt is the message you send to the chatbot.
- The model is especially good at writing fluent language. Fluency and truth aren't the same thing, and the model can guess.
- No. A search engine finds pages and links. A chatbot writes a reply. Some chatbots can search the web as they write, but they're still composing the answer, and you still check important facts.
- No. Variation is normal. The model chooses among likely next words; it doesn't return one stored answer.
- What you want done, and what you want back.
10. Recap
Transcript
Let's recap. A chatbot is a program you talk to. It writes by predicting likely next words, starting from what you wrote. That's why clear prompts help, and why a fluent answer can still be wrong. A clear ask says two things: what you want done, and what you want back. And you only need one free chatbot. This skill travels from tool to tool. Next time you open a chatbot, check your first message for both halves of a clear ask before you send it.
- An AI chatbot is a program you talk to, not a person, and not a search engine.
- It writes by predicting likely next words, starting from what you wrote. That's why clear prompts help and why fluent answers can be wrong.
- A clear ask says what you want done and what you want back.
- You only need one free chatbot. The skill travels from tool to tool.
You can now: send a clear first request and explain why the reply might vary or be wrong. Use it this week: next time you open a chatbot, check your first message for both halves of a clear ask before you send it.
Next: lesson 2, giving the chatbot enough context to help you specifically.
11. Talk it through with your coach
- Warm up: your coach reads the hook aloud and asks, "What do you think the houseguest needed to know?" You answer in a sentence.
- Quiz me: your coach asks the five retrieval questions aloud and gives short feedback on each.
- Text-only alternative: sections 2 and 9 above cover the same ground.
Talk to your coach
River is an AI voice coach, not a person. River knows this lesson and can quiz you, listen to a prompt you're working on, and give hints. River won't do the exercise for you. You can talk out loud or type.
Ideas to start:
Sources
- Stephen Wolfram, "What Is ChatGPT Doing … and Why Does It Work?" (2023), on next-word prediction: writings.stephenwolfram.com
- OpenAI, "Why language models hallucinate," on next-word prediction and plausible but false statements: openai.com
- OpenAI Help Center, "Does ChatGPT tell the truth?": help.openai.com
- OpenAI Help Center, "Prompt engineering best practices for ChatGPT" (be clear and specific), checked 22 November 2026: help.openai.com
- Free-plan pages checked mid-November 2026: ChatGPT Free, ChatGPT Free tier FAQ, Claude plans, Gemini subscriptions, Gemini Apps limits, xAI consumer FAQs