Lesson 2 of 11 in the free starter module · Progress is saved in this browser

All 11 lessons
  1. 1 What an AI chatbot actually is
  2. 2 Give it enough context to help you
  3. 3 Prompting Lab 1 — The Prompt Card
  4. 4 Prompting Lab 2 — Before and after
  5. 5 Improve the answer instead of starting over
  6. 6 Prompting Lab 3 — Fix-this-prompt drills
  7. 7 Prompting Lab 4 — Your personal prompt template
  8. 8 Check before you trust
  9. 9 Keep private things private
  10. 10 Run the full loop on a real task
  11. 11 Prompt Practice Lab and starter mastery check

Free starter · Lesson 2

Give it enough context to help you

Time
about 25–30 minutes
Prompting skill
P1 Clear ask, P2 Context, P3 Reader and tone, P4 Shape and limits
You'll need
Your free AI chatbot, and a new chat for this exercise.

Listen Talk to your coach


1. Objective

By the end of this lesson, you will be able to:

  • Rewrite a vague request into a useful prompt by adding context, the reader and tone, and limits.
  • Spot a detail in a reply that the chatbot invented, and remove it.

2. Hook

Listen: welcome 0:11 · AI voice: Bella
Transcript

Imagine asking a helpful stranger to write a text for you. They'd need to know who it's for and what it's about. A chatbot needs the same things. Today you'll learn to give it that context.

Two people send a chatbot a request about the same evening.

One types: "Help me write a text."

The other types two sentences about who it's for, what happened, and how it should sound.

Same chatbot, same minute. One gets something generic; the other gets something they could almost send. The difference wasn't luck or a better tool. It was what they put in the prompt.

3. Warm-up (from lesson 1)

From memory:

  1. A chatbot writes its reply by predicting what?
  2. What two things does a clear ask say?
Answers
  1. Likely next words (or pieces of words, called tokens), starting from what you wrote.
  2. What you want done, and what you want back.

4. The idea

Listen: the key idea 0:41 · AI voice: Bella
Transcript

A chatbot can't see your kitchen, your inbox, or the person you need to text. It only knows what you put in the prompt. So a vague request gets a generic answer. Before you send, glance over four questions. Who is this for? Why am I asking, and what facts must it get right? What tone do I want? And what are the limits, like length, format, and things to avoid? You won't need all four every time. Here's a handy test: if a helpful stranger would need it explained, explain it in the prompt. And remember, you stay in charge of the facts. If the reply adds a detail you never gave it, treat that as a suggestion, and delete or correct it.

In lesson 1, you saw that a chatbot continues from your words. It isn't sitting beside you. It can't see your kitchen, your inbox, or the person you need to text, unless you put that in the prompt.

So a short, vague request often gets a short, vague, or generic answer. The model fills the gaps with whatever seems typical. Typical isn't the same as useful for your situation.

This lesson adds three prompting skills to your clear ask:

  • Context (P2): the situation and the facts the reply must get right. Context means the background the chatbot needs so it doesn't have to guess.
  • Reader and tone (P3): who it's for and how it should sound. Tone is the manner of the writing: warm, formal, brief, playful, straight to the point.
  • Shape and limits (P4): length, format, what to include, and what to avoid. A constraint is simply a limit you set, like "under 400 characters" or "no jokes."

You don't need a formula or fancy wording. Everyday language is enough. Before you send, glance over four questions:

  1. Who is this for? You, a sister, a coworker, a landlord, a child's teacher? The reader changes the words.
  2. Why am I asking? What's going on, in a sentence or two? Include the facts the reply must get right.
  3. What tone do I want? If you don't say, the model will pick a generic one.
  4. What are the limits? Length, format, things to include, things to avoid.

You won't need all four every time. A one-line question can be plenty for a small job. Here's a handy test: if a helpful stranger would need it explained, explain it in the prompt. Anthropic, the company that makes Claude, suggests a similar test in its prompting guide: show your prompt to someone with little context on the task, and if they'd be confused, the model probably will be too (see Sources).

You stay in charge of the facts

Give the model the facts you want it to use. If you leave a fact out, it may invent a plausible stand-in, like a reason for leaving or a time that sounds right. That invented detail can slip into a message you almost send.

If a reply includes a detail you didn't provide, treat it as a suggestion, not as something that happened. Delete or correct it before you use the text.

5. Worked example

Before:

Help me write a text.

After:

Write a short, warm text message to my sister. I have to leave her birthday dinner at 8:30 because I have an early shift at 6 a.m. I don't want her to think I don't care. Keep it under 400 characters. Offer to take her out for coffee this weekend.

What changed, and which rubric row it fixes:

Change Row
"Write a short text message" says the task and what you want back 1. Task
"Leave her birthday dinner at 8:30… early shift at 6 a.m." gives the situation and the facts 2. Context
"To my sister… warm… I don't want her to think I don't care" names the reader, tone, and the feeling to get across 3. Reader and tone
"Under 400 characters… offer coffee this weekend" sets a length and one thing to include 4. Shape and limits

The model still chooses the wording. That's its job. It just doesn't have to guess the situation anymore.

6. Guided practice

Start a new chat so this exercise stands on its own.

First, send a weak prompt on purpose. You're doing this to see the difference, not because this is how you should usually ask.

Prompt 1 (copy and paste):

Help me write a text.

What to look for: you'll probably get one of these:

  • a generic text that could be for almost anyone, about almost anything; or
  • a polite question asking who it's for and what you need to say.

Both make the same point: the model didn't have enough context to write your message. (A chatbot that asks questions back is doing something helpful. It's asking for the context you left out.)

Now, in the same chat, send the stronger prompt.

Prompt 2 (copy and paste):

Write a short, warm text message to my sister. I have to leave her birthday dinner at 8:30 because I have an early shift at 6 a.m. I don't want her to think I don't care. Keep it under 400 characters. Offer to take her out for coffee this weekend.

What to look for: the key facts should appear: it's for your sister, you're leaving at 8:30, there's a reason (an early shift), the tone feels warm rather than stiff or jokey, and it offers coffee this weekend. The wording won't match anyone else's, and that's fine.

If it doesn't go as described: some tools land a little over 400 characters. You'll fix that in lesson 5. If the reply adds a detail you never gave (a restaurant name, a gift), cross it out. That's the "you stay in charge of the facts" habit.

Fill in the blanks

This time, you supply the details. Copy the frame, fill each [ ] with something you're comfortable typing, and send it. Use a made-up situation if you prefer.

Write a [short / medium-length] [text / email / note] to [who it's for]. 
The situation: [what happened, in one or two sentences]. 
Tone: [warm / polite and brief / friendly but firm]. 
Keep it [length limit]. Make sure it [one thing to include]. Don't [one thing to avoid].

Keep private details out for now: no account numbers, home addresses, health details, or other people's private information. Lesson 9 covers this in depth.

7. Independent practice

Spot the weak prompt

All three are asking for help with the same job. Which is weakest, and which rubric rows does it miss?

  • A. Write a polite, two-sentence email to my child's teacher, Ms. Lopez, saying my son will miss school Friday for a family event and asking how he can catch up.
  • B. Email to a teacher about missing school.
  • C. Write a short email to my son's teacher. He'll be absent Friday. Keep it friendly.
Answer

B is weakest. It has a rough task (row 1 = 1) but no situation details (row 2 = 0), no tone (row 3 = 0 or 1, since "a teacher" is the reader), and no limits (row 4 = 0). C is in the middle: it has the reader, a fact, and a tone, but nothing about what to ask for, and "short" is vague. A covers all four rows. (Using a teacher's name is fine here; it's not sensitive. You'll learn what is sensitive in lesson 9.)

Fix-it drills

Rewrite each prompt. First try the one with a hint, then the one without.

Drill 1 (with a hint):

Write something for my landlord.

Hint: What's the situation? What tone? How long?

Drill 2 (no hint):

Thank-you note.

Send your best version of each to your chatbot. Then score it in section 8.

One strong version of each (not the only one)

Drill 1:

Write a polite, brief email to my landlord. The kitchen faucet has been dripping for a week, and I'd like it repaired. I'm available weekday evenings after 5. Keep it under 100 words and don't sound annoyed.

Drill 2:

Write a warm thank-you note to my neighbor, who watered my plants while I was away for a week. Three or four sentences, friendly and casual. Mention that the tomatoes survived.

8. Rubric self-check

Score your Drill 1 and Drill 2 prompts. Rows 2–4 are new this lesson.

Row Question 0 1 2
1. Task Does it say what I want done and what I want back? Neither One Both
2. Context Did I give the situation and the facts it must use? No facts Some facts, key ones missing The facts a stranger would need
3. Reader and tone Did I say who it's for and how it should sound? Neither One of the two Both
4. Shape and limits Did I set length, format, include, or avoid as needed? None Vague ("short") Specific ("under 100 words; don't sound annoyed")

Reading your score: look at your lowest row, not your total. That's the one to fix next. A tiny job may not need every row; mark any that don't apply as "n/a."

Where to next: If row 2 is low, ask yourself, "What would a stranger need to know?" If row 4 is low, add one number (a length) and one "include" or "avoid."

9. Retrieval check

From memory first:

  1. What does "context" mean in this lesson?
  2. Name the four questions that can turn a vague prompt into a useful one.
  3. A chatbot writes a thank-you note and adds "Thanks again for the flowers on Tuesday," but you never mentioned flowers or Tuesday. What should you do?
  4. True or false: a good prompt has to sound technical or use special keywords.
  5. (From lesson 1) Why might the same prompt get a different reply the second time?
Answers
  1. The background the model needs (the situation, the reader, the facts, and your limits) so it doesn't have to guess.
  2. Who is this for? Why am I asking? What tone do I want? What are the limits (length, format, include, avoid)?
  3. Treat that detail as invented. Remove or correct it before you use the note.
  4. False. Everyday language is enough. Clarity matters more than special wording.
  5. The model chooses among likely next words each time; it isn't looking up one stored answer.

10. Recap

Listen: recap 0:23 · AI voice: Bella
Transcript

Quick recap. A chatbot can't see your situation unless you describe it. Vague prompts get generic answers. Before you send, think: who it's for, why you're asking, the tone, and the limits. And you stay in charge of the facts, so delete any detail the reply invented. This week, try writing a prompt for a message you actually need, using only details you're comfortable typing.

  • A chatbot can't see your situation unless you describe it.
  • Vague prompts get generic answers, or a round of questions you could have answered up front.
  • Before you send: who it's for, why you're asking, the tone, and the limits.
  • You stay in charge of the facts. Delete any detail the reply invented.

You can now: turn a one-line request into a prompt that scores well on rows 1–4. Use it this week: think of a message you actually need to send (a thank-you, a scheduling change, a request to a neighbor). Write the prompt using the four questions. Use only details you're comfortable typing. You don't have to send the result to anyone.

Next: lesson 3, Prompting Lab 1. You'll learn the Prompt Card, the parts of a strong prompt, so you can build one on purpose.

11. Talk it through with your coach

  • Coach my prompt: you say your Drill 1 prompt out loud. Your coach doesn't rewrite it for you. It asks the four questions ("Who's it for? What's the situation? How should it sound? How long?") and points out which rubric row is lowest. Then you say your improved version.
  • Text-only alternative: use the rubric table in section 8.
  • Privacy reminder: say only what you'd type, with no private details.

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.

Sources

  • OpenAI Help Center, "Prompt engineering best practices for ChatGPT" (be clear and specific, give enough context, name the tone), checked 22 November 2026: help.openai.com
  • Anthropic, "Prompting best practices" (be clear and direct; give context and the reason behind instructions; the "show it to a colleague with minimal context" test), checked 22 November 2026: platform.claude.com
  • On why models can fill gaps with plausible but invented details: OpenAI, "Why language models hallucinate": openai.com