Grade DBQs & Document-Based Essays

GradingPal is an AI grading assistant for teachers: upload student work and a rubric, and it drafts scores and specific, evidence-based feedback for you to review, edit, and release. In this use case, we'll follow an AP U.S. History DBQ body paragraph graded against a rubric built to mirror the College Board's own scoring categories.

The problem

AP DBQ writing has its own specific scoring logic, and it doesn't reward the same things a general essay rubric does. Citing a document isn't the same as sourcing it, explaining why its point of view or purpose matters to the argument. Mentioning outside knowledge isn't the same as citing a specific, dated historical fact distinct from anything already used for context. A student can write confidently and still miss the exact AP-specific moves that actually earn points.

Grading to that standard takes real fluency with the AP rubric itself, including its sometimes inconsistent decision-rule language, and applying it the same way across every student's paragraph. That's a lot to hold consistently across a full class, especially on a rubric with named performance levels rather than a simple point total.

This is where GradingPal helps. It grades against your AP-aligned rubric, distinguishes genuine sourcing from mere summary, and can even resolve an inconsistency inside the rubric itself by checking back against what the assignment actually asked for.

The assignment

Assignment

This 11th-grade AP assignment asks students to write a single, formal body paragraph evaluating why the Plains Indians' way of life ended between 1860 and 1900. Students have to use two documents from an earlier five-source packet, add one piece of specific outside evidence not found in the documents, open with a topic sentence tied to their thesis, and apply HIPP-style sourcing, historical situation, audience, purpose, or point of view, to the documents they choose.

It's a single, focused scaffold step inside a larger DBQ unit, deliberately built around the actual College Board scoring conventions for evidence and analysis rather than a generic classroom rubric.

The rubric

Rubric

The rubric mirrors the real AP DBQ scoring categories directly: Evidence From the Documents, Evidence Beyond the Documents, Analysis and Reasoning for Sourcing, and Analysis and Reasoning for Complexity, worth 5 points total. Each one carries the actual AP decision-rule language, requiring accurately describing a set number of documents to earn a given level, or explaining how a document's point of view is relevant rather than just naming it.

That's still your rubric, built to reflect the actual scoring standard students will face on the real exam, applied the same way to every paragraph in the class.

The graded submissions

The teacher uploads the student's paragraph, and GradingPal scores it against each AP criterion, applying the exact named performance level the response earns.

AI scores based on rubric and provides evidence

Detailed feedback for the student

Evidence that explains scores

Previous slideNext slide

1 / 3

One paragraph scores 5 out of 5, and along the way GradingPal catches a real inconsistency in the rubric's own wording, one level's title references a three-document requirement while its actual description and the assignment prompt both specify two. Rather than getting stuck on the contradiction, it defers to what the assignment itself asked for, confirms the student used and accurately described two documents, and even catches that the student mislabeled one of them.

On the outside evidence criterion, the feedback doesn't just confirm a fact was cited, it checks that it meets the AP standard specifically: a fact not detailed anywhere in the documents, given a specific timeframe, the 1870s, and a specific historical cause, post-Civil War railroad expansion. That's the difference between vague context and evidence that actually earns the point.

Even at a perfect score, the feedback coaches toward the habits that matter for the next paragraph: integrate a wider range of the provided documents, and make sure every piece of evidence gets its own brief sourcing analysis, not just a description of what it says. One comment even recommends a specific unused document by name as a next step. Every score and comment stays editable before anything is released.

Classwide analytics

Classroom performance overview

Strengths & weaknesses

Previous slideNext slide

1 / 2

Across the class, the mean sits at 58.8 percent, and the AI-written summary makes a genuinely AP-specific diagnosis: this class is already thinking like historians, building causal arguments about how violence, settlement, policy, and environmental change worked together, rather than just listing harms. The gap isn't historical understanding, it's the last layer of DBQ craft, using a wider range of documents, adding evidence that's truly separate from the documents, and explaining why a source matters instead of only what it says.

The criterion breakdown makes the gap precise: Evidence From the Documents sits at 97 percent, nearly universal strength, while Evidence Beyond the Documents lags at 71 percent, the clear relative weak point. The scores table shows this at the individual level too, several students score perfectly on document evidence and zero on outside evidence or sourcing, which is exactly what's capping their overall score even though half the rubric looks strong.

The outcome

Here's what changes when a DBQ paragraph runs through GradingPal:

The teacher gets scoring that actually applies the College Board's own categories and decision rules, including resolving ambiguity in the rubric itself, applied consistently to every student's paragraph.

The student gets coached in the exact AP vocabulary that earns points: your documents were accurate, now source them, and make sure your outside evidence is truly separate from anything already used. That's specific enough to change how the next paragraph is written.

And the class gets a diagnosis broken down by the exact AP scoring category that's capping scores, document evidence versus outside evidence versus sourcing, instead of one overall percentage.

That's the point of practicing DBQ writing this way in the first place. We don't just want students to understand the history. We want them to execute the specific craft that earns credit on the real exam, sourcing, genuinely separate evidence, complexity. GradingPal makes it practical to check for that craft, category by category, on every paragraph.