Generating Formative Assessments for Writing With AI

A formative assessment for writing has to work differently than a formative check in a fact-based subject. Instead of scoring one right answer, it evaluates a stage of an unfinished piece, a thesis statement, an outline, or a single paragraph's topic sentence and support, against specific writing-craft criteria, not a correct/incorrect key.

A 2015 meta-analysis by Steve Graham, Michael Hebert, and Karen Harris, published in The Elementary School Journal, reviewed decades of research on feedback, self-assessment, and peer assessment specifically inside writing instruction and found consistent positive effects on students' writing quality (Graham, Hebert & Harris, 2015). That's a different research thread than the one behind a formative quiz on rounding or vocabulary, and it changes what an AI-generated formative writing check should actually look like.

Quick Answer: A formative writing check can't just copy a "quick quiz" format, because writing rarely has a single correct answer for AI to detect. The stronger approach: target one stage of the writing process (a thesis, an outline, a topic sentence) against one named trait, such as the six-traits model (ideas, organization, voice, word choice, sentence fluency, conventions), grounded in writing-specific research rather than generic misconception-detection.

This guide covers:

• Why a formative check built for math or vocabulary doesn't transfer cleanly to writing
• How to anchor a formative writing check to a stage of the process instead of a finished product
• Using a trait-based framework, instead of a right/wrong key, to structure feedback
• What writing-specific formative assessment research actually supports
• A step-by-step AI workflow, a tools comparison, common mistakes, and FAQ

Why a Formative Writing Check Can't Just Copy a Formative Quiz

Formative assessment, broadly, means checking understanding during instruction so a teacher can adjust before a grade is ever assigned, a concept researchers Paul Black and Dylan Wiliam formalized in their influential review of classroom assessment research (Black & Wiliam, 1998). ASCD author James Popham later called this "assessment for learning" rather than "assessment of learning" (Popham, 2008). That principle holds for writing too. What changes is the mechanism.

The Quiz Mechanic Depends on a Discrete Right Answer

A formative check in a fact-based subject, rounding 47 to the nearest ten, naming a vocabulary word, identifying a fraction, works because there's one correct response. AI can compare a student's answer against that key instantly, sort wrong answers into misconception buckets, and hand a teacher a same-day dashboard.

Writing doesn't offer that shortcut. A thesis statement can be clear and arguable in several different valid ways, or vague in several different ways, and no single string comparison captures that difference.

What a Writing Check Needs Instead

A formative writing check needs criteria, not a key: does this paragraph have a clear topic sentence, does this lead avoid simply restating the prompt, does this transition actually connect two ideas. That's closer to rubric scoring than answer-matching.

• Fact-based subjects: one correct answer, instant match, misconception buckets
• Writing: a spectrum of quality against named criteria, judgment-based scoring

How to Create STEM Quizzes With AI covers formative and summative checks built around that first model, discrete answers a science or math class can score automatically. Writing needs a genuinely different structure, not a relabeled version of the same format.

Assess a Stage of the Process, Not a Finished Product

The strongest formative writing checks target one stage of an in-progress piece, a thesis, an outline, a single paragraph, rather than waiting for a finished essay. That single design choice is what separates a formative writing check from a summative one, and it's grounded in a specific strand of writing-instruction research, not just general assessment theory.

Why Process Matters More Than Product Here

Writing researcher Donald Graves argued that meaningful writing feedback happens during drafting, through short conferences on a work in progress, not only after a final draft is turned in (Graves, 1983). That writing-workshop tradition still shapes how writing teachers think about feedback timing today.

Steve Graham and Dolores Perin's Writing Next report, prepared for the Alliance for Excellent Education, named the "process writing approach," extended time for planning, drafting, and revising with feedback along the way, as one of eleven instructional elements with research support for improving adolescent writing (Graham & Perin, 2007).

The Common Core writing standards echo the same idea structurally. Standard W.5 asks students to "develop and strengthen writing as needed by planning, revising, editing, rewriting" (NGA Center & CCSSO, 2010), which only makes sense if a check-in happens somewhere inside that sequence, not only at the end of it.

Common Process-Stage Checkpoints

A teacher building AI-generated checkpoints across a grade band can pair this staged approach with age-appropriate scope; Building Upper Elementary Assessments With AI covers calibrating checkpoint difficulty and length to a specific grade band, which matters just as much for a thesis check as it does for a full test.

Keeping Each Checkpoint Genuinely Low-Stakes

A process-stage check only works as formative if it stays low-stakes. A thesis-statement check that gets a letter grade stops functioning like a check-in and starts functioning like a mini-summative test, which discourages the risk-taking early drafting requires.

• Keep each checkpoint short: one sentence, one paragraph, one outline
• Score it with a quick indicator (on-track / needs revision), not a full grade
• Return feedback the same class period or the next one, not days later

Using a Trait-Based Framework Instead of a Quiz Rubric

A quiz-style key can't score writing quality, so a formative writing check needs a named trait to score against instead, and the most widely used framework for that is the six-traits model. Built from teacher observation of what actually distinguishes strong from weak student writing, it gives a formative check something concrete to look for besides "is this good."

Where the Six-Traits Model Comes From

Teachers in Beaverton, Oregon, first documented the six traits in the 1980s, and the framework was later formalized by the Northwest Regional Educational Laboratory and popularized by writing-assessment author Vicki Spandel. It names six qualities that separate strong student writing from weak writing, independent of topic or genre.

Why One or Two Traits, Not All Six, at Once

A summative rubric scores all six traits on a finished piece. A formative check works better narrowed to one or two traits at a time, matched to whatever the current unit is teaching. A persuasive-writing unit focused on counterargument, for instance, formatively checks organization and ideas, not conventions.

That narrowing is also what makes a formative writing check faster to generate and faster to review than a full essay rubric. How to Generate 50 Quiz Questions in 5 Minutes with AI covers speeding up quiz creation for discrete-answer formats; a trait-narrowed writing check is the writing-specific equivalent of that same instinct, keep the check small enough to actually use daily.

What the Research Says About Feedback, Self-Assessment, and Peer Review

Formative writing assessment has its own dedicated research base, separate from general formative-assessment theory, and it specifically favors combining adult feedback with structured self-assessment and peer assessment rather than relying on any one source alone.

The Graham, Hebert, and Harris Findings

The 2015 meta-analysis referenced above examined formative writing assessment across several forms: feedback from teachers, feedback from peers, students assessing their own writing, and tests used to inform instructional decisions. Across the body of research reviewed, formative practices were associated with better writing quality than instruction without them (Graham, Hebert & Harris, 2015).

That's a meaningfully different evidence base than the one behind a formative quiz in another subject. It's specific to writing, and it explicitly includes peer and self-assessment as legitimate formative sources, not just teacher-generated feedback.

Self-Monitoring During a Draft

Researchers Karen Harris and Steve Graham developed Self-Regulated Strategy Development (SRSD), an approach that teaches students to monitor and check their own writing against a strategy's steps while drafting, rather than only after (Harris & Graham, various publications). A formative check that prompts a student to self-rate one trait before submitting a draft echoes that same self-monitoring habit.

The National Council of Teachers of English has similarly emphasized that effective writing feedback happens throughout drafting and revision, not solely at the end of a unit (NCTE, 2016).

Where General Formative-Assessment Principles Still Apply

Not everything about writing assessment is unique. Black and Wiliam's core formative-assessment principles, feedback that's immediate, specific, and actionable, still apply directly to writing (Black & Wiliam, 1998). What changes is the content of that feedback: a trait and a process stage, not a correct answer.

• Carries over from general formative assessment: speed, specificity, low stakes, immediate use
• Writing-specific: trait-based criteria, process-stage targeting, peer and self-assessment as core sources, not just extras

How to Build an AI-Generated Formative Writing Check-In

Building an effective AI-generated formative writing check means choosing a process stage and a trait first, then prompting for stage-appropriate criteria, not a full rubric. The sequence below keeps those decisions in the right order.

1. Pick one process stage the unit is currently working on, a thesis, an outline, or a single paragraph, rather than the whole essay.
2. Name one or two traits to focus on, matched to what the unit is teaching (organization for a structure-focused lesson, word choice for a revision-focused one).
3. Set the grade level and ability range so the AI's example and language match your class, not a generic grade-band default.
4. Ask AI for a short prompt plus lite scoring criteria: a one-line writing task, and two or three "look-fors" tied to the named trait, not a six-trait rubric.
5. Decide who reviews it first: the student (self-check), a partner (peer check), or the teacher, since each changes how much scaffolding the criteria need to spell out.
6. Set a same-day or next-day return window, since a formative check loses most of its value once feedback arrives days after the writing happened.
7. Decide the next action before handing it out: revise and resubmit, move to the next stage, or a short small-group conference for students who need it.
8. Repeat across the unit, tracking the same one or two traits at each stage so a student's growth on that trait is visible over time, not just a single snapshot.

You could use EduGenius's class profiles to hold a grade level and ability range, then generate a stage-specific prompt, a thesis-statement check for a persuasive-essay unit, for example, along with lite scoring criteria for one or two named traits, exported as a quick worksheet or a prompt to project. That's a workflow possibility worth testing against your own unit rather than a guaranteed formula.

For the broader landscape this workflow sits inside, AI Worksheet Generators: The Complete 2026 Guide covers how formative and summative generation fit together across subjects, writing included.

Tools and Technology for Formative Writing Assessment

No single tool automatically produces a good stage-specific, trait-based formative writing check; the difference comes from how much the prompt or platform lets a teacher target a stage and a trait rather than a full essay.

RAND's American Teacher Panel research has repeatedly flagged writing feedback and essay grading as among the most time-consuming parts of a teacher's week, a pressure that's part of why AI-assisted feedback tools have drawn growing classroom interest (RAND, 2023). EdSurge has reported similar rising teacher interest in AI-assisted writing feedback specifically, alongside caution about leaning on fully automated scoring for anything beyond a low-stakes formative check (EdSurge, 2024).

ISTE's guidance on classroom AI use has emphasized keeping a "human in the loop" for feedback that affects an actual grade, which is one more reason to keep formative writing checks low-stakes and teacher-reviewed rather than fully automated (ISTE, 2023).