How to grade a DBQ with a rubric, with AI as first reader

The document-based question is the most time-consuming thing an AP^® History teacher grades. A class set of thirty essays, seven documents each, seven rubric points each: that often takes a full weekend. This guide walks through the rubric the way a College Board reader applies it, names the places students lose points most often, and shows how to let an AI grader take the first pass to cut your grading time in half.

The rubric below is the shared AP^® US History, AP^® World History: Modern, and AP^® European History DBQ rubric. The point names are the College Board's; the commentary is ours.

The seven points, in the order a reader checks them

1. Thesis (1 point)

The student makes a historically defensible claim that answers the prompt and establishes a line of reasoning. It has to be in the introduction or the conclusion, and it has to be more than a restatement of the prompt.

Where students lose it: "The Progressive Era changed America in political, social, and economic ways" answers nothing. A defensible thesis takes a position that someone could argue against: "Progressive reforms expanded federal power over business far more than they expanded democratic participation."

2. Contextualization (1 point)

The student describes a broader historical context relevant to the prompt. This is a few sentences of what was happening before or around the events in question, connected to the argument.

Where students lose it: A single phrase ("after the Civil War") is not contextualization. Readers are looking for a developed description, usually a short paragraph, that situates the prompt in time.

3 and 4. Evidence from the documents (up to 2 points)

One point for using the content of at least three documents to address the prompt. A second point for using at least four documents to support an argument. The difference is the difference between summarizing and arguing.

Where students lose it: Quoting a document without saying what it shows. "Document 3 says workers were unhappy" describes; "Document 3, a union pamphlet, shows that workers framed their demands in the language of citizenship, which supports the claim that…" argues.

5. Evidence beyond the documents (1 point)

One specific piece of historical evidence not in the documents, used to support the argument. It must be described, not just named.

Where students lose it: Name-dropping. "Also the Sherman Antitrust Act" earns nothing. "The Sherman Antitrust Act of 1890, which was used more often against unions than against trusts in its first decade, shows the same pattern of…" earns the point.

6. Sourcing (1 point)

For at least two documents, the student explains how the document's point of view, purpose, historical situation, or audience is relevant to the argument.

Where students lose it: Identifying the source without connecting it. "This document was written by a factory owner" is identification. "Because this was written by a factory owner for a Senate committee, its portrait of contented workers should be read as an argument against regulation, not as a description" is sourcing.

7. Complexity (1 point)

The essay demonstrates a complex understanding of the historical development that is the focus of the prompt. Readers award this for essays that explain nuance, consider multiple variables, qualify an argument, or make connections across periods.

Where students lose it: Almost always. The complexity point is the rarest on the exam. It usually comes from a paragraph that takes the counterargument seriously, or from a thesis that already contains a "but" or an "although." In the recent rubric change it can be earned by successfully using 6 of 7 documents or sourcing 4.

A grading routine that scales

Readers do not read a DBQ the way they would read a term paper. They check for points in roughly the order above, mark each one as earned or not, and move on. That routine is what makes it possible to score reliably at speed, and it is the routine an AI grader should follow too.

1. Score the thesis and contextualization first. Both are usually in the first paragraph. Decide, mark, move on.
2. Count document use. Tally which documents are cited and, for each, whether the use supports an argument or only describes.
3. Check for outside evidence and sourcing. These are the two points where a specific sentence either exists or does not.
4. Decide on complexity last, with the whole essay in mind.
5. Write one comment per missed point, not a general note. Students act on "your sourcing of Document 2 identifies the author but does not explain why that matters" far more than on "work on analysis."

Using AI as the first reader

Studiorum's auto-grading scores a DBQ against exactly this rubric, one point at a time, and writes a sentence on each. Here's how:

• Attach the official rubric to the question. For an AP^® US History, AP^® World, or AP^® Euro DBQ generated in Studiorum, it is attached already.
• Grade at College Board strictness. AP^® work is scored to reader standards. Everyday classwork uses a gentler reading, so make sure the activity is marked as an AP^®-style DBQ.
• Read the AI's reasoning, not just the score. Each point comes with the sentence that justifies it. When you disagree, change the point.
• Release grades when you are done. Students see nothing until you release, and released scores post back to your LMS gradebook if you use LTI.

Two rules keep the AI honest. First, treat its scores as a draft, especially the complexity point, which is a judgment call for human readers too. Second, spot-check the essays at the top and bottom of the distribution and adjust accordingly.

Before the next DBQ

Students improve fastest when they can see the rubric applied to their own writing. Return the per-point breakdown, then assign a revision that targets the missed points. In Studiorum, the mastery dashboard records each rubric criterion as evidence, so you can see which point a whole class is missing and generate targeted practice for it.

More on the tools behind this workflow: assessment generation and AI auto-grading.