AI detection you can defend.
A score alone convinces no one. Get the verdict and the reasoning behind it - with the language detected automatically and the text analyzed as it was written. Processed with controlled infrastructure and clear privacy safeguards.
Is it AI? Check it now - free.
Paste 200–2,000 words. Language detected automatically. No account.
Processed on controlled infrastructure. Handling follows our Privacy Policy. No model training without authorization.
Featured in
Covered by the newsrooms of the languages we detect - from Le Monde and Le Figaro to Radio-Canada and La Presse.
Built within Canada’s AI research ecosystem
From IVADO to the Vector Institute - research-grade methodology, startup-grade speed.
The leading detection models are trained overwhelmingly on English text. When they score French, Spanish or Romanian, they're pattern-matching against a language they barely read - and their error rates climb exactly where you can't see them.
That's how a student in Montréal gets flagged for writing the way French is actually written. How an editor in Madrid kills a clean article. How an enterprise builds a compliance workflow on a signal that quietly degrades outside English.
We took the opposite path: language-aware detection built to work across languages, evaluated language by language, and explained in plain terms. Because a detector you can't question is just an accusation machine.
Built for the moment someone asks: “are you sure?”
Every result is designed to survive scrutiny - from a student, a client, or a board.
A result you can explain
No black-box verdicts. Every result shows the signals evaluated, the confidence level and the limits - so you can stand behind the call in a difficult conversation.
Read the methodology 02Language-aware by design
The language is detected automatically and analyzed directly, never machine-translated into English first. Detection that holds beyond the anglosphere.
Explore languages 03Private by design
Detection runs on controlled Google Cloud infrastructure. Retention and provider safeguards are documented clearly, and submitted text is not used to train public or general-purpose models without authorization.
How we protect your dataTrained against the models people actually use.
OpenAI GPT-5.5 
Anthropic Claude Opus 4.8 
Google Gemini 3.5 Flash 
xAI Grok 4.5 
Meta Llama 4 
Mistral AI Mistral Large 3 
DeepSeek DeepSeek V4 
Alibaba Qwen 3.7 Max Model-agnostic by design - because the text you're checking never tells you which model wrote it. Includes paraphrased and lightly edited output.
Built for the teams that can't get it wrong.
A wrong call costs something different in every job. Same detector, tuned to your stakes.
Education
The hardest conversation in teaching is “did you write this?” Walk into it with a reviewable, explainable report - not just a percentage.
Learn morePublishers & SEO
Verify contributor and agency copy at scale - before Google, or your readers, do it for you.
Learn moreEnterprise
Detection that passes procurement. Privacy, volume, SSO, DPA, GDPR and Law 25 - built to clear your legal review, not fight it.
Learn moreWriters
Prove your work is human before anyone demands it. Evidence for your voice, your byline, your reputation.
Learn moreWhat we won't promise you.
No detector on earth is 100% accurate - including ours, and including the ones advertising 99.98%. AI writing changes every month, and probabilistic tools have error bars. Anyone who hides that is selling you confidence, not detection.
That’s exactly why we built for explainability. You get signals and confidence levels, not just a number - so a human makes the final call. Treat every result as evidence for a conversation, never as a verdict on its own.
Questions buyers ask first
Are AI detectors accurate?
Good ones are accurate enough to be useful and imperfect enough to require judgment. We publish our methodology and per-language evaluation openly, and we design every result to be reviewed by a human - not treated as a verdict.
Which languages do you support?
All languages. The language is detected automatically and the text is analyzed directly rather than translated into English first.
Is my text kept private?
Analysis runs on controlled Google Cloud infrastructure. Retention and service-provider processing follow our published Privacy Policy, and submitted text is not used to train public or general-purpose models without authorization.
Which AI models can you detect?
The detector is model-agnostic and evaluated against today’s major families - GPT, Claude, Gemini, Grok, Llama, Mistral, DeepSeek and Qwen - including paraphrased and lightly edited output.
Do you offer an API?
Yes. The same detection engine is available via API for volume workflows, with enterprise options for SSO, DPAs and dedicated support.
Make the call with evidence, not a hunch.
Tell us your language, volume and privacy requirements - we'll show you exactly how the detector performs on your kind of text.