Writing a Prompt That Says What You Actually Want · seed 1 · A4, ink-friendly. The answer key prints on its own page for grown-ups.

Say what you actually want

Computing · Artificial Intelligence · ages 14-16
Name ______________________   Date ____________

  1. A vague request earns a vague answer. Why?

    • The model dislikes short prompts
    • Vague answers cost extra
    • The model fills gaps you left open

  2. Which request is strongest?

    • Write something about stuff
    • Make five quiz questions for a 14 year old, each explanation under 30 words
    • Surprise me with words

  3. What material should you supply with an essay request?

    • Your password and address
    • Nothing at all
    • Topic, length, stance and your notes

  4. Before reading the answer, what should you state?

    • How fast you can read
    • What would make that answer wrong
    • The model's favorite color

  5. What is the best first round when drafting with a chatbot?

    • Demand the final essay at once
    • Ask for sources and stop
    • Ask for a plan, then expand only the part you chose

  6. A student assumes the chatbot knows their class novel. What should they do instead?

    • Repeat the vague request louder
    • Switch to a different novel
    • Paste the key passages and the exact question

  7. A model answers the request in your head even when your words say less.

    Circle one:   True   False

  8. An answer looks polished but states a false date. Who must catch it?

    • You, by checking facts yourself
    • Nobody, polished means verified
    • The model, which verifies everything
LightMySky · lightmysky.comW1-mt_dJ5-SDX3Ck-s1

Answer key

For grown-ups. Fold this page away before handing over the rest.

Say what you actually want W1-mt_dJ5-SDX3Ck-s1

  1. The model fills gaps you left open · The model answers what is typed, guessing at the rest.
  2. Make five quiz questions for a 14 year old, each explanation under 30 words · It names the task, audience, format, and limits in one sentence.
  3. Topic, length, stance and your notes · The model needs your case details to argue your point.
  4. What would make that answer wrong · A wrongness test lets someone else judge whether the answer met it.
  5. Ask for a plan, then expand only the part you chose · Plans let you keep what helps before spending words on the rest.
  6. Paste the key passages and the exact question · Supply the material rather than assuming the model knows the case.
  7. False · It answers the request in front of it, not the one unspoken.
  8. You, by checking facts yourself · The tool cannot verify facts, so checking stays your job.
Worksheet · LightMySky