Transparency
We show our reasoning. Every verdict comes with span-level evidence you can inspect, not a mystery number.
Our story
Founded by researchers who saw the wave coming
TrueText AI was founded in 2023 by a small group of NLP researchers and former educators who watched AI-generated text go from a curiosity to a daily problem in classrooms and newsrooms. The detection tools they tried were either black boxes or wildly overconfident — flagging honest writers and missing obvious machine output.
So they built the tool they wished existed: one that reports a real probability, shows its reasoning sentence by sentence, and is honest about its limits. What began as a research prototype now scans more than 180 million documents a year for schools, publishers, and content teams.
We're a remote-first team of fifteen across four time zones, united by one belief — trust in the written word is worth protecting, and the only way to protect it is in the open.
Mission & values
Our mission is trustworthy text
We exist to help people tell human writing from machine writing — fairly, transparently, and without punishing honest authors. These values keep us honest while we do it.
Fairness
We tune relentlessly against false positives, because a wrongly flagged writer is a real person who deserves better.
Privacy
Your text is never used to train our models, and scanned content is purged on a rolling schedule.
Scientific honesty
We publish our benchmarks and limitations. Detection is hard, and pretending otherwise helps no one.
Built for the real world
Fast, reliable, and easy enough that a teacher or editor can use it between two cups of coffee.
Human in the loop
We design for judgement, not automation. TrueText informs your decision; it never makes it for you.
A young company, measured in trust
2023Year founded
15People, remote-first
180M+Documents scanned
15+Languages supported
Our team
Meet the people behind the scanner
Researchers, engineers, and educators who care as much about fairness as they do about accuracy.
Elena Castillo
Co-founder & CEO
Former NLP researcher. Leads product and the detection roadmap.
James Mensah
Co-founder & CTO
Built the model pipeline that scores text in under three seconds.
Aisha Khan
Head of Research
Runs our benchmarking and false-positive reduction program.
Robert Tanaka
Head of Education
Ex-teacher who keeps the product grounded in classroom reality.
An honest note
What our accuracy numbers mean
No detector is perfect — and we won't pretend otherwise.
Our 99.1% accuracy and 0.4% false-positive rate are measured on a balanced, independently held benchmark of human and AI-written samples. Real-world text is messier: heavily edited AI output, non-native writing, and very short passages are genuinely hard to classify.
That's why TrueText reports a probability rather than a verdict, highlights its evidence sentence by sentence, and is built to inform a human decision — never to replace one. We recommend using our score as one signal among several, especially when the stakes are high. We publish our methodology and update it as models evolve.
Put TrueText to the test
Try it on writing you already know the answer to. We think you'll trust the result — and we'll show you why.