Last updated October 4, 2026.
An AI rubric generator turns an assignment description into a criteria grid you can edit, and the better ones also apply that grid to student work. Six are worth using, and they split by who is doing the grading. MagicSchool drafts a standards-aligned grid fastest. Brisk Teaching builds one inside the Google Doc you already have open. EnlightenAI applies the rubric you already wrote to student writing. GradingPal reads handwritten and scanned work. Knowt gives you a free grid with students in the loop. Gradescope is the one built for college and Advanced Placement grading at scale.
Vendors sell the same thing as a rubric generator, a rubric maker and a rubric creator, and the three words mean one tool. What they hide is the distinction that actually decides your choice: whether the tool follows the rubric you wrote or quietly writes its own. Three of these six accept your criteria. Three draft their own and leave you to edit them.
Search for any of those three names and almost every result is a vendor recommending itself. MagicSchool's page says MagicSchool. Knowt's page says Knowt. None of them tells you where the free tier ends, whether the tool actually follows the rubric you wrote, or what happens when you grade 140 essays instead of 14.
AI rubric generator, AI rubric maker, rubric creator: one tool, three names
The category is searched under all three names and sold under all three, sometimes by the same vendor on the same page. Knowt's rubric tool is headed "Free AI Rubric Maker," describes itself in the next line as a "free AI rubric creator for teachers," and answers the cost question with "our AI rubric generator is completely free" (Knowt rubric generator). Three names, one product, no difference.
So the name on the button tells you nothing. The thing that does is rubric control: whether the tool generates criteria, accepts yours, or holds yours still while you grade. That is the next section, and it is the only distinction in this category worth shortlisting on.
One practical consequence if you are searching rather than reading. A tool that only appears under "maker" or "creator" is not a different class of product, so widen the search rather than assuming you have found a gap. And a tool that calls itself a generator is still, in almost every case, generating criteria rather than applying yours.
How this list was built
The pool is the assessment and grading category of the directory at aieducator.tools, which held 53 tools when it was checked on September 15, 2026. From those, and from the wider directory, this list keeps only tools that genuinely build rubrics or apply them to student work, rather than mentioning rubrics somewhere in the marketing. The test for inclusion is fit for the job, evidence of classroom use, safety posture, honest pricing, and whether the tool leaves the judgment with the teacher. Nobody pays for inclusion in the directory and nobody paid for a place on this list.
Three exclusions worth naming. chalkd.Ai is a capable marking tool, the British term for grading software, but it is built for scoring exam papers against mark schemes and its record carries an 18+ minimum age, so it is not a rubric builder for a K-12 class. EduArchive AI generates rubrics and launched in 2026 with no reviews in the directory yet, so there is nothing to weigh. CoGrader, RubricAI and EssayGrader rank well for this query and are not in the directory at all, which means their compliance records cannot be checked here, and an unverified claim is not a recommendation.
Pricing and feature claims below were checked against live vendor pages on October 4, 2026. Directory ratings are quoted as the directory holds them.
What the directory data shows, and what changed since August
Of the 53 tools in the assessment and grading category, 13 carry an independent 9ine security assessment, which is about one in four. Nineteen declare FERPA compliance and 15 declare COPPA. Thirty say they collect student data, 10 say they do not, and 13 leave the field blank. On hosting, 12 record the United States, 11 the United Kingdom, 10 the European Union, two Australia and one Canada, while 16 do not say where student data lives at all.
That last number is the one to carry into a procurement conversation. A rubric generator becomes a grading tool the moment student work goes into it, and at that point the blank field is the problem, not the answer.
Two figures in this article moved between its first publication on August 29, 2026 and this update, and both moved the same way.
GradingPal halved its free tier. This article recorded 100 submissions a month on August 29. The pricing page now states 50 submissions a month. Its paid tiers were also restructured: Lite is now $14 a month or $168 a year and Pro is $26 a month or $312 a year, replacing the earlier monthly-versus-annual split.
Knowt's teacher pricing rose and the free plan is narrower than it looks. Teacher Ultra is now $19.99 a month or $120.00 a year (Knowt help center), against the $12.49 a month billed annually recorded in August. The same help page states that free teacher accounts allow one class and have no access to AI features, while Knowt's rubric generator page states that the generator itself "is completely free." Both can be true, because the standalone generator sits outside the class features, but a teacher reading only the second page will be surprised by the first. Check which one you are relying on before you plan around it.
Re-checked October 4, 2026: both figures still stand, and nothing else moved. Every price and free-tier limit in this article was verified again against the live vendor pages on October 4, 2026. GradingPal still lists 50 submissions a month free, Lite at $14 a month or $168 a year with 250 submissions, and Pro at $26 a month or $312 a year with unlimited submissions (pricing). Knowt's Teacher Ultra is still $19.99 a month or $120.00 a year (help center). MagicSchool Plus is still $8.33 per user per month billed annually, which the pricing page gives as $99.96 a year, or $12.99 billed monthly (pricing). Brisk Teaching still runs free for individual teachers with 20+ tools against 35+ on paid plans, with no individual paid tier and custom data privacy agreements on paid plans only (plans). Gradescope Basic is still free and the institutional plan still carries no public price (pricing).
One correction made in this update. An earlier version of this article said Knowt's free teacher accounts carry ads. The help center sets out the free-account limits as one class and no access to AI features and does not support the ads claim, so it has been removed.
Free tiers move quietly and in one direction. Any figure in an article like this one is a snapshot, including these.
Does the tool follow your rubric, or write its own?
This is the question the category is actually being asked, and most comparisons skip it. There is a real difference between a tool that drafts a rubric for you and a tool that takes the rubric you already wrote and applies it faithfully to 120 pieces of work. The second is harder, and it is where tools separate.
Three levels are worth telling apart.
It generates a rubric. You describe the assignment and the tool returns a grid you then edit. MagicSchool, Brisk Teaching and Knowt all do this. The grid is a draft, and the criteria are the machine's until you change them.
It accepts your rubric. You paste in or upload the rubric your department already agreed, and the tool scores against those criteria rather than inventing its own. EnlightenAI states that it "learns your rubric and writes feedback in your voice," and GradingPal offers a choice of AI-generated or custom rubrics on every paid tier and on its free plan.
It holds the rubric still while you grade. This is the one that matters at scale and the one only Gradescope does. In Gradescope, the instructor defines each rubric item and its point value, and can build the rubric before submissions arrive or during grading (Gradescope guides). Edits to a rubric item propagate through the grading interface, so a criterion you rewrite in week two does not leave week one graded against a different standard.
The practical test, whichever tool you pick: generate or upload the rubric, grade five pieces of work, then read the scores against the criteria yourself. If the tool has quietly rewarded something your rubric does not mention, you have found out at a cost of five pieces rather than 120.
The six, compared
| Tool | Best use | Rubric control | Strengths | Limitations | Price / free tier | Privacy considerations |
|---|---|---|---|---|---|---|
| MagicSchool | Drafting a standards-aligned grid fast | Generates, you edit | 80+ teacher tools on one account; converts existing rubrics to other formats | Free plan caps generations; rubric is a draft until you change the criteria | Free $0; Plus $8.33/user/mo billed annually or $12.99/mo | Carries a 9ine assessment; hosting and minimum age not stated in our record |
| Brisk Teaching | Building a rubric inside a Google Doc | Generates, in place | Free with 20+ tools; runs over Docs, Slides and your LMS | Browser extension only; no individual paid tier | Free $0; school and district plans quote-only | Carries a 9ine assessment; custom data agreements are a paid feature |
| EnlightenAI | Applying your rubric to student writing | Accepts yours | Teacher approves, edits or overrides every score before release | Vendor states grades 3-12, so not a higher-ed tool; no reviews in our record | Free for teachers grades 3-12; paid tiers quote-only | Carries a 9ine assessment; UK hosting, minimum age 8+ |
| GradingPal | Handwritten and scanned work | AI-generated or yours | OCR reads handwriting; standards-aligned rubric library | No independent security assessment; free tier halved since August | Free 50 submissions/mo; Lite $14/mo or $168/yr; Pro $26/mo or $312/yr | US hosting, no minimum age, no 9ine assessment; GDPR status conflicts between our record and the vendor |
| Knowt | A free grid with students in the loop | Generates, you edit | Rubric generator free; connects to flashcards and practice | Free teacher accounts have no AI features and allow one class | Generator free; Teacher Ultra $19.99/mo or $120/yr | Carries a 9ine assessment; US hosting, no minimum age |
| Gradescope | College and AP grading at scale | Holds yours still | Dynamic rubrics; AI groups similar answers for instructor review | Advanced rubric features and AI grading are institutional-only | Basic free; Institutional quote-only, no public price | Carries a 9ine assessment; hosting and minimum age not stated in our record |
The privacy column records what each vendor declares or what our directory holds. It is not an audit and it is not a compliance finding.
Rubric grading for college and AP instructors
If you teach a 200-student lecture course or an AP section, most of the tools above were not built for you, and the mismatch is worth naming before you trial one. EnlightenAI's own site states it is free for teachers across grades 3-12. MagicSchool, Brisk and Knowt are pitched at K-12 classrooms and their standards pickers assume state standards rather than a syllabus. None of that makes them bad; it makes them the wrong shape for a course where 12 teaching assistants grade the same problem set and every score has to survive a regrade request.
One directory discrepancy is worth flagging here, because it would mislead anyone shortlisting from the record alone: EnlightenAI's directory entry lists Higher Education / University among its year groups, while the vendor's own grading tool page states grades 3-12. Where a directory field and a vendor page disagree, the vendor page is the one to act on, and the record needs correcting.
Gradescope is the best fit for rubric grading in higher education and AP courses because its rubric is the grading interface rather than a document beside it. The instructor writes each rubric item with its point value, then grades question by question across the whole class instead of paper by paper. Its AI groups identical and near-identical answers so a rubric item can be applied to a whole group at once, and every suggested group has to be reviewed and confirmed by the instructor before any score lands. Individual grades stay adjustable afterward.
Watch out for: the free plan is narrower than the trial that precedes it. Gradescope Basic remains free after an institutional trial ends, and keeps dynamic rubrics, question-by-question grading and basic PDF assignments. What it does not keep is collaborative grading by multiple staff, AI-assisted grading and roster matching, anonymous grading, section management, programming and bubble-sheet assignments, advanced rubric functions such as item groups and rubric import, and LMS and single sign-on integration (Gradescope pricing). For a solo instructor that is a workable free tool. For a course with a grading team, the features you came for are the institutional ones.
Pricing: Gradescope Basic is free; Gradescope Institutional is quoted directly with no public price (pricing, checked October 4, 2026). Note that Gradescope is now a Turnitin product and its pages sit on turnitin.gradescope.com; Turnitin is listed separately in the directory and holds 3.20 from 5 educator reviews.
Safety: Gradescope carries an independent 9ine assessment in our record. Hosting location, minimum age and FERPA and COPPA status are all blank in that record, which for a tool this widely used in US universities means the information lives in institutional agreements rather than in a directory field. Your registrar or provost's office will have it, or will need to ask for it.
Two others are worth a line for this audience. Class Companion runs AP-level assignments and AP exam preparation with a free tier for individual teachers, and carries a 9ine assessment and 4.00 from a single review, which is too thin to rank. Turnitin remains the default where academic integrity rather than rubric fidelity is the driver.
MagicSchool: best for drafting a standards-aligned grid fast
MagicSchool is the best fit for the teacher who wants a usable rubric in under a minute because its rubric generator takes an assignment, a grade level and a standard and returns an editable grid, with a companion tool that converts a rubric you already have into a different format or point scale. It sits inside a platform of 80+ teacher tools sharing one account, so the rubric feeds into quiz generation, feedback and IEP drafting without a second login. Our directory holds a rating of 4.74 from 19 educator reviews, the largest review count of any rubric tool here apart from Brisk.
Watch out for: the criteria are the machine's until you change them, and the free plan meters how often you can ask. Unlimited generations, unlimited output history and unlimited AI-output editing are all listed as Plus features, which means the free tier caps all three (pricing). A teacher who drafts a rubric, dislikes it and redrafts twice will meet that ceiling in a single planning period.
Pricing: free at $0 with 80+ teacher tools, the Raina chatbot, quizzes and class writing feedback; Plus at $8.33 per user per month billed annually, which the pricing page gives as $99.96 a year, or $12.99 billed monthly; Enterprise quoted directly (pricing, checked October 4, 2026).
Safety: carries an independent 9ine assessment in our directory record, which is more outside scrutiny than most tools in this category have. That record does not state a hosting location or a minimum age. The vendor publishes a privacy policy and terms of service, and what your district can actually agree to sits in those documents and in the agreement it signs.
Brisk Teaching: best for building a rubric inside a Google Doc
Brisk Teaching is the best fit for the teacher who wants the rubric built where the assignment already lives, because it runs as a Chrome and Edge extension over Google Docs, Slides and learning management systems rather than as a separate site. Open the assignment document, describe the task, and the rubric is drafted in place, then powers Brisk's feedback tools against the same criteria. It is the most-reviewed tool in this comparison: 4.77 from 35 educator reviews, with an independent 9ine assessment.
Watch out for: the extension is the product, so a district that blocks browser extensions blocks Brisk entirely, and there is no way around that from the teacher's side. There is also no individual paid tier at all. Pricing goes from free straight to a school or district contract, so a teacher who outgrows the free plan cannot simply upgrade (plans). The free tier carries 20+ tools against 35+ on paid plans, and standard language models rather than the faster ones.
Pricing: free at $0 for individual teachers; Premium and Intelligence are school and district plans, both custom-priced and quote-only with no public figures (plans, checked October 4, 2026).
Safety: this is the detail to carry to an administrator. Custom data privacy agreements appear on the plans page as a paid-plan feature, so a department adopting Brisk on free teacher accounts has no negotiated agreement behind it, however well the tool works. Brisk carries a 9ine assessment; our record does not state a hosting location or a minimum age. The vendor publishes a privacy notice.
EnlightenAI: best for applying your rubric to student writing
EnlightenAI is the best fit for the English or humanities teacher who already has a department rubric and needs it applied consistently across a stack of essays, because it scores against the rubric you supply rather than one it drafts. The vendor states that it learns your rubric and writes feedback in your voice, and its workflow puts the teacher last in the chain: "You approve, edit, or override as needed before releasing final scores" (AI grading tool). Class-level views show which criterion students missed most, which is where a rubric starts changing instruction rather than just recording an outcome.
Watch out for: two things. The vendor states the tool is free for teachers across grades 3-12, which is generous and also a scope limit, so it is not the answer for a college course. And our directory record holds no educator reviews at all, so the compliance record is the only independent signal available and classroom evidence is not.
Pricing: the vendor states it is free for teachers across grades 3-12 and does not publish figures for paid or district tiers, which are quoted directly (AI grading tool, checked October 4, 2026). No public pricing.
Safety: the strongest declared record in this comparison. Our directory record shows FERPA, COPPA and GDPR all declared, data hosted in the United Kingdom, a stated minimum age of 8+, and an independent 9ine assessment. UK hosting is worth checking against your own district's requirements rather than assuming it helps or hurts; some US districts require domestic hosting and some do not care. The vendor publishes a privacy policy.
GradingPal: best for handwritten and scanned work
GradingPal is the best fit for the teacher whose students still write on paper, because its OCR reads scanned handwriting and scores it against a rubric, which no other tool in this comparison does. Optical character recognition, the technology that turns a photograph of writing into text a computer can read, is the whole reason to choose it. You can supply your own rubric or have one drafted, pick from six feedback styles, and pull work in through Google Classroom.
Watch out for: the free tier is half what it was. It allowed 100 submissions a month when this article first checked it in August 2026 and now allows 50 (pricing). Several assignment types on the Lite plan, including quizzes, exams, worksheets, problem sets, video and audio, are paid add-ons rather than included features. GradingPal also carries no independent security assessment in our record. Its privacy notice, last updated April 28, 2026, sets out broad collection covering teacher account and billing data, technical data such as IP address and device type, and student submissions and grades where a school uses the service, and states that data is processed and stored in the United States and other jurisdictions. Note one conflict worth resolving before purchase: our directory record lists GDPR compliance as in progress, while the vendor's own privacy notice displays a GDPR Compliant badge. Ask the vendor which is current.
Pricing: free at $0 with 50 submissions a month and no card required; Lite at $14 a month or $168 a year with 250 submissions; Pro at $26 a month or $312 a year with unlimited submissions; School and District quoted directly and including a data privacy agreement (pricing, checked October 4, 2026).
Safety: our record shows FERPA and COPPA declared, data hosted in the United States, no minimum age and no independent 9ine assessment, with GDPR marked in progress against the vendor's own GDPR Compliant badge. Of the tools here, this is the one where the gap between what is declared and what has been checked by anyone outside the company is widest, and a tool with no stated minimum age that accepts student submissions is the one to put in front of your district before a class uses it.
Knowt: best free rubric grid with students in the loop
Knowt is the best fit for the teacher with no budget who wants students working with the rubric rather than only being graded against it, because its rubric generator is free and sits beside flashcards, practice and monitored AI chats students can actually use. Our directory holds a rating of 5.00 from a single review, which is a placeholder rather than a verdict; one review tells you almost nothing.
Watch out for: the word free is doing two jobs on Knowt's site. The rubric generator page states the generator is completely free. The help center states that free teacher accounts allow one class and have no access to AI features (help center). Read together, the standalone generator is free and the classroom workflow around it is not, so plan around the page that describes what you actually intend to do. Rubric features are also simpler here than in the dedicated tools, because the platform is built for studying.
Pricing: the rubric generator is free; Teacher Ultra is $19.99 a month or $120.00 a year, and buying it upgrades students on shared class files without separate student subscriptions (help center, checked October 4, 2026).
Safety: our record shows FERPA, COPPA and GDPR declared, data hosted in the United States, no minimum age and an independent 9ine assessment. No minimum age makes it one of the few tools here a whole elementary class could touch, which raises rather than settles the COPPA question for under-13s, and your district will want that in writing. The vendor publishes a privacy policy.
The big names, in one line each
Khanmigo's teacher side leans toward lesson plans and quiz questions rather than rubric construction, and its student tutor is the stronger half of the product. Quizlet builds study sets and practice activities, which is a different artifact from a rubric and a better one for revision. ChatGPT and Claude will write a serviceable rubric and will not hold it still across a class, export it into a grading interface or keep a record of how you applied it. Gemini Notebook, the tool Google renamed from NotebookLM in July 2026, answers only from documents you upload, which makes it a good way to draw criteria out of a syllabus or a department handbook and a poor way to grade against them. Diffit levels a text rather than assessing one. Canva will make the finished grid look like a document your department is happy to hand out.
Four questions that pick your rubric tool
Ask them in order. Most teachers have their answer by question two.
- Do you already have the rubric? If your department agreed the criteria last year, you need a tool that accepts them: EnlightenAI, GradingPal or Gradescope. If you are starting from a blank page, MagicSchool, Brisk or Knowt will draft one faster than you will.
- Who is grading, and how many? One teacher with 30 essays is a different problem from four teaching assistants with 400 problem sets. Collaborative grading, anonymous grading and rubric import are institutional features in Gradescope and absent from the K-12 tools entirely.
- What does the work arrive on? Paper points to GradingPal's OCR. A Google Doc points to Brisk. A PDF upload at scale points to Gradescope.
- Who signs for the data? One teacher on a free plan is a personal decision. A department is a purchase with a named owner, a signed agreement and somebody who has read the privacy policy. Brisk's custom data agreements sit on paid plans only, which catches departments out.
Underneath all four sits the rule worth keeping: outsource the doing, not the thinking. Writing descriptor cells is the doing. Deciding what counts as quality, and what weight each criterion carries, is the thinking, and no subscription buys it. Generate the grid, then change at least the criteria and their weighting before it touches student work. A rubric you accepted untouched is a definition of quality somebody else wrote.
A worked example: one essay set, one rubric, five checks
Take a grade 10 English class, 28 argumentative essays, and a department rubric with four criteria: thesis, evidence, analysis and conventions. Here is the sequence as a plan.
- Put your rubric in, do not generate one. Upload or paste the four agreed criteria with their weightings rather than accepting a draft. Two minutes, and it removes the largest source of drift.
- Grade five by hand first. Score five essays yourself against the rubric before the tool sees them. Keep the scores.
- Run the same five through the tool. Compare criterion by criterion, not on the total. A total can match while the parts disagree, which is the failure mode that matters.
- Look for the criterion it inflates. Analysis is usually the one. A tool that cannot tell a claim from a warrant rewards essays that assert confidently, and your rubric probably does not.
- Decide what the tool is allowed to do. If it tracks you on conventions and evidence but drifts on analysis, let it draft those two and grade analysis yourself. That is a legitimate result, not a failed trial.
- Release nothing unreviewed. EnlightenAI and Gradescope both build teacher confirmation into the workflow. Where a tool does not, you supply it.
The whole check costs about forty minutes once, against a stack you were going to grade anyway, and it tells you something no feature list can: whether this tool agrees with you about what good work looks like.
What to do this week
Pick one assignment you were setting regardless. Build or upload the rubric, grade five pieces both ways, and compare by criterion. If the tool tracked your judgment, you have a case to take to your department and the evidence to make it. If it did not, you spent forty minutes and learned which criterion to keep for yourself.
Then ask the question this article cannot answer for you. Adoption is running ahead of policy: RAND's national survey found that 53 percent of English language arts, math and science teachers used AI for school in 2024-25, up more than 15 percentage points in two years, while only 45 percent of principals reported having any school or district AI guidance in place (RAND, September 2025). If student work is going into one of these tools, somebody in your building needs to have signed for it.
The bottom line
MagicSchool drafts the fastest standards-aligned grid, Brisk Teaching builds it inside the document you already have open, EnlightenAI applies the rubric you wrote to student writing, GradingPal handles the paper, Knowt does it free with students involved, and Gradescope is the one built for college and AP grading at scale. If you take one thing from the update: check whether the tool follows your criteria or its own, because that is the difference the product pages do not describe.
For scoring whole papers rather than building the grid, the AI grading tools comparison goes deeper. For the feedback that follows a score, see AI writing feedback tools, and for the assessment either side of it, AI quiz makers and AI worksheet generators. Before student work touches any of them, the 2026 compliance guide for schools covers how to vet a tool, and the full directory holds the records these recommendations were built from.
Build the rubric with the machine. Decide what good looks like yourself.
Dan Fitzpatrick is a Forbes contributor, bestselling author and founder of The AI Educator, and has trained more than 150,000 educators across 30+ countries.
