AI-Assisted Feedback Tool

Paste a student paragraph. Choose focus and level. Get structured teacher feedback - plus the exact CLEAR prompt to use with ChatGPT, Claude, or Gemini.

Student text

CEFR Level Essay / Writing Type Feedback Focus Feedback Tone

How this tool works

Based on Chapter 19The 2 strengths, 2 improvements, 1 guiding question model. Assess the process - not just the product.

Use the CLEAR promptCopy the generated CLEAR prompt into ChatGPT, Claude, or Gemini with the student text. AI asks questions - student revises. AI does not rewrite.

AI Collaboration LogAsk students to document any AI feedback they receive in their Collaboration Log. Download it from the Templates page.

How the demo above actually works

This page generates a sample of what structured AI feedback looks like, using a fixed bank of realistic comments for each focus area, so you can see the format before trying it live with a real AI tool. It does not call an AI model itself - it is a demonstration of the output shape (2 strengths, 2 improvements, 1 guiding question) that you should expect when you copy the generated CLEAR prompt into ChatGPT, Claude, or Gemini together with actual student writing.

Step-by-step workflow

Paste a student's paragraph or short essay into the box above and choose level, essay type, focus area, and tone.
Click Generate Feedback to see the demo output shape, then copy the CLEAR prompt it produces.
Open ChatGPT, Claude, or Gemini in a separate tab and paste the CLEAR prompt, replacing the placeholder with the actual student text.
Read the AI's response critically before sharing anything with the student - check that it actually followed the "2 strengths, 2 improvements, 1 question" structure and did not rewrite the student's sentences.
Share only the verified feedback with the student, and have them log the interaction in their AI Collaboration Log before revising.

Sample teacher prompt (typed directly, not using the generator)

"I am an EFL teacher assessing a B1 student's narrative paragraph. Here is the text: [paste]. Give exactly 2 strengths and 2 areas for improvement focused on sentence variety. For each improvement, ask one guiding question instead of rewriting the sentence. Do not correct grammar directly. Keep the tone encouraging."

Sample student-facing prompt

"I am a B1 EFL student. Here is my paragraph: [paste]. Do not rewrite it. Ask me two questions that help me notice what to improve, and tell me one thing I did well."

What a real AI feedback output typically looks like

A well-constructed response, whether generated by the demo or by a live AI tool, should always separate strengths from improvements clearly, name a specific pattern rather than a vague generality ("several sentences begin with I" rather than "vary your sentences"), and end with a question the student can act on, not a corrected version of their text. If an AI response includes a fully rewritten paragraph, that is a sign the prompt needs the explicit "do not rewrite" instruction repeated more forcefully.

Interpreting AI feedback: a teacher's checklist

Does the feedback name a specific, verifiable pattern in the actual text, or could it apply to almost any student paragraph? Generic feedback is a sign the AI did not engage closely with the specific text.
Does the feedback respect the level requested? Feedback written in B2 vocabulary for an A2 student is not actionable.
Has the AI invented anything about the student's intent or background that was not in the text? This is a hallucination risk even in feedback tasks, not just content generation.
Would acting on this feedback require the student to think, or could they simply copy a provided fix? If the latter, the feedback has crossed from coaching into ghostwriting.

Limitations of AI feedback

AI feedback can miss context a human teacher has: a student's prior drafts, their typical error patterns, or a recent change in effort or engagement. It evaluates the text in isolation.

AI feedback tools occasionally praise something that is not actually correct, or flag something that is actually a legitimate stylistic choice. Always treat AI feedback as a draft for the teacher to verify, not a final verdict.

AI feedback also cannot replace the relational dimension of teacher feedback - knowing when a student needs encouragement versus a harder push, or recognising when a pattern in writing reflects something happening outside the classroom. Use AI feedback to handle the repetitive, surface-level pass quickly, and reserve your own attention for the judgment calls only a teacher can make.

Ethical considerations, privacy, and academic integrity

Student data: avoid pasting full names or other identifying information into third-party AI tools. Use initials or anonymised reference numbers when sharing student text for feedback generation.
Consent and transparency: tell students and, where age-appropriate, parents, that AI tools may be used to help generate feedback on their writing. See Appendix O for parent communication letter templates.
Academic integrity in reverse: using AI to generate feedback is not the same category of action as a student using AI to generate their essay. The Chapter 19 AI Use Spectrum framework - scaffolding versus replacement - applies to teacher use of AI as well as student use.
AI detectors: as Chapter 19 notes, AI detection tools are unreliable and should not be the basis of an accusation against a student. Build process-based evidence (drafts, AI Collaboration Logs, oral defence) instead.

Best practices

Always read AI-generated feedback before it reaches a student - treat it as a draft, never a final product.
Keep feedback specific to the actual text, not generic advice that could apply to any paragraph.
Use a consistent prompt structure across a class so feedback is comparable and fair.
Pair every AI feedback pass with a fast, in-person check-in for at least your most vulnerable writers.
Rotate the focus area (structure, vocabulary, sentence variety, coherence, voice) across drafts rather than trying to fix everything in a single pass.

Frequently asked questions

Does this replace my own marking?

No. It accelerates the repetitive, surface-level pass so you can spend your own time on judgment calls - content quality, voice, and individual student context - that AI cannot assess reliably.

Can students use this tool directly?

Yes, with the student-facing prompt format above, which explicitly forbids rewriting and requires guiding questions instead - this keeps the revision work with the student.

What if the AI's feedback seems wrong?

Trust your own judgement over the AI's. Edit or discard any feedback point that misreads the text before sharing it with a student.

How does this connect to Chapter 21's multi-agent system?

This page demonstrates a single-pass, teacher-initiated feedback workflow. Chapter 21 and the Chapter 21 companion page describe a fully automated pipeline that runs this kind of feedback at scale, without per-essay teacher effort.

Troubleshooting

Problem	Likely cause	Fix
AI rewrites the student's paragraph despite instructions	The "do not rewrite" instruction was not explicit or forceful enough	Repeat the instruction in capital letters or as a separate final line: "FINAL REMINDER: do not rewrite any sentence."
Feedback is too generic	The prompt did not include enough of the actual student text, or the focus area was too broad	Paste the full text and narrow the focus area to one specific category at a time
Feedback vocabulary is above the student's level	Level was not specified, or the AI ignored it	Restate the level explicitly twice in the prompt, and add "use only vocabulary a [level] learner would understand"
Copy button does not work	Browser clipboard permissions blocked	Select the text manually and copy with Ctrl/Cmd+C