AI hallucination examples for grade 7 ELA: 5 to teach Monday

It happens in middle school ELA classrooms across the country: a student turns in a research paragraph citing a paper that does not exist. ChatGPT fabricated it — confident author, plausible journal, fake DOI. The student didn’t lie. The AI did. Teachers we read on TPT reviews keep asking for the same thing: 5 real cases to project, a clear 3-step verification routine, and a printable worksheet that fits in one period. This post — AI hallucination examples for grade 7 ELA — gives you exactly that.

This post is the anchor lesson for hallucination fact-checking in Track B: Going Deeper — AI Ethics and Critical Thinking, the track for teachers whose students are ready to investigate AI errors and bias.

What an AI hallucination actually is (in grade 7 terms)

A hallucination is when an AI generates confident, plausible-sounding text that is factually wrong. Not a typo. Not a misunderstanding. A made-up source the AI presents with the same tone it uses for true ones.

Your students already know this concept. You taught it last unit. Holden Caulfield. Nick Carraway. The narrator sounds sure. The narrator is wrong. ChatGPT is an unreliable narrator wearing a research-assistant costume — same skill set, new disguise.

Here is the why, in one paragraph you can read aloud. Large language models predict the most-plausible next word, not the most-true next word. They were trained on a giant pile of human writing and rewarded for sounding right. Nobody trained them to be right. When a model does not know an answer, it does not pause and say so. It generates a sentence that fits the pattern of an answer. That sentence often contains a real-looking name, a real-looking date, and a real-looking citation that was assembled, not retrieved.

This sits inside AI4K12 initiative Big Idea #3 (Learning — computers learn patterns from data) and lands hard on AI4K12 initiative Big Idea #5 (Societal Impact — what happens when the confident wrong answer shows up in a courtroom, in a Google ad, in your third-period homework).

5 AI hallucination examples to project tomorrow — fact-check Monday

These are real. Each takes 90 seconds on the projector. Each one your students can verify themselves before the bell.

1. Mata v. Avianca — the lawyer’s fake cases. In 2023, two New York lawyers submitted a federal brief drafted with ChatGPT. The brief cited six court cases — Varghese v. China Southern Airlines among them — that did not exist. The judge fined the firm $5,000 and the story made the front page of the New York Times. Verify it live: have a student search “Varghese v. China Southern Airlines” in Google Scholar. Zero results. Source: Wikipedia — Mata v. Avianca.

2. Google Bard’s $100 billion sentence. February 2023: in Google’s promotional ad introducing Bard, the chatbot claimed the James Webb Space Telescope took the first picture of an exoplanet. The European Southern Observatory did that in 2004. Alphabet’s stock fell roughly 9 percent the same day — about $100 billion in market value. Verify it live: open NASA’s First Images page next to the ESO 2004 announcement. Source: CNN — Bard demo error.

3. The fake footnote your students will write. A 2023 study in Scientific Reports tested 300 ChatGPT-generated citations. 32.3 percent were fully fabricated. Real-sounding authors. Plausible journals. Convincing DOIs. None of them existed. Verify it live: have a student paste any ChatGPT citation into Google Scholar. The zero-result page appears in 60 seconds. Source: Scientific Reports — Nature.

4. Google AI told America to put glue on pizza. May 2024: Google’s new AI Overviews told users to add non-toxic glue to pizza sauce so the cheese would stop sliding off. The source was an 11-year-old Reddit joke. Google pulled the result after viral screenshots. Verify it live: open any Food Network pizza recipe. Zero credible chefs recommend adhesive. Source: Evidently AI — hallucination examples.

5. The expert who got fooled by his own warning. December 2024: a Stanford communication professor — a national authority on deepfakes — filed a sworn court declaration partly drafted with ChatGPT. It contained hallucinated citations. The deepfake expert himself missed them. Verify it live: search his cited paper titles in Google Scholar. Zero results. Source: Stanford Daily — Hancock admitted to AI use.

The 3-step verification routine your class runs in one period

Three verbs. Students remember verbs better than rules. Put them on the wall.

1. Pause. Read the AI output. Find the specific verifiable fact — a name, a date, a number, a quoted statistic. Underline it. If a sentence contains no verifiable fact, there is nothing to check.
2. Trace. Ask the question: what primary source could prove or disprove this? A court record. A NASA archive page. The publisher’s own website. Google Scholar for any paper. Wikipedia is a starting point, not a finish line.
3. Verify. Open the primary source. Compare. Write the result on the worksheet — “confirmed,” “contradicted,” or “no source found.”

This is the middle-school version of SIFT (Stop, Investigate the source, Find better coverage, Trace claims) — the framework librarian Mike Caulfield built for college students. For the full SIFT unit with the verification rubric, see the SIFT Hallucination Triage lesson.

Standards the routine hits, by anchor code: CCSS.ELA-LITERACY.W.7.8 (gather information from multiple sources, assess credibility and accuracy of each), and ISTE 1.3.b (evaluate accuracy, perspective, credibility, and relevance of information).

Where this slots into your ELA scope (CCSS + ISTE + AI4K12)

It is part of the AI literacy teaching guide for grades 6–12, a structured path from first lesson through full standards documentation.

This is not a bolted-on AI lesson. It extends the unreliable-narrator work your grade 7 unit already runs. Same critical-reading muscle, new text type.

Standards crosswalk by anchor code:

• CCSS.ELA-LITERACY.W.7.8 — gather relevant information from multiple sources; assess credibility and accuracy of each
• ISTE 1.3.b — evaluate accuracy, perspective, credibility, and relevance of information
• ISTE 1.3.d — build knowledge by exploring real-world issues and pursuing active investigation
• AI4K12 initiative Big Idea #3 — Learning (the mechanism behind why models hallucinate)
• AI4K12 initiative Big Idea #5 — Societal Impact (court fines, retracted ads, wrong homework on Tuesday)

Standards source: ISTE Student Standards. ISTE is a registered trademark of the International Society for Technology in Education. These resources are not affiliated with or endorsed by ISTE.

Suggested slot: one 45-minute period inside your research-writing unit, before students begin independent source work. If you want the companion piece on teaching students to write better prompts in the first place — fewer hallucinations to triage downstream — the CRAFT prompting framework for middle school post covers the prompt-writing side of the same problem.

Print Monday — worksheet, deepening unit, and full triage

Three offerings matched to where your class is this week:

1. Free worksheet — the AI Hallucination Fact-Check Worksheet (grades 6–8) turns the 3-step routine into a paragraph-length student deliverable. One page front, one page back. Grab it free at /free — no email gate.
2. 3-day prompting unit — the CRAFT Framework Prompt Writing Unit (grades 6–8) teaches students to write prompts that produce fewer hallucinations to begin with. Three lessons, scripted. Pairs with this post as the upstream skill.
3. Full SIFT triage lesson — the SIFT Hallucination Triage (grades 6–8) is the deep version for teachers building a media-literacy unit. Includes the verification rubric, answer key, and extension discussion for the fake-court-case scenario above.

PDF, ready to print — Monday lesson plan sorted before the weekend.

This post was drafted with AI assistance and human-finalized.