Activity Bank
Two eighth graders check an AI summary line by line against a handwritten historical letter.

All activities Student learning

Primary Source vs. AI Summary

Students read a clearly fictional 1911 letter, then catch an AI summary that changes dates, invents a quote, turns a rumor into a fact, and erases the neighbors who helped.

Print handouts

Overview

AI summaries are fast and fluent, which makes them easy to trust. But a summary is only useful if it is faithful to the source. Students read a primary-source-style letter (fictional and written for this lesson) from a 13-year-old describing a flood in her town, then compare it line by line with a mock AI summary. The summary sounds right but contains six distortions: a wrong date, an exaggeration, an invented quote, an uncertain rumor stated as fact, an added detail, and a family whose help disappears. Students learn a habit historians already use: go back to the source.

Objectives

  • Students will compare an AI summary with a primary source and identify at least four distortions, citing the paragraph in the source that shows each.
  • Students will classify each distortion (wrong detail, exaggeration, invented, certainty added, left out, added detail) and explain why it matters to someone relying on the summary.
  • Students will write a corrected summary and explain when an AI summary is and isn't a trustworthy stand-in for a source.

Materials

On paper

  • Handout A: The Letter (1 per student)
  • Handout B: The AI Summary (1 per pair)
  • Handout C: Distortion Detector (1 per pair)
  • Two colors of highlighters or pencils per pair

On screen

  • Optional: a generative AI chatbot your district approves, teacher-driven on the projector, to summarize the letter live
  • Optional: a digital annotation tool so pairs can link each summary claim to a paragraph in the letter

Before you start

  1. Print Handouts A, B, and C. Handout A is labeled fictional; point that out so students don't search for Willow Fork.
  2. Read the key in this plan's Practice step so you know the six distortions (and the accurate parts).
  3. Optional: paste the letter's text into your district-approved chatbot, ask for a summary, and save the result. Real outputs vary; compare whatever you get using the same organizer.

Step by step

  1. 10–4 min

    Hook

    The telephone game, historian edition

    Ask: "Have you ever heard a story from a friend who heard it from someone else, and it changed along the way?" Take one or two examples. Say: "Historians have a rule: when you can, go back to the original. Today we'll test whether an AI summary follows that rule." Explain primary source (made by someone who was there, at the time) vs. secondary source (made later, about it). Ask: "Which is an AI summary?"

    Facilitator noteStudents often call an AI summary a primary source because it quotes the original. Clarify: it's a secondary source, written later, by something that wasn't there.

  2. 24–12 min

    Explore

    Read the letter first

    Students read Handout A individually, before they see the summary. In the margin they note: who wrote it, when, what happened, what Ada was sure of, and what she was unsure of. Pairs quickly compare notes.

    Facilitator noteReading the source first matters. If students read the summary first, its version tends to stick.

  3. 312–27 min

    Practice

    Distortion detectives

    Pairs read Handout B and check each sentence against the letter, using one color for matches the source and another for doesn't match. They log each mismatch on Handout C with the paragraph number and a distortion type. Key: (1) date given as May 1912; the letter is April 1911 (wrong detail). (2) "the flood destroyed the town of Willow Fork"; the letter says water reached the mill floor and the lower road, and some cellars flooded (exaggeration). (3) the quote "We lost everything" never appears (invented). (4) "caused by the new dam upstream"; Ada writes that some people say so but Papa says no one knows yet (certainty added). (5) the Castillo family, who sheltered Ada's family and rowed the books to safety, is missing (left out). (6) "a young farm girl"; her father runs the mill, and the letter doesn't say they farm (added detail). Accurate: school closed for four days; no one was hurt; the Petersons lost two cows; neighbors carried grain to the loft.

    Facilitator notePush pairs to explain why each distortion matters. Leaving out the Castillos erases who actually helped; the dam claim turns a rumor into history.

  4. 427–35 min

    Debrief

    Why did the summary do that?

    Collect distortions on the board by type. Ask: "Which distortion would matter most if you only read the summary for a history project?" Then ask: "Why might an AI tool do this?" Guide students to the idea that AI tools generate likely-sounding text from patterns. Dramatic words and neat causes are common in writing about floods, so they can slip in even when the source doesn't say them. If you ran the letter through a district-approved chatbot, show that output now and check it the same way.

    Facilitator noteAvoid "AI lies." Better: "AI predicts plausible words. Plausible isn't the same as faithful."

  5. 535–42 min

    Create

    Write a faithful summary

    On Handout C, pairs write a three-sentence summary that is faithful to the letter: correct date, Ada's uncertainty kept, the Castillos included. Pairs swap with another pair, who check it against the letter and mark one thing to strengthen.

    Facilitator noteLook for summaries that preserve hedges ("some people said") instead of flattening them.

  6. 642–45 min

    Reflect

    My summary rule

    Each student finishes the last prompt on Handout C: "I can trust an AI summary when… I need to go back to the source when…" Collect Handout C as the exit ticket.

    Facilitator noteStrong answers name stakes: "when I'm going to quote it, cite it, or make a decision based on it."

Paper or screen

Unplugged

Designed for paper: the letter, the mock AI summary, and the organizer are all printed. Two highlighter colors make matches and mismatches visible at a glance, and the paragraph numbers make citing easy.

Digital

The teacher pastes the letter into a district-approved chatbot on the projector and asks for a summary; the class checks that live output with the same organizer, since real distortions vary. With a digital annotation tool, pairs link each summary sentence to the matching paragraph in the letter. Students under 13 do not use AI tools directly.

Does it need a screen? A live summary of the same letter shows students that distortions aren't planted by the teacher; real tools produce them too, and differently each time. The core comparison is stronger on paper, where students can lay the two texts side by side.

Evidence of learning

What you should be able to see or collect if it worked.

  • Handout C lists at least four distortions, each with a paragraph number from the letter.
  • Students explain why a distortion matters, especially the dam claim (a rumor turned into fact) and the missing Castillo family.
  • Rewritten summaries keep Ada's uncertainty and include the people the AI summary left out.

Adaptations

Grades 3–5
Read the letter aloud and use a four-sentence summary with two distortions (the date and the missing Castillos). Students hunt with "Does the letter say that? Where?"
Grades 9–12
Pair with a real primary source from your social studies curriculum, have students run it through a district-approved tool, and write a short analysis of the distortions they find.
Emergent bilinguals
Pre-teach flood, cellar, loft, mill, rumor, and provide a picture glossary. Pairs may discuss in any language and mark the matching sentences with arrows.

Standards connections

Students compare an AI summary with a primary-source letter, cite paragraphs for each distortion, and write a faithful summary, core social studies source skills.

See how all activities align

Reflect

  • Can I identify trustworthy sources? How did going back to the original change what I believed happened?
  • Why might an AI summary sound confident even when the source was unsure?
  • When is it worth the time to read the original instead of a summary?

What comes next

In the next research task, students must attach one "source check" to any AI summary they use: one sentence from the summary and the matching line from the original. At home, students can ask a family member to tell a family story, then compare it with a sibling's or cousin's version.

Pairs well with