AI detectors for school districts

AIDI-U-02

Updated ·

An AI detector for school districts is chosen on four fields, and two of them decide whether a card appears here at all: a named learning management system in integrations, and a published false positive figure in rank_fields.published_fpr. Of the 72 cards, 14 name an LMS, 26 carry a false positive figure, 24 state at least one of the two, and 9 state both. Those 9 cards make up this view, Compilatio having joined with the fourth wave of cards on 9 November 2026 and PlagiarismCheck with the ninth on 10 November 2026, with the student data statement from each privacy policy and the institutional plan as the other two columns. This page sits under AI detectors by use case beside the views for recruiters and business, and the same cards can be compared side by side on the full board. Educational institutions beyond K-12 read the same fields.

  1. Field 1LMS integrations as stated
  2. Field 2Published false positive rate
  3. Field 3Source
  4. Field 4Institution plan as stated

The LMS field is the first filter because a tool that lives outside Canvas, Moodle, Blackboard, D2L, Schoology or Google Classroom is a tool teachers will not open. The LMS integrations slice lists every named integration; this view keeps only the cards that also publish a false positive figure, because a district is the reader most exposed to a wrong flag on a student. The figure is taken from a study in the index where one tested the detector, and from the vendor where none did, with the source named in the table. Student data policy quotes the privacy policy; pricing is the institutional tier as stated. Grammarly Authorship and Turnitin Clarity, added on 9 November 2026, name Canvas and LTI 1.3 in Canvas, Moodle, Blackboard and D2L but publish no false positive figure, because each records a student's writing process rather than scoring the text; the LMS integrations slice carries both (grammarly.com/authorship, Turnitin Clarity administrator FAQs, read 9 November 2026). The rules are in the methodology, and the false positive rate ranking orders every card on the same field. Trinka, added on 10 November 2026, names Canvas, Moodle, Blackboard and D2L on a platform-wide institutional line and publishes no false positive figure ("Low False Positive Rate" with no number), so it stays out of this view and sits on the LMS integrations slice (trinka.ai/pricing, read 10 November 2026); Passed.ai, added the same day, names Google Classroom only on its closing notice as an Originality.ai feature.

Which fields decide fit for a district?

AI detectors for school districts: the cards in this view and the fields that decide fit
DetectorLMS integrations as statedPublished false positive rateSourceInstitution plan as stated
Winston AIGoogle Classroom0 false accusations in 18 human-written documentsWeber-Wulff 2023Enterprise: Contact sales
PangramGoogle Classroom, Canvas, Moodle, Brightspace (D2L)1 in 10,000Vendor claimTeam: $40 / month, 2 seats
GPTZeroGoogle Classroom, Canvas0.00% at naive threshold 0.5RAID 2024Professional Team Plan: 500,000 words per member, per month; min 2 seats
CopyleaksCanvas, D2L, Moodle, Blackboard, Schoology, Edsby, Sakai, Google Classroom0 of 42 human papersWalters 2023Education and Enterprise: Custom
Originality.aiGoogle Classroom (via Chrome extension), Moodle Plugin 2.00 of 42 human papers (2 uncertain)Walters 2023Enterprise (yearly) $136.58/month
TurnitinLMS (Moodle, Blackboard, Canvas, etc.) via the Similarity Report0 of 42 human papersWalters 2023Licence for Turnitin Originality; no free tier
CodequiryCanvas, Moodle and Blackboard (bulk import), LTI LMS integration1.3% of human-written code falsely flagged (no sample size)Vendor claimGrow US$200/month base (C$272 /mo as displayed); Enterprise: contact sales
CompilatioMoodle, Canvas, Brightspace (D2L), Blackboard (Learn and Learn Ultra), Microsoft Teams, Open LMS, Cineca, LTI0 false accusations in 18 human-written documentsWeber-Wulff 2023; vendor states below 1% on 3,700 human textsMagister+ by demo and quote; Magister Standard excludes AI detection
PlagiarismCheckCanvas, Moodle, Google Classroom, Schoology, Brightspace, Blackboard, Populi0 false accusations in 18 human-written documentsWeber-Wulff 2023Institutional plans by request; "educators can control their repository within integration"

Values as each vendor or study states them, read 8 November 2026; the Codequiry row, a detector of AI-written source code rather than essays, was read 9 November 2026.

LMS lists come from Winston AI, Pangram, GPTZero, Copyleaks, Originality.ai and the Turnitin FAQ, all read 8 November 2026, and Codequiry, read 9 November 2026. False positive figures come from Weber-Wulff 2023, Dugan et al. 2024 (RAID), Walters 2023 and, for Pangram and Codequiry, the vendor's own pages (pangram.com, codequiry.com). Plans come from gowinston.ai/pricing, pangram.com/pricing, gptzero.me/pricing, copyleaks.com/pricing, originality.ai/pricing, codequiry.com/pricing and the Turnitin FAQ.

AI detectors for school districts: register plate with the page title and its register ID
The view plate: register ID and title, drawn from the index data.

Why each card is in this view

  1. AIDI-D-0005Winston AI

    the product page names Google Classroom (gowinston.ai, read 8 November 2026), and Weber-Wulff 2023 recorded 0 false accusations in 18 human-written documents (arXiv). The privacy policy dated July 17, 2026 says "We do not use your individually identifiable submitted content to train AI models without your explicit consent" (privacy policy).

  2. AIDI-D-0006Pangram

    the home page names Google Classroom, Canvas, Moodle and Brightspace (D2L) and states a false positive rate of 1 in 10,000 "calculated on an aggregate of public datasets containing tens of millions of written documents" (pangram.com, read 8 November 2026). No study in the index tested Pangram, so the figure is marked as the vendor's. The data privacy page says the detection technology "doesn't include user submitted content" in its training data (data privacy).

  3. AIDI-D-0001GPTZero

    the home page names Google Classroom and Canvas (gptzero.me, read 8 November 2026); RAID 2024 Table 4 records 0.00% of human-written documents flagged at threshold 0.5 (arXiv). The privacy policy states compliance with FERPA, GDPR and CCPA and deletion of personal data on request (privacy policy).

  4. AIDI-D-0003Copyleaks

    eight LMS names on the product page (copyleaks.com, read 8 November 2026); Walters 2023 Table 4 records 0 of 42 human papers classified as AI (publisher PDF). The privacy policy revised August 24, 2026 allows use "including by training Copyleaks' models" with an opt-out for direct customers (privacy policy).

  5. AIDI-D-0004Originality.ai

    Google Classroom via the Chrome extension and a Moodle Plugin 2.0 (extensions page, read 8 November 2026); Walters 2023 records 0 of 42 human papers flagged, with 2 returned uncertain. The privacy policy amended June 17, 2026 says "We do not use your personal data or any Student Data to train AI or large language models" (privacy policy).

  6. AIDI-D-0002Turnitin

    the AI indicator appears inside the Similarity Report delivered through Moodle, Blackboard, Canvas and other LMSs, not in the LMS interface itself (Turnitin FAQ, read 8 November 2026); Walters 2023 records 0 of 42 human papers flagged. The FAQ says "customers retain the ability to choose whether to add their student papers into the repository or not".

  7. AIDI-D-0032Compilatio

    the LMS integration page names Moodle, Canvas, Brightspace, Blackboard Learn and Learn Ultra, Microsoft Teams, Open LMS, Cineca and LTI (compilatio.net, read 9 November 2026), and Weber-Wulff 2023 Table 9 records 0 false accusations in 18 human documents beside 11 of 36 AI documents undetected (arXiv 2306.15666, read 9 November 2026); the vendor's own below 1% on 3,700 human texts is a vendor figure (support.compilatio.net, read 9 November 2026). The privacy policy stores submitted documents for 5 years from the end of the contract on servers in France (support.compilatio.net, read 9 November 2026), and Magister+ is sold to institutions by demo and quote.

  8. AIDI-D-0020Codequiry

    the AI code detection page names Canvas, Moodle and Blackboard bulk import and an LTI LMS integration, and states that model v3e "falsely flags 1.3% of human-written code", with no sample size published (codequiry.com, read 9 November 2026). No study in the index tested Codequiry, so the figure is marked as the vendor's. The privacy policy last updated October 22, 2026 says source code submitted without an active Rapid add-on is added in anonymised form to a cross-account matching corpus and retained indefinitely, and that the Rapid add-on carries the right to delete submissions (privacy policy). Codequiry scores source code in programming courses, not essays.

  9. AIDI-D-0046PlagiarismCheck

    seven LMS platforms named and 0 of 18 in Weber-Wulff 2023, where the detector also ranked 13th of 14 and missed 19 of 36 AI documents (plagiarismcheck.org/ai-detector, read 9 November 2026; Weber-Wulff 2023); the privacy policy states institutional retention is governed by the institution's agreement. Detecting-AI.com, added the same day, names Moodle but states "under 1%" false positives with no base, a bound the index does not record as a figure, so it stays in the LMS integrations slice only.

What the evidence says for school districts

Who may buy the detector is its own field: the institution-only view, added 10 November 2026, lists the five cards sold only through a licence, demo or sales team (Turnitin, Turnitin Clarity, Compilatio Magister+, Pindrop Pulse, Sensity) and the fourteen that sell a self-serve plan beside an institutional tier, with how each grants access.

Student writing is the sample in most of the studies indexed here. In Walters 2023, 126 first-year composition and GPT-written papers went through 16 detectors: Copyleaks and Turnitin classified all 126 correctly, Originality.ai (model 1.4) classified 98% correctly with 0 incorrect, and GPTZero (Pro) 81% with a 2% false positive rate on 42 human papers (Walters 2023, publisher PDF, read 8 November 2026). The author notes the student papers could have been used to train the Turnitin detector.

The sharpest finding for a district is Liang 2023. Across seven detectors, the paper reports an average false positive rate of 61.22% on 91 human-written TOEFL essays and near-perfect accuracy on 88 US 8th-grade essays; 18 of the 91 TOEFL essays were flagged as AI by all seven detectors, and GPTZero and Originality.ai were among the seven (Liang 2023, arXiv, read 8 November 2026). A district with English learners is the population this study describes.

Weber-Wulff 2023 ranked Turnitin first of 14 tools with 41 of 54 documents correct and 0 false accusations in 18 human-written documents, Winston AI fifth with 36 of 54 and 0 false accusations, and GPTZero eighth with 29 of 54 and 9 false accusations in 18 (Weber-Wulff 2023, arXiv, read 8 November 2026). In Perkins 2024, Turnitin fell from 50% on non-manipulated AI text to 7.9% after adversarial techniques, the largest drop of the seven tools (Perkins 2024, arXiv, read 8 November 2026). Turnitin's own statements, recorded as a separate entry, put the document false positive rate under 1% for documents with over 20% AI writing, validated on over 700,000 pre-ChatGPT papers (Turnitin claims record, Turnitin FAQ, read 8 November 2026).

No detector score should be the only evidence in a misconduct or hiring decision. GPTZero's FAQ says "These results should not be used to punish students", and Turnitin's FAQ says the indicator "should not be used as the sole basis for action", both read 8 November 2026.

Turnitin deletes a student's whole submission on an institutional request and cannot remove the AI writing component alone, as the deletion routes question, added 9 November 2026, records. One card names FERPA on a page the index has read, GPTZero, and Turnitin's trust centre names no standard; the compliance view, added the same day, lists every named standard across the index.

Quote-priced programmes for institutions are listed by name in the nonprofit and education question, twelve cards from Turnitin's licence to Codequiry's class plans, with the finding that none states a discount, read 10 November 2026; the student plan question lists what a student can buy alone.

Sources

  • https://gowinston.ai/ai-content-detector/ (read 8 November 2026)
  • https://gowinston.ai/pricing/ (read 8 November 2026)
  • https://gowinston.ai/privacy-policy/ (read 8 November 2026)
  • https://www.pangram.com/ (read 8 November 2026)
  • https://www.pangram.com/pricing (read 8 November 2026)
  • https://www.pangram.com/data-privacy (read 8 November 2026)
  • https://gptzero.me/ (read 8 November 2026)
  • https://gptzero.me/pricing (read 8 November 2026)
  • https://gptzero.me/faq (read 8 November 2026)
  • https://gptzero.me/privacy-policy.html (read 8 November 2026)
  • https://copyleaks.com/ai-content-detector (read 8 November 2026)
  • https://copyleaks.com/pricing (read 8 November 2026)
  • https://copyleaks.com/privacy-policy (read 8 November 2026)
  • https://originality.ai/extensions (read 8 November 2026)
  • https://originality.ai/pricing (read 8 November 2026)
  • https://originality.ai/privacy-policy (read 8 November 2026)
  • https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs (read 8 November 2026)
  • https://codequiry.com/detect-ai-written-code (read 9 November 2026)
  • https://codequiry.com/pricing (read 9 November 2026)
  • https://codequiry.com/privacy-policy (read 9 November 2026)
  • https://www.degruyterbrill.com/document/doi/10.1515/opis-2022-0158/pdf (read 8 November 2026)
  • https://arxiv.org/pdf/2304.02819v3 (read 8 November 2026)
  • https://arxiv.org/pdf/2306.15666 (read 8 November 2026)
  • https://arxiv.org/pdf/2403.19148 (read 8 November 2026)
  • https://arxiv.org/pdf/2405.07940 (read 8 November 2026)

Revision log: 2026-11: view created2026-11-09: Codequiry row added; the counts in the opening paragraph now follow the index data file at build time2026-11-09: six wave 3 cards checked; Grammarly Authorship and Turnitin Clarity noted as LMS cards with no false positive figure, compound count 18; 2026-11-09: ten wave 4 cards checked, Compilatio added, compound counts 21 and 8; the College Board question page records that the AP and SAT policies name no detector2026-11-09: results as evidence question, Canvas question and LMS coverage ranking linked (seventh wave)2026-11-09: eighth-wave pages linked2026-11-10: PlagiarismCheck added (ninth wave), 9 cards; Detecting-AI.com names Moodle with a bound and no figure2026-11-10: institution-only view linked2026-11-10: twelve tenth-wave cards checked; Trinka (LMS line, no false positive figure) and Passed.ai (closing notice only) noted and left out2026-11-10: twelfth-wave pages linked

FAQ: AI detectors for school districts

Which AI detectors fit school districts?

Seven cards name an LMS and publish a false positive figure: Winston AI, Pangram, GPTZero, Copyleaks, Originality.ai and Turnitin, as read on 8 November 2026, and Codequiry, a detector of AI-written source code for programming courses, read 9 November 2026. Copyleaks names the most LMS platforms at eight; Turnitin is the only card with no self-serve plan. Fit beyond those two fields depends on your student data policy and budget, and the table carries both as stated.

Which detectors integrate with Canvas and Moodle?

Pangram and Copyleaks name both Canvas and Moodle on their product pages, Turnitin delivers through both via the Similarity Report, GPTZero names Canvas, and Originality.ai names a Moodle Plugin 2.0, all read 8 November 2026. Winston AI names Google Classroom only, and Codequiry names Canvas, Moodle and Blackboard for bulk import of code submissions, read 9 November 2026. The LMS integrations slice carries every named platform.

What does the false positive evidence mean for students?

A false positive is a human essay flagged as AI; the rate changes with the sample. Copyleaks, Originality.ai and Turnitin each recorded 0 of 42 human papers flagged in Walters 2023, while Liang 2023 found an average of 61.22% of 91 TOEFL essays flagged across seven detectors. English learners are the students at risk.

Do these detectors train on student writing?

Policies differ. Originality.ai says student data is never used to train models, Pangram says user submissions are excluded from its training data, and Winston AI requires explicit consent, while the Copyleaks privacy policy allows training with an opt-out for direct customers, all read 8 November 2026. Read the retention field on each card before a district contract.

Which detectors fit educational institutions beyond K-12?

The same seven cards. Colleges read identical fields, and the Walters 2023 sample was first-year university composition. Turnitin's licence and Copyleaks' Education and Enterprise tier are written for institutions; Pangram, GPTZero and Originality.ai publish plans a department can buy directly. Two pages added 9 November 2026 bear on a district's choice: the results as evidence question, which lists the seven primary documents in the index on acting on a score, and the Canvas without Turnitin question, which tables the seven non-Turnitin cards that name Canvas with their institution plans; the LMS coverage ranking, added the same day, counts the LMS platforms each vendor names.

What does a district pay?

Pangram states $40 / month for a 2-seat Team plan and Originality.ai states $136.58/month for Enterprise billed yearly (pangram.com/pricing; originality.ai/pricing, read 8 November 2026). Copyleaks, Winston AI and Turnitin quote on request; GPTZero's Professional Team Plan starts at 2 seats with 500,000 words per member per month (gptzero.me/pricing).

Related reading