AI vs human writing activity grade 7: spot the bot

Students already know AI writes things. They say so every time you bring it up. But push them to explain how AI writing is different from human writing, and the room goes quiet — which is the exact gap that makes this a teachable moment. The class has a stack of papers to return, a reading unit running, and one period this week where something new can fit. This activity fills that period: after one 45-minute session comparing AI vs human writing, grade 7 students can name the specific mechanics that give AI away and defend their reasoning on paper.

This activity is part of the critical-reading toolkit in Track B: Going Deeper — AI Ethics and Critical Thinking, which covers hallucination, bias, and the difference between AI and human writing.

Why the “Spot the Bot” game works

Detection challenges tap something discussion guides never do: the competitive instinct. When students are asked to compare two samples and decide which one the machine wrote, the analytical engagement is qualitatively different from reading a passage and answering questions. They’re not retrieving — they’re hunting. Reading-level attention climbs when students have something to win.

That shift maps directly onto academic standards you already cover. ISTE 1.3.b asks students to evaluate accuracy, perspective, and credibility of information — and an AI-generated essay is exactly the kind of information that requires that evaluation. CCSS.ELA-LITERACY.RI.8.6 asks students to determine an author’s point of view and analyze how it shapes content — a mechanic-first comparison activity is the concrete exercise that makes that standard real.

The gap in the free resources currently out there: most are discussion guides (“talk about whether AI writing is good or bad”) rather than mechanics-first, print-ready comparison activities. The can-students-spot-AI-writing-middle-school problem needs a repeatable checklist students can annotate, not an open-ended Socratic circle.

The 5 mechanics that give AI writing away

This is the teaching section. Each mechanic below has a brief concrete example so you can project or read aloud before students see the paired samples.

1. Sentence-length evenness. Human writers vary sentence length instinctively — a short punch after a long buildup. AI models produce sentences that cluster around a statistical average. Every sentence in a paragraph lands between 18 and 22 words. The rhythm is suspiciously steady.

AI sample: “Renewable energy sources offer many important benefits to society. Solar power and wind energy are becoming increasingly common choices. Both options reduce our dependence on fossil fuels.”

Human sample: “Renewable energy sounds obvious. Why aren’t we there yet? The answer, frustratingly, is money — but it’s more complicated than that.”

2. Hedging phrases. AI writing is littered with pre-emptive apologies: “It is important to note,” “It should be mentioned,” “It is worth considering.” These phrases are common in formal academic writing and so they show up constantly in training data — and therefore constantly in AI output. Human writers hedge too, but not three times per paragraph.

3. Generic transitions. Count the “Furthermore,” “In addition,” and “Moreover” in the AI sample. Then count them in the human sample. This alone is often enough to tip the verdict. Human writers vary transitions or drop them entirely. AI writing treats “Furthermore” as a default clause starter.

4. Vocabulary density without specificity. AI writing often deploys complex vocabulary that floats free of any concrete detail. Big words appear, but no specific person is named, no specific place is described, no particular Tuesday-afternoon moment is invoked. “The multifaceted implications of this phenomenon are significant” contains zero verifiable information. Human writers anchor abstract claims to something real.

5. Missing personal anecdote or emotional register. The AI mechanics lesson ELA grades 6–8 teachers most reliably recognize: AI writing has no embarrassment, no affection, no frustration. It knows what those emotions are. It cannot accidentally reveal one. Human writing regularly does. A slight bitter edge on a sentence, a clause that sounds like the writer actually cares — those register differently, and students can feel the difference once they know what to look for.

Practice these five before the activity. Post them as a checklist. That checklist becomes the analytical tool students use in Step 2.

How to run the comparison activity — 45-minute arc

The arc has four moves, timed tightly.

Step 1 — 2-sample reveal (5 minutes). Project two paragraphs on the same topic, side by side. Label them only “Sample A” and “Sample B.” Ask: which one did the AI write? Students write their vote and one-sentence gut-instinct reason. No discussion yet. The silence here is useful — students are reading closely because they have a stake in being right.

Step 2 — Group analysis with the 5-mechanics checklist (15 minutes). Small groups (3–4 students) annotate both samples using the checklist: sentence-length evenness, hedging phrases, generic transitions, vocabulary-without-specificity, emotional register. Each mechanic gets a yes/no vote per sample. Groups record their tally. This is where the compare AI and human writing classroom activity shifts from guessing to evidence-gathering — which is the move your admin needs to see in a lesson walkthrough.

Step 3 — Class reveal and discussion (5 minutes). Reveal which is which. Groups report their tally. Discuss: which mechanics were decisive? Did any mechanics appear in the human sample too? (They often do — that’s a productive complication, not a problem.) The point is not that AI writing is always detectable; it is that specific mechanics make it probable.

Step 4 — Students write, then judge (20 minutes). Each student writes one short paragraph on a prompt you assign (the same topic as the samples works well — familiar context, reduced cognitive load). Students swap paragraphs with a partner. Partner applies the 5-mechanics checklist. Not to label the paragraph as AI or human — to flag which mechanics are present. That flag-and-return becomes the feedback.

For the printable comparison pack — the two paired samples, the student checklist, the group tally sheet, and the Step 4 paragraph prompt — it’s ready to print by Monday morning with no prep beyond photocopying.

What to do with results — rubric + reflection

After Step 4, students rate each mechanic in their partner’s paragraph on a 1–4 scale: 1 = clear AI signature present, 4 = distinctly human. The rubric does two things at once. It gives students a concrete language for what they observed (“your transitions were generic — three ‘Furthermore’s in four sentences”), and it generates evidence of analytical thinking that a pure vote-for-A-or-B never could.

Close with one discussion question: “Where did AI fool you? Where did your gut catch it first?” Gut instinct before checklist versus checklist result is a rich comparison — some mechanics register intuitively (emotional register), others only appear when students slow down and count (hedging phrases).

The rubric-completion step is what earns CCSS.ELA-LITERACY.W.7.8 alignment — students gather information from multiple sources (two samples), assess credibility and accuracy (mechanic by mechanic), and document their findings. For the follow-up lesson where students build their own proof-of-writing workflow, the AI-Resistant Writing Pack continues directly from this foundation.

Differentiation — grade 6 vs grade 8

Grade 6: Provide the 5-mechanics checklist explicitly before students see any samples. Use sentence-level excerpts only — 2 to 3 sentences per sample, not full paragraphs. The reduced scope keeps the cognitive load on the analytical task rather than on reading stamina.

Grade 8: Remove the checklist scaffold entirely. Give full paragraphs and ask students to generate their own list of tells before the class share. Then compare student-generated lists against the 5-mechanics framework. The mismatch is instructive — students often catch emotional register and miss hedging phrases; that gap is worth a follow-up conversation.

The underlying reason AI writing has predictable mechanics is a useful grade-8 extension: AI4K12 Big Idea #3 — computers learn from data. AI writing systems learned by reading enormous volumes of human writing and predicting which word comes next. The mechanics that show up are the patterns that were statistically common in formal academic text. Once students understand that the writing reflects training data patterns, the checklist isn’t a parlor trick — it’s an application of how machine learning works.

For the scaffold that pairs well here — teaching students to write prompts that produce better, less robotic AI output — the CRAFT prompting framework post covers the writing-prompt side of the same skill set.

Standards crosswalk — print it, keep it handy

For the full AI literacy curriculum map, including where this activity fits in a unit arc, see the AI literacy teaching guide for grades 6–12.

• CCSS.ELA-LITERACY.RI.8.6 — determine an author’s point of view or purpose; distinguish what is directly stated from what is implied; analyze how the author distinguishes their position from others
• CCSS.ELA-LITERACY.W.7.8 — gather relevant information from multiple sources; assess credibility and accuracy of each source
• ISTE 1.3.b — evaluate accuracy, perspective, credibility, and relevance of information, media, data, or other resources
• AI4K12 Big Idea #3 — Learning (computers learn patterns from data — the reason AI writing mechanics are predictable, not random)

This activity double-dips. It covers ELA reading-analysis standards and AI literacy standards inside one 45-minute period. That is the real answer to the packed-ELA-curriculum problem: not a standalone AI day, but an activity where AI is the text type being analyzed under existing standards. Your admin sees CCSS. Your students get AI literacy. One prep.

For the companion lesson that covers verifying AI claims rather than detecting AI authorship, the hallucination worksheet post pairs well here — one covers spotting AI mechanics, the other covers verifying AI claims.

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