Using Generative AI to Provide Targeted Writing Feedback

Overview

Providing students with timely, specific feedback enhances their learning by reinforcing strengths, clarifying misunderstandings, and guiding them toward deeper engagement with the writing process–all while supporting the development of a growth mindset. However, delivering high-quality feedback takes time, and many educators struggle to provide the depth and frequency of feedback they know would benefit students. Generative AI tools offer a way to expand these opportunities, giving students immediate, structured insights while allowing teachers to focus on deeper instructional support. When students receive feedback while they’re still working on an assignment or soon after, they can immediately reflect, revise, and refine their work, strengthening both their confidence and writing skills.

Teachers implementing AI-powered feedback tools have found success in tailoring them to their instructional needs, whether by customizing rubrics, refining AI grading accuracy, or integrating student self-assessments. AI can help scale feedback across a variety of writing-intensive tasks, from short responses to analytical essays, offering students multiple opportunities to improve their work while ensuring students still do the majority of the cognitive lift, presenting them the opportunity to intentionally develop their skills and knowledge.

Example From the School Teams AI Collaborative

Zach Kennelly, a Civics and AP Psychology teacher at DSST: College View in Denver and participant in the School Teams AI Collaborative, used AI to deepen writing feedback and student engagement (see the accompanying strategy card on generating AAQs more efficiently). He wanted to ensure his students were fully prepared for the Article Analysis Question (AAQ) on the AP Psychology exam, a task that requires analyzing research studies and interpreting statistical findings. He estimated that students needed to complete about 35 AAQs to build the necessary skills. To make feedback more immediate and actionable, he integrated Class Companion, an AI-powered feedback tool, into his instruction. Kennelly estimated that grading AAQs and providing actionable feedback, and then getting that data into an actionable format on each incorrect response would take him anywhere from 4-6 hours, but when he decided to use generative AI to support the process, his time spent grading decreased to less than 2 hours. Kennelly approached this by taking the following steps:

• Investing Time in Setup to Ensure High-Quality AI Feedback: Kennelly customized Class Companion to align with the AP exam’s scoring criteria, ensuring that students received feedback relevant to the expectations they would encounter on the test. Because the platform’s built-in rubric library included an AAQ-specific rubric, he only needed to adapt it versus building it from scratch. He also fine-tuned the AI’s grading logic to address inconsistencies, such as recognizing when students correctly identified a study — an adjustment that significantly improved the tool’s accuracy. The upfront time spent customizing the tool allowed him to provide consistent, high-quality feedback throughout the year, helping students refine their skills with each assignment.

• Encouraging Students to Engage With AI Feedback: Rather than treating AI-generated feedback as a final judgment, Kennelly encouraged students to critically assess their scores, just as they would if receiving feedback from a teacher. Students were empowered to challenge discrepancies, compare AI feedback to rubric expectations, and use their insights to refine their responses. This approach helped them develop a more active role in their learning process, strengthening their ability to self-assess, revise, and improve their writing. Additionally, Kennelly ensured another human element by personally reviewing all of the feedback students received so he could see the progress his students made, and override any feedback from the AI he believed was incorrect.

• Using Time Saved for Targeted Instruction: By automating the initial rounds of feedback, Kennelly recovered the capacity to provide more targeted support to his students. Instead of spending hours grading each response and providing feedback manually, he was able to pull students into small groups or conduct 1:1 check-ins focusing on specific areas where they needed additional instruction. Class Companion’s data insights helped him quickly identify patterns in student performance, allowing him to address common misconceptions and adjust his teaching accordingly.

• Scaffolding Supports to Increase Cognitive Lift: Kennelly recognized that students would not have access to immediate feedback during the AP exam, and he wanted to develop students’ writing skills to draft whole responses without the use of this tool. To build their writing skills, he scaffolded the supports. At the beginning of the school year, students received immediate feedback after each question, giving them time to revise their answers before progressing. As the year progressed, this step was removed and students received feedback only after submitting the entire assignment. This gradual release allowed students to develop the skills and knowledge to take on more cognitive lift without the support of AI. Explore this example of feedback students receive from Class Companion below.

Apply This Strategy in Your Context

AI-powered feedback tools can be applied across a variety of subjects and writing formats – from research essays to short response questions in math and science. Educators looking to implement AI feedback in their classrooms can take the following steps:

1. Identify a Writing Task: Find a task where frequent, timely, formative feedback could help students refine their work (e.g., research papers, letters to public officials).

2. Customize an AI Tool: Modify tools to align with learning objectives, ensuring AI-generated feedback is meaningful and relevant. Provide guardrails by uploading reference documents like rubrics or guidelines. ChatGPT, Gemini, Claude, and similar tools allow for uploading such documents. Direct these tools to refer to these documents by inputting prompt language such as “feedback should be aligned to the rubric document I have uploaded/copied and pasted below.”

3. Teach Students to Engage Critically With AI Feedback: Directly share with students that AI is not perfect, and while the feedback generated may often be helpful and accurate, there are times that the technology may not get it right. Rather than accept the feedback passively, encourage students to review feedback and to raise any disagreements or concerns with you. Explore this reflection protocol developed by another member of the School Teams AI Collaborative.

4. Use Time Saved on What Matters Most: Relationships drive success in teaching and learning, and time saved automating critical but time-intensive tasks can be used to double-down on relationship-building or personalizing instruction through small-group discussions or 1:1 coaching.

While Kennelly’s approach focused on AP-level writing, AI feedback can benefit students at all levels. Many AI tools offer free or trial-based versions, allowing teachers to experiment with the technology before making a full investment. Additionally, AI feedback can be scaffolded for students with diverse learning needs by pairing it with structured peer review or teacher check-ins.

By leveraging AI as a support tool rather than a replacement for teacher feedback, educators can create a more dynamic and responsive writing process, helping students build confidence, refine their ideas, and grow as independent thinkers.

This AI-enabled strategy was developed by a member of the School Teams AI Collaborative — a partnership between Leading Educators and The Learning Accelerator (TLA). The Collaborative was developed to bring together innovative educators from schools across the country to share ideas and discover effective ways to use AI in the classroom.