IsItAI vs Pangram: AI detector comparison
A side by side look at pricing, methodology, language support and best-fit use case. Every claim on this page is linked to a source at the bottom.
Last updated . Reviewed by IsItAI Editorial.
At a glance
IsItAI
UK, 2025UK-built AI detection that gives you an AI-pattern score and the passages behind it, not a verdict.
- Pricing
- Teacher plans from £9.99/month; one-off student checks from £4.99
- Free tier
- 3 scans/day with a free account, 1/day without
- API
- Yes
- Languages
- 30
Pangram
US, 2024Well-funded 2024 entrant with aggressive accuracy claims and strong LMS integrations.
- Pricing
- From $20/month (Individual, 600 scans)
- Free tier
- 4 scans/day free
- API
- Yes
- Languages
- 20
Side by side
| Feature | IsItAI | Pangram |
|---|---|---|
| Free tier | 3 scans/day with a free account, 1/day without | 4 scans/day free |
| Starting price | Teacher plans from £9.99/month; one-off student checks from £4.99 | From $20/month (Individual, 600 scans) |
| Cost per 1k words | Not word-metered; about £0.10 per scan on Teacher Pro | $0.05 per 1,000 words via API |
| API available | Yes | Yes |
| Languages supported | 30 | 20 |
| Founded | 2025 | 2024 |
| Country | UK | US |
| Primary audience | Educators, recruiters and content teams | Schools, universities, publishers and content-moderation teams |
| Best for | Teachers, students and recruiters who want a clear AI-pattern score and source-passage highlighting | Schools and teams that want AI detection inside the grading workflow with plagiarism checking combined |
| Known weakness | Smaller brand than US incumbents | Headline 99.98% accuracy and 1-in-10,000 false-positive claims come from the founders' own benchmark, not peer review; independent writers report inconsistent repeat results on the same text |
How each tool works
IsItAI
Method: A trained model that combines a language-model reading of the text with local statistical measurements and a curated phrase library. The layers are inputs to one decision, not votes. Returns a screening result with an AI-pattern score rather than a binary verdict, with per-layer results shown transparently.
Data handling: Submitted text is not stored after the scan completes, on any tier; account holders keep a scan history of results and scores, never the text. Text is processed by Anthropic during the scan and is not used to train models under its API data policy.
Pangram
Method: Deep-learning classifier with segment-level analysis and an 'AI assistance' mode for edited text. Founded in Brooklyn by Max Spero and Bradley Emi (ex-Google, ex-Tesla) as Checkfor.ai, renamed Pangram in 2024, with around $4m in seed funding.
Data handling: States submissions are not used to train models and are deleted within 30 days of account closure. Acts as a FERPA 'School Official' for US education customers.
How accurate are AI detectors? What the research says
Two findings from independent academic research apply to almost every detector on this list, including IsItAI and Pangram.
1. False positives on non-native English writers. Liang et al. (Stanford, 2023) found that GPT detectors flagged over half of TOEFL essays written by non-native English speakers as AI-generated, while flagging almost none of the essays from US students. If your use case involves non-native English writers, never treat a detector verdict as proof on its own.
2. Paraphrasing reliably defeats classifiers. Sadasivan et al. (Maryland, 2023) showed that running AI output through a paraphraser drops detection performance close to chance. Both IsItAI and Pangram are vulnerable to this.
For a deeper breakdown of how accuracy claims hold up across 2026 testing, see our guide to how accurate AI detectors actually are in 2026.
Where IsItAI wins
- Publishes an honest out-of-sample benchmark with the false-positive rate reported separately: 4 of 400 real student essays wrongly flagged (5 of 517 held-out professional passages), catch rates stated per case from about 24% on copy-paste student essays to about 42% on articles, and confidence intervals throughout. Pangram's 99.98% headline comes from a benchmark written by its own founders.
- Submitted text is never stored on any tier (a one-way hash with scores is kept for 24 hours to avoid re-billing identical scans), which keeps a UK or EU school's GDPR review short. Pangram retains content while an account is open and deletes within 30 days of closure.
- Flagged passages come with plain English reasons a teacher can weigh and a head of department can defend in an appeal. Teacher plans from £9.99 per month versus $20.
Where Pangram wins
- LMS integrations with Canvas, Moodle, Google Classroom and Brightspace put the score inside the grading workflow, where Is It AI runs in the browser with no LMS integration yet.
- Combines plagiarism checking with AI detection on paid tiers, plus browser extensions and a Google Docs integration.
- Detection across 20+ languages and an 'AI assistance' mode aimed at edited and humanised text, with a slightly more generous free tier at 4 scans per day.
Which one should you pick, IsItAI or Pangram?
Pangram vs Is It AI comes down to workflow versus evidence. Pick Pangram if you need detection inside Canvas or Classroom, multilingual coverage, or combined plagiarism checking. Pick Is It AI if you want explained flags, a published false-positive rate you can actually cite in a dispute, and a strict no-storage privacy posture. Treat Pangram's 99.98% accuracy and 1-in-10,000 false-positive claims with care: they come from the founders' own non-peer-reviewed benchmark, and independent writers have documented different verdicts on the same text scanned twice. No detector's flag is proof, whichever you choose.
Frequently asked questions
Is IsItAI free?
3 scans/day with a free account, 1/day without
Is Pangram free?
4 scans/day free
Can either tool be fooled by paraphrasing?
Yes. Sadasivan et al. (2023) demonstrated that paraphrasing reliably defeats most AI classifiers. Treat any detector verdict as a screening signal, not proof.
Are AI detectors fair to non-native English writers?
Independent research (Liang et al. 2023) found high false-positive rates for non-native English writing. Always combine a detector signal with other evidence, never use it as a sole basis for an academic or hiring decision.
Which one should I pick if I have to choose one, IsItAI or Pangram?
Pangram vs Is It AI comes down to workflow versus evidence. Pick Pangram if you need detection inside Canvas or Classroom, multilingual coverage, or combined plagiarism checking. Pick Is It AI if you want explained flags, a published false-positive rate you can actually cite in a dispute, and a strict no-storage privacy posture. Treat Pangram's 99.98% accuracy and 1-in-10,000 false-positive claims with care: they come from the founders' own non-peer-reviewed benchmark, and independent writers have documented different verdicts on the same text scanned twice. No detector's flag is proof, whichever you choose.
Try IsItAI free
A free scan every day, three with a free account. No credit card. UK company, text never stored.
Sources and further reading
- IsItAI homepage
- Pricing
- How it works
- Pangram homepage (accuracy claims)
- Pangram pricing
- Pangram data privacy commitment
- Founders' technical report (self-published benchmark)
- Seed funding announcement (Business Wire)
- Independent criticism: repeat-scan inconsistency and base-rate risk (Requarth)
- Liang et al. (Stanford, 2023): GPT detectors are biased against non-native English writers
- Sadasivan et al. (2023): Can AI-generated text be reliably detected?