How to Use AI to Teach Reading Comprehension (Grades 6-8)

Tuesday’s text is picked. You’ve got six reading levels staring back at you, a handful of students who struggle with academic vocabulary, and a plan that calls for close reading of an informational article on climate migration. The article is grade-appropriate. The spread in your room is not. The instinct is to simplify — find an easier version, or ask AI to rewrite the passage at a lower Lexile. There’s a better move: keep the text hard, and let AI do your prep instead of the students’ reading.

TL;DR — how to use AI to teach reading comprehension in middle school: Keep the complex text. Use AI to build the scaffolding around it — question banks, vocabulary previews, discussion prompts, exit tickets. The student does the reading; AI does your prep. Every step below maps to CCSS.ELA-LITERACY.RI.7.1, CCSS.ELA-LITERACY.SL.7.1, and ISTE 1.3.d. A standards crosswalk at the end is ready to hand an administrator.

What does using AI for reading comprehension actually mean — without doing the reading for students?

The direct answer: AI for reading comprehension means AI does the teacher’s prep — generating questions, building vocabulary lists, drafting discussion prompts — while students do the reading and the thinking.

Literacy researcher Timothy Shanahan has written plainly about this: simplifying texts for struggling readers can actually slow reading growth, because the cognitive friction of a hard text — decoding unfamiliar syntax, working through dense paragraph structure — is precisely where reading skill is built. The struggle is not a bug in the lesson plan. It is the lesson plan. Shanahan’s argument is not that AI has no place in reading instruction. It is that using AI to take the complexity out of the text takes the reading out of the lesson.

The shift this post is built on: keep the text complex, and use AI to do the preparation that helps every student get into that text — the questions, the vocabulary scaffolds, the discussion prompts. AI generates those in minutes. The close reading stays with the student. For a related framework on how students can use AI as a research thinking partner rather than a shortcut, see how to teach students to use AI for research in middle school.

Prep the questions, not the text

The fastest move in AI-assisted reading instruction is using AI to generate a question bank for one text you already plan to teach. Here is the actual chain for a single informational article — not a description of the process, but the real steps.

The teacher’s prompt to the AI:

“I’m teaching a grade-7 ELA class. We’re reading a 600-word informational article about climate migration — people who move because drought or flooding has made their home region unlivable, written at about an 8th-grade Lexile level. Generate 10 comprehension questions: 3 recall questions that require locating specific details, 4 inference questions that require using two or more details together, and 3 synthesis questions that connect the article’s argument to a broader concept or a second source. Label each question by type.”

A realistic short AI output (excerpt):

Recall: What two geographic conditions does the article identify as primary drivers of climate migration? Inference: The article describes both economic and environmental pressures on migrants — what does the author’s word choice suggest about which pressure the author considers more significant? Synthesis: The article’s central claim is that climate migration is preventable with policy intervention. Do you agree, and what evidence from a second source would you need to support or challenge that claim?

What the teacher edits before handing it to students:

The synthesis question asks students to find a second source — a strong task, but only if you have already named which sources are available on this day. Edit it to be specific: “The article claims climate migration is preventable with policy. The UN Environment Programme’s 2023 Adaptation Gap Report argues adaptation funding falls far short. Using both, write one paragraph that agrees or disagrees with the article’s claim.” That edit takes about 90 seconds and turns a generic AI question into a citable task.

What the student does:

Students read the original article. They answer in writing, citing paragraph and line. The AI did not read the article for them — it generated the task structure. The reading and the evidence-gathering remain the student’s work.

Build a differentiated question bank for one text

A differentiated question bank — built on the same article, same reading — answers a 3-to-5-grade-level spread without handing out three different texts. Ask AI to generate questions across three tiers for one passage. CCSS.ELA-LITERACY.RL.6.1 (cite textual evidence) is the anchor across all three tiers; the cognitive demand changes, not the text.

Here is a worked table for the climate migration article:

Tier	Cognitive demand	Example question
Recall	Locate and restate	”Name two regions the article mentions as likely sources of climate migration by 2050.”
Inference	Connect two or more details	”The author describes both the economic cost and the human cost of climate migration. Which does the article treat as more urgent? Cite at least two details (CCSS.ELA-LITERACY.RL.6.1).”
Synthesis	Connect text to concept or source	”The article argues climate migration is a policy failure. Choose one of the author’s claims and explain how a reader who disagrees might respond. What evidence would that reader need?”

Students at the recall tier are still reading the same article. Students at the synthesis tier are still reading the same article. The differentiation lives in the question’s entry point, not in the complexity of the text — which preserves the cognitive work Shanahan describes as irreplaceable.

For a ready-to-print set of AI ELA leveled comprehension materials, the AI ELA Differentiated Worksheets cover 7 skills across 3 reading levels for grades 6-8.

Turn AI into your discussion partner

AI-generated discussion prompts, mapped to CCSS.ELA-LITERACY.SL.7.1 (collaborative discussion, grades 6-8), give a class a structured way into a text without having the AI summarize or analyze it for them. The student still has to have read the article to take part.

Ask AI for three prompts that require students to have read, not just listened. For the climate migration article:

• Position + evidence: “The article suggests wealthier countries bear some responsibility for climate migration from poorer regions. Do you agree? Cite one specific detail to support your position, then respond to someone who holds the opposite view.”
• Perspective shift: “The article is written from the perspective of a policy analyst. How might a climate migrant who has already relocated describe the same situation? What details from the article support your prediction?”
• Vocabulary + meaning: “The article uses the phrase ‘managed retreat.’ Using only the context in the article, write your best definition. Then compare as a class — whose is closest to what the paragraph actually says?”

The vocabulary prompt is deliberately last. It gives students who struggled with the article a concrete entry point without pre-teaching the term. They earn the definition through the discussion, not before it. For a full set of AI discussion protocols for literary and informational texts, see AI discussion partner for literary analysis in middle school.

A 3-step check for understanding you can run any day

The formative loop — exit ticket, pattern check, reteach decision — is where most reading lessons lose ground. A class produces 28 exit tickets; you read them the next morning and decide whether to reteach. AI can speed that read, with one hard rule: the teacher makes the call, not the AI.

• Step 1 — exit ticket. At the end of the period, students write 2-3 sentences answering one question tied to CCSS.ELA-LITERACY.RI.7.1 (cite textual evidence in informational text): “What is the article’s central claim, and what is one piece of evidence the author uses to support it? Quote the article directly.”
• Step 2 — AI pattern check. Paste 6-8 anonymized responses into an AI tool and ask: “These are student responses to a reading comprehension exit ticket. Which correctly identify a central claim AND cite specific evidence? Which identify a topic but not a claim? Which cite evidence but misread the claim? Sort into three groups and explain why.” The AI surfaces the pattern in about 90 seconds instead of 10 minutes.
• Step 3 — teacher makes the reteach call. Read the AI’s grouping against what you know about the class. If 60% misread the claim, the next day opens with a 5-minute whole-class return to the most claim-dense paragraph. If most cited evidence well but missed the claim, run a small group, not a whole-class reteach.

ISTE 1.3.b (evaluate accuracy, perspective, credibility and relevance) applies to your read of the AI’s grouping: you are checking whether the AI’s read of your students’ work is accurate before you act on it. The AI flagged the pattern. The teacher read the students. The teacher made the decision.

The standards crosswalk (send this to your admin)

When an administrator asks how AI-assisted reading comprehension maps to standards, this table is the answer:

Lesson step	ISTE code	CCSS code
Close-read the complex text + cite evidence	ISTE 1.3.d (Knowledge Constructor — actively explore real-world issues)	CCSS.ELA-LITERACY.RI.7.1
Differentiated question bank — recall / inference / synthesis	ISTE 1.3.d	CCSS.ELA-LITERACY.RL.6.1
AI-generated discussion prompts, student-led discussion	ISTE 1.3.b (evaluate accuracy, perspective, credibility and relevance)	CCSS.ELA-LITERACY.SL.7.1
Check AI’s grouping of exit-ticket responses for accuracy	ISTE 1.3.b	CCSS.ELA-LITERACY.RI.7.1

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.

The design principle is the one this post opened with: you can’t out-detect or out-ban AI in a reading classroom, and you shouldn’t simplify texts to dodge the struggle — you out-design the lesson so the reading, the student’s cognitive work, stays with the student and in front of you, while AI handles your prep. That is the same stance Maya, a former teacher, built the catalog on.

For a ready-to-teach set of AI-generated leveled reading passages and differentiated comprehension materials for grades 6-8, the AI Reading Literacy Pack covers 10 modules across middle and high school reading levels. Browse the full catalog at the shop.

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