Pangram vs Turnitin: the 14 fields and the study results side by side
Turnitin was tested in 3 independent studies indexed here, recording 0 false accusations in each (18, 42 and 10 human-written documents), while no indexed study tested Pangram, by the count compiled on 8 November 2026 from the papers linked in the studies index. The two cards differ on almost every other field as well: Pangram sells to individuals from $20 per month and Turnitin sells only institutional licences, so the comparison below reads each field as published by each vendor.
Updated Fields checked 8 November 2026
Short answer
Turnitin recorded 0 false accusations in 18 documents in Weber-Wulff 2023, 0 of 42 human papers in Walters 2023 and 0 of 10 in Perkins 2024, but missed 88 of 104 AI samples in Perkins 2024; Pangram's 1 in 10,000 false positive rate is the vendor's own figure (pangram.com). Pangram states 24 named languages and text, image and video detection with self-serve plans; Turnitin states 7 languages, text only, and access through a Turnitin Originality licence with no published price (pangram.com, guides.turnitin.com). All read 8 November 2026. No detector score should be the only evidence in a misconduct or hiring decision.
How do Pangram and Turnitin compare on the 14 fields?
| Field | Pangram | Turnitin |
|---|---|---|
| 1. Name, vendor, official domain | Pangram, Pangram Labs, pangram.com | Turnitin AI writing detection, Turnitin, LLC, turnitin.com |
| 2. Modalities detected | text, image, video | text |
| 3. Generators covered | OpenAI, Anthropic, Google, Meta, DeepSeek, xAI, Claude Opus 5, GPT 5.6, Gemini 3, Grok 4.3, Claude Sonnet 5, Claude Fable 5, GPT-5.4, GPT-OSS 120B, DeepSeek V4, Llama 3.3, Mistral Medium 3.5, Kimi K2.6, Qwen 3.7 Max, Gemma 4, Nemotron 3 Ultra, GLM 5.2, 26 models benchmarked | 36 named models including Amazon Nova 2 Lite, Claude-3-Haiku through Claude Opus 4.6, DeepSeek V3.2, Gemini-1.0-pro through Gemini-3.1-pro-preview, GPT-4o through GPT-5.4-pro, o1-mini, Grok-4.1, Llama-3.3-70b, Llama-4-Maverick-17b, Mistral-large-3; Spanish model: GPT-3.5 and GPT-4; Japanese model: GPT-4 (0613), GPT-4o, GPT-4o-mini |
| 4. Languages supported | 24 named: English, Arabic, Chinese, Czech, Dutch, French, German, Greek, Hindi, Hungarian, Italian, Japanese, Korean, Persian, Polish, Portuguese, Romanian, Russian, Spanish, Swedish, Turkish, Ukrainian, Urdu, Vietnamese | 7: English, Spanish, Japanese, Modern Arabic, Dutch, German, Turkish |
| 5. Input limits | Supports PDF, DOCX, RTF files and up to 100 files at a time. Minimum and maximum words per text check not published. Image: JPG, PNG, WebP, min 512x512px, up to 30 MB each. Video: MP4, MOV, WebM, min 720x720px, up to 500MB. API: 5 QPS realtime; bulk 1,000 billable units. | English: file size under 100 MB; at least 300 words of prose text in a long-form writing format; must not exceed 30,000 words; accepted file types .docx, .pdf, .txt, .rtf. Lists, headers, references, tables, formulas and code are excluded from scoring. |
| 6. Free tier | 20 free credits daily (twenty free checks a day after signing up for the free plan, including AI assistance detection) | None. AI writing detection is only available to customers that license Turnitin Originality (or the AI writing add-on for iThenticate 2.0). |
| 7. Paid plans | Individual $20 / month; Professional $65 / month; Team $40 / month, 2 seats; API metered: Text (Pangram 4.1) $0.05 per 100 words, Image 1.0 $0.25 per image, Image Flash 1.0 $0.05 per image, Video $0.05 per second | |
| 8. API | Yes, docs.pangram.com | |
| 9. Integrations | Chrome extension, Firefox extension, Google Docs, Gmail, Google Classroom, Canvas, Moodle, Brightspace (D2L), MCP server | LMS (Moodle, Blackboard, Canvas, etc.) via the Similarity Report; no AI indicator embedded in the LMS user interface; Microsoft Teams (Feedback Studio integration only) |
| 10. Vendor accuracy and false positive claims | States accuracy of over 99% and third-party verification by researchers at the University of Chicago and The University of Maryland; "accurate on 99.6% of these AI-generated samples"; false positive rate 1 in 10,000 on an aggregate of public datasets containing tens of millions of written documents | States a false positive rate under 1% for documents with over 20% AI writing, validated on over 700,000 pre-ChatGPT academic papers; ELL writers 0.014 and native English writers 0.013 in its ELL research; no accuracy rate stated |
| 11. Independent studies | 0 records | 3 independent records (Weber-Wulff 2023, Walters 2023, Perkins 2024) plus the vendor statements record |
| 12. Data retention and privacy statement | Submitted text stored for as long as the account is active, deletable from the history tab; detection trained on a proprietary dataset that does not include user submissions; never shared or sold; zero data retention option for enterprise API | No separate repository for AI writing detection; customers choose whether student papers enter the repository; AI results retained as part of the similarity report; full deletion of submissions on request, no partial deletion of the AI component |
| 13. Launch year and ownership | 2023; Pangram Labs, co-founders Max Spero (CEO) and Bradley Emi (CTO), founded in Brooklyn, NY; no parent company stated | 2023; Turnitin, LLC; no parent company stated on pages read |
| 14. Rankings the card appears in | Languages supported | Independently tested, false positive rate, free tiers, languages supported, input limits |
Which studies tested both Pangram and Turnitin?
None of the nine studies indexed on the morning of 9 November 2026 tested Pangram, so the table below sets Turnitin's measured results, and Turnitin's own statements, beside Pangram's own statements; the two vendor rows are labelled so that a company's figure is never read as a researcher's (studies index, compiled 8 November 2026). One record added later on 9 November 2026 tested both tools: Van Vlasselaer 2026 reports Pangram at 65% strict and 97.5% inclusive accuracy on 40 fully AI-generated papers against Turnitin scoring all 40 under 20%, Pangram at 37 of 40 and Turnitin at 24 of 40 hybrid papers within their known AI range, and both tools correct on all 40 human papers (DOI 10.1007/s40979-026-00226-w, read 9 November 2026). Jabarian and Imas 2025 tested Pangram alone of the pair, reporting a 0.0000 false positive rate on 1,992 human passages (NBER w34223, read 9 November 2026).
| Source | Pangram | Turnitin |
|---|---|---|
| Vendor statement, pangram.com (read 8 November 2026) | False positive rate 1 in 10,000 on an aggregate of public datasets of tens of millions of documents; accurate on 99.6% of AI-generated samples | |
| Vendor statement, Turnitin claims record (2023 to 2026) | Document false positive rate under 1% for documents with over 20% AI writing, validated on 800,000 (May 2023) and over 700,000 (current FAQ) pre-ChatGPT papers; sentence-level false positive rate about 4% (2023); no recall or accuracy figure published | |
| Weber-Wulff 2023, 54 documents | 41 of 54 correct (76%, rank 1 of 14); 0 false accusations in 18 human-written documents; 10 of 36 AI documents undetected, all among edited and paraphrased texts; Turnitin knew the documents were a test suite | |
| Walters 2023, 126 documents | 100% of 126 correct, 0 incorrect, 0 uncertain; 0 false positives in 42 human papers; the author notes the papers could have been used to train the detector | |
| Perkins 2024, 114 samples | Baseline mean accuracy 61% (rank 2 of 7); 50% on non-manipulated AI text falling to 7.9% after adversarial techniques; 0 false accusations in 10 human samples; 88 of 104 AI samples undetected |
Results as each study reports them: arXiv 2306.15666, De Gruyter 10.1515/opis-2022-0158, arXiv 2403.19148, Turnitin FAQ, all read 8 November 2026.
What do the two cost over 12 months?
| Plan | Monthly as stated | Annual as stated | Inclusions as stated | Checked on |
|---|---|---|---|---|
| Pangram Individual | $20 / month | Save $60, annually (12 x $20 less $60 is $180 by arithmetic) | Scan up to 300,000 words per month, 100 image detection scans monthly, similarity detection, Pangram MCP, Google Docs and Gmail integration, $10 in monthly API credits; 7-day free trial | 8 November 2026 |
| Pangram Professional | $65 / month | Save $240, annually ($540 by arithmetic) | 5x more words, 5x more image scans, $200 of API credits monthly | 8 November 2026 |
| Pangram Team | $40 / month, 2 seats | $20 / seat per month | 8 November 2026 | |
| Pangram API | Text (Pangram 4.1): $0.05 per 100 words; Image 1.0: $0.25 per image; Image Flash 1.0: $0.05 per image; Video: $0.05 per second | Bulk text API: 20% discount | 8 November 2026 | |
| Turnitin AI writing detection | Included with a Turnitin Originality licence or the AI writing add-on for iThenticate 2.0; no price published; institution-level contract | 8 November 2026 |
Prices in USD as shown with the $ symbol on pangram.com/pricing, checked 8 November 2026. Turnitin publishes no price on the FAQ or product page read the same day, so its 12-month cost is a contract figure the institution holds.
One of three comparisons in the compare hub. The cards are Pangram and Turnitin, and the home board answers which AI detectors exist and who has tested them?
Which card fits which reader?
- Choose Pangram ifyou are an individual, a department or a small team: plans start at $20 per month with a 7-day trial, and Turnitin sells no individual access (pangram.com/pricing, guides.turnitin.com, read 8 November 2026).
- Choose Pangram ifyour texts span many languages or include images and video: 24 named languages and three modalities against Turnitin's 7 languages and text only (pangram.com, guides.turnitin.com, read 8 November 2026).
- Choose Turnitin ifyour institution already licenses Turnitin Originality and wants results inside the Similarity Report, where 3 indexed studies recorded 0 false accusations on 18, 42 and 10 human documents (arXiv 2306.15666, De Gruyter, arXiv 2403.19148).
- Choose Turnitin ifyou submit long academic documents: 300 to 30,000 words of prose per submission, while Pangram publishes no words-per-check figure (guides.turnitin.com, pangram.com, read 8 November 2026).
- IfRead Perkins 2024 before relying on either for edited text: Turnitin fell from 50% to 7.9% on AI text after adversarial techniques, and Pangram's statement that humanized text is detected stands on the vendor's page alone in this index (arXiv 2403.19148, pangram.com, read 8 November 2026).
How does the evidence compare, tool by tool?
Pangram
- Register ID
- AIDI-D-0006
- Study records
- 4 in the index: Jabarian and Imas 2025, Van Vlasselaer 2026, DAMAGE 2025, Epoch AI 2026
- Shared with Turnitin
- 1: Van Vlasselaer 2026
- Vendor accuracy claim
- Pangram states accuracy of over 99% and third-party verification by researchers at the University of Chicago and The University of Maryland. We're accurate on 99.6% of these AI-generated samples.
- Free tier
- 20 free credits daily (twenty free checks a day after signing up for the free plan, including AI assistance detection)
- Languages
- 24 stated
Turnitin
- Register ID
- AIDI-D-0002
- Study records
- 5 in the index: Weber-Wulff 2023, Walters 2023, Perkins 2024, Van Vlasselaer 2026, Turnitin claims
- Shared with Pangram
- 1: Van Vlasselaer 2026
- Vendor accuracy claim
- None published
- Free tier
- None. AI writing detection is only available to customers that license Turnitin Originality (or the AI writing add-on for iThenticate 2.0).
- Languages
- 7 stated
How this comparison is built
- Fields
- The 14 fields of every card, in card order
- Shared study records
- 1: Van Vlasselaer 2026
- Fields checked
- 8 November 2026
- Records in the studies index
- 35 records
Both columns are drawn from the same 14 fields in the index data file, read on 8 November 2026 from the vendor page recorded on each card, and from the studies index, where each result is copied as the paper reports the figure, with the base stated. Turnitin's own statements are kept in a separate record and labelled as vendor statements, exactly as Pangram's are here. The rules, the recheck schedule (next pass 8 December 2026) and the correction route are on the methodology page; the false positive rate ranking sorts every card on the evidence field used in the first sentence.
Where each figure comes from
Values as each vendor states them on pangram.com, pangram.com/pricing, pangram.com/data-privacy, pangram.com/about-us, the Turnitin AI writing FAQ and turnitin.com, all read 8 November 2026. An empty cell means the card has no published value for the field.
Questions about Pangram and Turnitin
Has any study tested Pangram against Turnitin?
No study indexed here as of 9 November 2026 tested Pangram at all, so no head-to-head result exists in the index. Turnitin appears in Weber-Wulff 2023, Walters 2023 and Perkins 2024 (studies index).
What false positive rate does each publish?
Pangram states 1 in 10,000 on an aggregate of public datasets (pangram.com); Turnitin states under 1% for documents with over 20% AI writing on over 700,000 pre-ChatGPT papers (guides.turnitin.com), both read 8 November 2026. Both are vendor figures; Turnitin's 0 false accusations in three studies are the measured ones.
Can an individual buy Turnitin AI detection?
No: the FAQ states that AI writing detection is available only to customers licensing Turnitin Originality or the iThenticate 2.0 add-on, read 8 November 2026 (guides.turnitin.com). Pangram's Individual plan is $20 per month (pangram.com/pricing).
Which detects more languages?
Pangram names 24 languages; Turnitin names 7, with paraphraser and bypasser detection limited to English, both read 8 November 2026 (pangram.com, guides.turnitin.com).
How did Turnitin do on edited AI text?
In Perkins 2024, Turnitin's accuracy fell from 50% on non-manipulated AI text to 7.9% after adversarial techniques, the largest drop of seven tools, and 88 of 104 AI samples went undetected (arXiv 2403.19148, read 8 November 2026). In Weber-Wulff 2023 the 10 AI documents Turnitin missed were all edited or paraphrased (arXiv 2306.15666).
What does each do with submitted text?
Pangram stores text while the account is active and lets users delete history; Turnitin lets institutions choose whether papers enter the repository and keeps AI results within the similarity report, both read 8 November 2026 (pangram.com/data-privacy, guides.turnitin.com).
What is the minimum text each accepts?
Turnitin requires at least 300 words of prose in a long-form format; Pangram publishes no minimum or maximum words per text check, both read 8 November 2026 (guides.turnitin.com, pangram.com).
Sources and revisions
Sources
- Pangrampangram.com
- Pangram pricingpangram.com/pricing
- Pangram data privacypangram.com/data-privacy
- Pangram aboutpangram.com/about-us
- Turnitin AI writing FAQguides.turnitin.com/hc/en-us/articles/284...
- Turnitin AI writing product pageturnitin.com/solutions/topics/ai-writing
- Weber-Wulff et al. 2023arxiv.org/pdf/2306.15666
- Walters 2023degruyterbrill.com/document/doi/10.1515/o...
- Perkins et al. 2024arxiv.org/pdf/2403.19148
