LAW · AI FEEDBACK
Best AI Feedback Tool for Law Essays: What Good Legal Feedback Looks Like and Which Tools Deliver It
The best AI feedback tool for law essays is the one that tells a student where their legal reasoning fails, not where their sentences could be smoother. Here is what that feedback looks like, how the main tools compared on a real law essay, and which one fits which job.
By Eduface · September 2026 · 12 min read
A student sends you a draft of their judicial review essay two weeks before the deadline. You have forty more like it and time to comment properly on perhaps ten. So most students get two lines, and the two lines say the same thing you wrote last year: “engage more critically with the authorities”. The student does not know what that means for their essay. They submit, and the final mark tells them what the feedback should have.
What is the best AI feedback tool for law essays?
The best AI feedback tool for law essays is one that comments on legal reasoning (issue identification, use of authority, application to the facts, engagement with counter-argument) rather than on language alone, uses your rubric, and lets the lecturer review feedback before the student sees it. In our test on a real law essay, general AI assistants gave fluent but flattering feedback. Eduface, a purpose-built grader with a Law model, flagged the same weaknesses the lecturer did.
What does good feedback on a law essay look like?
Good feedback on a law essay points at the legal reasoning and tells the student what to do next. Hattie and Timperley’s review of feedback research found it among the strongest influences on learning, but only when it answers three questions: where am I going, how am I going, and where to next.¹ Law feedback often answers the first two and skips the third.
For a law essay, useful feedback is:
Specific to a passage. “Paragraph 4 states the rule in Wednesbury but does not apply it to the council’s decision” is usable. “More application needed” is not.
Tied to a criterion. The student should see which part of the rubric a comment belongs to, so they can see what it costs them.
About the law, not only the language. Comments on grammar and flow are helpful. They are not what separates a 2:2 from a 2:1 in law.
Timely. Feedback on a draft is worth far more than feedback on a final submission. In the 2025 National Student Survey, 83.3% of respondents were positive about the timeliness of feedback.² That still leaves a sixth of students who were not, and the draft stage is where the difference shows.
To make this concrete, here is one weak paragraph from a public law essay and three kinds of feedback a tool could give it.
Example paragraph (illustrative)
“The decision of the council was unreasonable. In Associated Provincial Picture Houses v Wednesbury Corporation the court held that a decision can be challenged if it is so unreasonable that no reasonable authority could have made it. Therefore the council’s decision can be challenged and the claimant will succeed.”
Feedback type
What the student is told
Is it useful?
Language only
“Consider varying your sentence openings. ‘Therefore’ could be replaced by a stronger transition.”
No. The paragraph’s problem is legal, not stylistic.
Generic
“Good use of case law. Try to be more critical and develop your analysis further.”
Barely. The student cannot act on “be more critical”.
Legal reasoning
“You state the Wednesbury test correctly, but you jump from the test to the conclusion. What exactly did the council decide, and why would no reasonable authority have decided it? The threshold is high. Address why this decision meets it, and consider whether a proportionality argument would be stronger on these facts.”
Yes. It names the missing step, the legal standard and a next move.
Table 1: Three kinds of feedback on the same paragraph. The example is written for this article.
The third kind is what a good law tutor writes. It is also what most AI tools do not write unless they were built for assessment.
How did the main AI tools compare on a real law essay?
We tested this directly. As part of our guide to AI grading tools, two students in the Netherlands ran eight tools against papers with a known lecturer grade, the lecturer’s written feedback and the official rubric. One paper was a Dutch-language constitutional law essay graded 4.4, a fail. Its real problems, as the lecturer identified them: a research question restated six times, sources cited but not applied to the legal criteria, a missing bridging sentence between two parts of the argument, several typos, and a conclusion that did not reach a legal determination.
Here is what each tool’s feedback did with those problems.
Tool
Grade given (lecturer: 4.4)
What the feedback caught
What it missed or got wrong
Eduface
5.5
The typos, the weak link between sources and argument, the missing bridging sentence, the non-committal conclusion
Still graded above the lecturer, by 1.1 points
Claude
7.2
The missing bridging sentence, inconsistent heading numbering, the conclusion not answering the research question
Saw the problems, then weighted them far too lightly
Gemini
6.8
The research question stated six times; called the conclusion “an open-ended evasion rather than a legal determination”
Refused to commit to a grade in most runs; output changed between runs
ChatGPT
7.1
Readable, well-organised comments
Praised the work generally; responded to how well the student writes, not how well they reason
CoGrader
10.0
Nothing
Called the paper an “excellent structure that flows seamlessly from your creative concept into a professional legal analysis”
Source: Eduface independent student test, July 2026. One law paper, so read it as directional evidence.
Tool
Question repeated
Sources not applied
Bridging sentence
Typos
No determination
Eduface
–
Yes
Yes
Yes
Yes
Claude
–
–
Yes
–
Yes
Gemini
Yes
–
–
–
Yes
ChatGPT
–
–
–
–
–
CoGrader
–
–
–
–
–
Figure 1: How closely each tool’s feedback matched the lecturer’s own criticism of the same law essay.
The pattern matters more than the scores. Claude and Gemini both made individual observations a law tutor would recognise. But feedback that notices a problem and then calls the essay a 7 sends the student a mixed message they will resolve in their own favour. Feedback and grade have to agree.
Why do general AI assistants give weak feedback on law essays?
General assistants like ChatGPT, Claude, Gemini and Copilot were trained on general text, not on how law is assessed. Three things follow from that.
They reward polish. A model trained on general text responds to linguistic quality. A model trained on academic assessment responds to disciplinary argument quality. In law, those two diverge constantly.
They do not weight criteria like a law marker. A structural weakness that costs 0.3 points in a general model’s weighting can cost two full points with a university assessor. A missing application step is not a minor flaw in law. It is the assessment.
They have no workflow. Every session means pasting the rubric again, there is no lecturer review before the student sees anything, and student work goes to US servers under default settings. For formative feedback in a law school, that last point is a GDPR question, not a technical detail.
There is also a writing-assistant category, tools like Grammarly, which work on clarity, grammar and tone. They are useful for a final polish. They are not built to tell a student whether their application of Wednesbury is convincing. We looked at this in our Grammarly for Education review.
What should you look for in an AI feedback tool for law?
Use these six criteria when you compare tools. They come from what law feedback needs to do, not from feature lists.
Criterion
What good looks like
Red flag
Legal reasoning
Comments on issues, rules, application, counter-argument and conclusion
Comments are mostly about grammar and flow
Rubric-grounded
Each comment is linked to a criterion in your rubric
One block of general praise and advice
Passage-level
Annotations point at the exact sentence or paragraph
Feedback refers to “the essay” as a whole
Lecturer in control
The lecturer can edit and approve before release, or choose to release directly for low-stakes drafts
Feedback goes straight to the student with no option to review
Consistent
The same essay gets the same feedback on a second run
Different advice every time
Data-safe
Processed in the UK or EU under a DPA, not used for training
Student work sent to a consumer chatbot
How does Eduface give feedback on law essays?
Eduface’s Paper Grader gives formative feedback on drafts and full rubric-based grading on final submissions, in one tool. For a law module, the lecturer selects the Law model, one of six discipline models, and sets three things: the assignment brief, the rubric, and instructions for how feedback should be phrased.
Draft by draft. Students can get feedback on a first draft, a second draft and the final version. Eduface tracks how each student’s work develops across those rounds, so feedback on a second draft can say where the student has improved since the first.
Four feedback styles. The lecturer picks the voice of the feedback:
Reflective and Socratic: asks the student questions that lead them to the gap. This suits law teaching well, where the Socratic method has a long history.
Constructive and Direct: names the problem and the fix.
Went Well and Needs Improvement: a two-column structure students find easy to scan.
Supportive and Encouraging: for first-year students and formative work early in a module.
Annotations in the text. Every comment is anchored to the passage it is about and linked to a rubric criterion.
The lecturer decides how much oversight. For summative work, no grade reaches a student without lecturer approval. For low-stakes drafts, the institution can choose to let feedback go directly to students, or only after a lecturer has read and approved it. Approved feedback carries a “Lecturer + AI” label.
1
Student submits draft 1
2
Eduface annotates and comments per criterion
3
Lecturer reviews (or releases directly for drafts)
4
Student revises
5
Draft 2 with feedback that references the progress since draft 1
6
Final submission graded against the rubric
7
Lecturer approves the mark
Formative feedback and the final grade run through the same rubric, so what the student hears on the draft matches how the final version is marked.
What Eduface doesn’t do
It does not read handwritten exam scripts (Gradescope is stronger there), it does not offer a library of ready-made law rubrics, and it is not a citation formatter. If your main need is checking OSCOLA formatting line by line, use your library’s referencing guidance alongside it.
Which tool is best for which job?
“Best” depends on what you need the feedback for. Here is our honest recommendation.
If you need…
Best fit
Why
Feedback on legal reasoning in drafts, for a whole cohort
Eduface Paper Grader
Law model, rubric-grounded, lecturer review, draft-to-draft tracking
A quick language polish before submission
A writing assistant such as Grammarly
Built for clarity and grammar, not for legal argument
A rough second opinion when no dedicated tool is available
Microsoft Copilot
Best accuracy of the general tools in our test (±0.7 on psychology papers), but generic feedback
Brainstorming counter-arguments or reading lists
ChatGPT or Claude
Useful for ideas, not for judging the quality of an essay
Short essays in school settings
EssayGrader.ai or CoGrader
Built for K-12; on university law, CoGrader gave a failing essay 10/10
How do you introduce AI feedback in a law module?
Start with a draft deadline. AI feedback is most valuable before the final submission. Add a formal draft point three weeks before the deadline if the module does not have one.
Tell students what the feedback is and isn’t. It is a structured reading of their draft against your rubric, reviewed by you. It is not a predicted grade.
Calibrate on last year’s essays. Run five essays you have already marked and compare the feedback with your own comments. Adjust the feedback instructions until the tone and focus match what you would write.
Watch the second draft. The real test is whether students who received feedback on draft 1 improve on the criteria the feedback addressed. Eduface’s draft-to-draft tracking shows you that per student.
If you are choosing grading software for a whole law school, read AI grading software for law school assignments. If your students ask whether they can check their own essays with AI, send them this guide for law students. For more on keeping formative and summative uses separate, see formative vs summative assessment with AI.
Frequently asked questions
Can ChatGPT give good feedback on a law essay?
It can give fluent, readable comments, and it is useful for brainstorming. But in our test it graded a failing constitutional law essay at 7.1 against a lecturer’s 4.4 and praised the work rather than identifying its legal weaknesses. It also processes student work on US servers by default, which is a GDPR issue for European institutions.
What makes feedback on a law essay useful to the student?
It points at a specific passage, names the legal step that is missing or weak (usually application to the facts or engagement with counter-argument), links it to a rubric criterion and tells the student what to do next. Comments on language alone rarely change a law mark.
Should AI feedback on law drafts go straight to students?
That is the institution’s choice. For low-stakes drafts, many lecturers are comfortable releasing feedback directly. For anything that carries a mark, the lecturer should review first. Eduface supports both, and never releases a grade without approval.
Does AI feedback work for law essays in languages other than English?
It depends heavily on the tool. On a Dutch-language constitutional law essay, general tools were 2.4 to 2.8 points too generous and CoGrader gave full marks. Eduface returned 5.5 against the lecturer’s 4.4 and identified the specific errors. Test any tool in the language you actually assess in.
Does Eduface know OSCOLA or IRAC?
Eduface scores against the rubric you give it, with a Law model trained on the conventions of legal writing. If your rubric rewards a clear issue-rule-application-conclusion structure or correct use of authority, feedback will address those criteria. For line-by-line citation formatting, use your library’s OSCOLA guidance.
References
1. Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112. [Feedback is among the most powerful influences on learning when it answers where the learner is going, how they are going and where to next.]
2. Office for Students. (2025). National Student Survey 2025. [83.3% of respondents positive about the timeliness of feedback; 83.5% felt marking and assessment had been fair.]
3. Eduface. (2026). The Complete Guide to AI Grading Tools for Higher Education. [Independent student test of eight tools on six papers with known lecturer grades, including a Dutch-language constitutional law essay graded 4.4.]
See it on your own assignments
Eduface drafts criterion-level grades and feedback for your lecturers to approve. Book a demo or start free.