How to Use AI Conversations with Historical Figures in Your Classroom

The first time a class of 7th-graders opens a primary-source-grounded chat with the AI Frederick Douglass, two things happen at once. Half the room types a guarded version of “is this real?” and the other half asks a question the textbook never let them ask. Both reactions are appropriate to the moment. They are also the reason the pedagogy around AI historical-figure conversations needs to be done deliberately, not improvised. This piece is the playbook for the deliberate version: how to introduce the activity to students, what classroom moves work across grade bands, what to do when a chat goes sideways, and which guardrails matter more than the rest.

The audience is a middle school or high school social studies teacher who has access to a source-grounded AI history platform (Humy or equivalent) and wants to use it well, not loosely. If you are still upstream of that decision, our 2026 buyer’s guide for history teachers covers what to evaluate. This guide assumes you have a tool and need the classroom playbook.

The frame: the chat is rehearsal, not retrieval, and definitely not roleplay

The single most important framing move with students is naming what the activity is for, and naming what it is not. An AI historical-figure chat is not an information-lookup tool; the textbook and the primary source do that job. It is also not a role-play of the figure’s experience, and the distinction is one the discipline has been clear about for a long time. The IHRA Recommendations for Teaching and Learning about the Holocaust , updated in 2019 and adopted by all 35 IHRA member countries, are explicit: “empathetic exercises” that ask students to imagine themselves directly inside a historical experience are “often in poor taste and pedagogically flawed because it is impossible to really imagine what it would feel like to be in circumstances so far removed from our own life experience.” That guidance was written about Holocaust education, and it generalizes. The chat is rehearsal space for the disciplinary moves a historian actually makes, which the NCSS C3 Framework  names as sourcing, contextualization, corroboration, and close reading. The student is doing the thinking; the figure is the conversational partner who pushes the thinking forward.

Michael Calvert, the assistant superintendent at South Allegheny School District in Pennsylvania, puts  the value plainly when he says that “allowing them to chat with Lincoln will do that better than giving them a printed sheet of facts.” The point is not that the chat replaces the document; it is that the chat lets a student turn the document over and ask the figure why it says what it says. Roger Campbell, a 7th-grade World History teacher in Lancaster County, Pennsylvania, describes  the work he does before students touch the platform: he has them practice “good conversation techniques beforehand,” teaching them to “listen (or read) carefully and formulate thoughtful follow-up questions rather than just interrogating.” The chat does not replace that skill. It gives every student in the class a low-stakes place to practice it at the same time, which is the part the teacher cannot replicate solo across 28 desks.

Students who hear the framing once on day one usually take to it. Students who do not hear it at all treat the chat as a glorified search engine, which is the failure mode the rest of the playbook is built to avoid.

A four-phase classroom workflow

A workable workflow inside a 45-minute period runs in four phases.

The opening 10–15 minutes belong to the primary source itself. Students get Frederick Douglass’s 1852 “What to the Slave is the Fourth of July?”  speech, or an excerpt from the Narrative of the Life, or a contemporaneous newspaper account, on paper or in their LMS. They read, they annotate, they argue with each other about what the figure is saying. The chat is closed and the cognitive work of close reading lives with the student, in pencil if you can swing it.

For the next 15–20 minutes the chat opens. Students start a conversation with the figure (the AI Douglass for the example above) and practice three specific moves: asking about audience and purpose, asking why the figure used a particular phrase, asking what was happening politically when the document was written. Those are HIPP-style sourcing questions; the chat is the space to rehearse them. The teacher circulates, listens, and intervenes when a follow-up question would sharpen the inquiry. Melissa Vanicky, a teacher at Staten Island International School in New York, describes  what this looks like in practice: during her unit on the French Revolution, students asked Robespierre to reflect on his use of the death penalty during the Terror, and “Robespierre’s response sparked an interesting class debate about justice in today’s society.” The chat surfaced the historiographical question; the debate is where the discipline happened.

A short corroboration step, usually 8–10 minutes (or carried over to a second day for younger grades), introduces a second figure on a contrasting vantage point. For Douglass, that might be the AI Henry Clay on the doctrine of compromise. The exercise is to find where the two figures agree, where they diverge, and what evidence each cites.

The closing segment, whatever time remains in the period or assigned overnight, is the student’s writing. The chat is closed again. The student writes a short response (a thesis, a paragraph, an annotation) on their own. The discipline-defining output never travels through the model.

Jacob Chisom, a World and American History teacher in Monticello, Arkansas, reports that his students learn to “actively explore the past rather than passively consuming information” through this kind of structured sequence. The structure is the lesson, not the technology.

Grade-band variations

The playbook above bends to fit grade-band realities. In the upper elementary band (5th and 6th grade), the document is shorter, the chat is more guided, and the teacher’s framing is more present moment-to-moment; a unit on Mansa Musa or the Indus Valley might use a five-line excerpt rather than a full document, and the chat duration is closer to 10 minutes than 20.

Middle school (7th and 8th grade) is where the documents lengthen and the chat moves toward independent practice. This is the band Roger Campbell teaches in, and his structured pre-chat coaching (“teach them to listen carefully and formulate thoughtful follow-up questions”) is exactly right for the developmental moment. Students at this age have the metacognitive capacity to articulate what they are practicing, which means the chat-as-rehearsal frame lands.

By the time students reach high school and AP coursework, the workflow scales to support DBQ-style writing. The four phases above are, in effect, the DBQ scaffold we cover separately, and the chat becomes pre-writing rehearsal for the kind of argument the AP rubric measures.

What to do when a chat goes sideways

A few common failure modes and what to do about them.

When a student asks the figure something the platform restricts, the teacher’s framing controls should already have shaped what the figure will and will not engage with for the unit. The platform’s job is to produce a polite redirect rather than a generic refusal, and the teacher’s move is to use the moment as a teaching beat about historical context, not as a content-moderation incident.

The “just give me the answer” attempt is the second-most-common pattern, and it tells you the chat-as-rehearsal frame did not land for that student yet. The classroom move is to ask the student to explain, in their own words, what they actually want to know from the figure. Most of the time the question reframes itself once the student articulates it.

Occasionally, a chat surfaces a perspective the student finds upsetting. This is more common than it sounds, especially in units that touch atrocity (slavery, the Holocaust, residential schools, civil rights atrocities), and the teacher’s prior framing matters far more than anything the platform does in the moment. The right time to plan that framing is before the lesson, not after a student raises their hand.

The last pattern is a student testing the figure with a deliberately provocative question. The platform’s topic-scope controls should hold the line; if the figure responds in a way that surprises you, screenshot the exchange and send it to the vendor. A serious platform takes that feedback seriously and ships a fix.

Guardrails for sensitive history

Some units demand more than ordinary care, and the reference points are worth keeping a tab open for.

Holocaust units sit at the top of the list, both because the documentary record on AI failure modes is the best-developed (the UNESCO 2024 report AI and the Holocaust: Rewriting History? is the canonical reference) and because the IHRA Recommendations cited above give the discipline’s clearest published stance on what AI-assisted Holocaust pedagogy must avoid. The USHMM teaching materials  and Yad Vashem’s education resources  are the operational companions; both prioritize survivor voice and primary documentary evidence over any kind of imagined-perspective exercise. Our longer treatment of the procurement implications is in the ethics piece on AI historical figures, which spells out which figures should never be conversational characters at all.

Slavery and Reconstruction need similar care, anchored in the Library of Congress’s Teaching with Primary Sources  program and the National Museum of African American History and Culture’s education resources . For Indigenous histories, the Smithsonian’s Native Knowledge 360  materials are the discipline’s strongest current K-12 curriculum, with explicit emphasis on tribal voice and contemporary Indigenous scholarship. Civil rights units pair naturally with the SNCC Digital Gateway  and the King Center’s classroom resources .

In each case the same operational rule holds: the platform’s job is to give the teacher the controls; the teacher’s job is to know which units demand them and why, and to plan the framing before the lesson rather than after a student raises their hand.

What a teacher reports after a year of doing this well

When teachers have a year of this kind of structured AI-conversation use under their belt, the consistent reports are not about time savings; they are about specific changes in what students do when the AI is closed. Vanicky’s Robespierre debate is a concrete instance of the pattern: the chat surfaced a question, and the discipline-defining work happened in the room afterward, on the students’ own thinking. Paul Lepore, a social studies department chairperson who uses  the platform across his department, describes the same pattern at unit scale: through the platform, students “engage in inquiry-driven explorations, where they not only interact with historical personas but also unearth leads to primary and secondary documents.” The chat is the door; the documents are the room.

Behaviorally, three shifts show up in non-AI work after a semester of structured use. Follow-up questions get sharper across conversation contexts, including with the teacher and with each other. Primary-source citation gets more confident, because students who have spent time inside a document with a figure on the other end write about that document as if they actually know what is in it. And the argument paragraphs get less generic and more anchored, because the disciplinary moves were rehearsed before the writing began rather than improvised at the desk.

If you want to bring this playbook to a unit you are teaching this month, try Humy free  and run the four-phase workflow on one lesson with one section. The opinion you form on what students actually do with the activity is the right input to your next round of unit planning.