AI Story Remix

The Idea

AI language models are extraordinarily good at generating text that looks like a natural continuation of something. But they don’t understand your story. They don’t know your character. They don’t care about your world. They’re doing something more like very sophisticated pattern-matching — and that shows up in interesting ways when you give them something personal to continue.

This activity uses the gap between “looks like a story continuation” and “actually continues your story” to teach something real about how large language models work.

Part 1: Write Your Opening (10 minutes)

Have your child write 3–5 sentences that start an original story. Give them one rule: include at least one specific detail that only they would know — a character trait, a setting detail, something personal.

Example openings:

• “Mia had been saving the acorn since September. Not because acorns were valuable, but because it was the last thing her grandfather gave her before the hospital. She kept it in her jacket pocket where a hole was forming…”
• “The robot in the basement wasn’t supposed to make sounds at night. My dad said it was just the cooling system. But the cooling system didn’t hum in rhythms of three.”

The more specific and personal, the better. The specificity is the test.

Part 2: Get the AI’s Version (5 minutes)

Open a free AI chatbot. Paste in the opening. Then give this prompt:

“Continue this story for 2–3 more paragraphs. Keep the character’s voice and the tone consistent.”

Generate the response. Read it together before the child sees it alone — the reaction is often more candid.

Part 3: The Critique (15–20 minutes)

This is the heart of the activity. Work through these questions together:

Voice and character:

• Does this sound like the character you invented? What’s different?
• Did the AI invent new character traits or personality details? Are they consistent with your vision?
• Did it use words or phrases that feel wrong for this story?

Plot and logic:

• Did the AI continue the story in a direction you expected? Wanted? Neither?
• Did it resolve the tension too quickly? AI systems often try to “wrap things up” rather than let a story breathe.
• Did it add anything surprising that actually worked?

Specific details:

• Did the AI remember and use the specific details you included? (Often it ignores them or handles them generically.)
• Did it invent plausible-sounding but wrong details about your world?

The quality question:

• On a scale of 1–10, how good is this continuation?
• If a classmate had written this, what would you say to them about it?

Part 4: The Experiment (5–10 minutes)

Now try the same opening but with a different prompt:

“Continue this story in a completely different direction. Take it somewhere I wouldn’t expect.”

Or:

“Continue this story as if it’s a scary story. Or a funny story. Or a mystery.”

Notice how dramatically the output changes based on the instruction. Then ask: “Did the story actually change? Or just the surface tone?” Often the AI will just add spooky adjectives rather than genuinely restructuring the narrative.

The AI Connection

“Language models like this one don’t actually understand your story. They don’t know who Mia is or why the acorn matters to her. They’ve been trained on millions of stories, and they’re very good at predicting what text usually comes after text like yours. But they don’t have the intent you had when you wrote it.”

This is a useful distinction: prediction vs. understanding. The AI is predicting likely continuations based on patterns in training data. You had an intention when you wrote your opening. Those are very different things — and they produce very different results.

A useful follow-up: “What would you have to put in the prompt to get the AI to do what you actually wanted?”

This is the beginning of prompt engineering — the skill of communicating precisely with AI systems.

Optional Worksheet: Story Critique

Rate the AI’s continuation: __ / 10

One thing the AI got right: ___

One thing the AI got wrong or missed: ___

One thing I’d tell the AI to do differently: ___