We've been doing this since 2014.
Twelve years of plagiarism detection. From a Kyiv apartment to a San Francisco headquarters, through a New York accelerator, a million checks, and one deliberate rebuild. This is how Noplag got here — and where it goes next.
We started Noplag in Kyiv in 2014 because every plagiarism checker we could find was either institutional-only, prohibitively expensive, or a black box. None gave you a real explanation of why a passage was flagged.
We built a free version, opened it up, and learned what people actually needed: not just a score, but the source. Not just a number, but an audit trail. Twelve years and one rebuild later, that's still what we're trying to make true.
Twelve years in milestones.
- 2014
Founded
Noplag launched as a free plagiarism checker from a Kyiv apartment. First version checked plain text against Google search results via a rotating proxy pool. Got our first 100 users in three weeks.
- 2015
100,000 checks · first institutional pilots
Three small US colleges adopted Noplag for student work. Added .docx and .pdf upload support. Discovered that scanned PDFs need OCR — a problem we'd still be working on a decade later.
- 2016
Document conversion pipeline
Built the unoconv + pdf2html chain that converted any input format to indexed HTML. Sphinx full-text search backend. The state machine — Upload → Convert → Analyse → Search → Report — that survives into v5.0 was designed this year.
- 2017
Multilingual coverage
Added Spanish and Portuguese. Discovered through customer support that ESL writers were getting flagged more — the first signs of a problem the whole industry would confront years later.
- 2018
NY EdTech accelerator · $200K seed
Joined a New York-based accelerator focused on education technology. Raised $200,000 in seed funding from angel investors. Built the React 16 dashboard, the WebSocket-driven progress UX, and the first folder organization.
- 2019
1,000,000 checks · 4,700 free-tier users
Crossed the million-check threshold. Free-tier user base grew to 4,700. Began work on payment integration with Stripe and PayPal. The codebase from this era is the legacy v1 we rebuilt v5.0 from.
- 2020
Active development paused
We stopped shipping to think about what came next. ChatGPT happened. AI-generated content became the dominant question. Plagiarism detection without AI detection started looking incomplete. Watched the field, talked to customers, took notes.
- May 2022
Headquarters moves to San Francisco
Three months after the invasion of Ukraine, one co-founder relocated to San Francisco and re-established the company headquarters there. Engineering operations continued from Kyiv with the team that remained.
- Aug 2023
Co-founders reunited in San Francisco
The other co-founder joined the team in San Francisco after being medically dismissed from the Ukrainian Armed Forces, having served since early 2022. Planning for the rebuild began in earnest.
- 2024
Rebuild begins
Decided to start over with an open-source-first architecture. Picked Python + FastAPI for the engine, React/Next.js for the surfaces, and Apache 2.0 for the core. Made the bet that auditable beats black-box for the next decade of the market.
- 2025
Engine v1 · Layer W · cloud corpus
Designed the four-layer retrieval cascade (winnowing → MinHash → embeddings → reranking) and shipped the first layer, integrated 3 search engines, ingested Common Crawl + 8 academic indexes. MinHash (v1.1), embeddings (v1.2), and Binoculars-based AI detection (v1.2) are in development.
- 2026
v5.0 launch
LIVENoplag returns. Apache 2.0 engine on GitHub, cloud tier on noplag.com, four pricing tiers, dashboard from the ground up, and the Noplag Database — anonymized fingerprints that grow with every check. Here we are.
Three commitments that won't change.
Auditable
The detection engine is Apache 2.0. Read the code, fork it, run it on your own hardware. Trust shouldn't require taking our word for it.
Honest
AI detection is a signal, not a verdict. Every match shows its source. ESL writers get explicit disclaimers. We don't claim accuracy we can't measure.
Yours
Your documents are never used for training. EU residency available. Right-to-erasure is one click. The corpus moat is built on consent.
A small group that has been at this a long time.
Originally three engineers in Kyiv. Today we're headquartered in San Francisco with engineering and research teams in Ukraine, Poland, and Spain — plus one full-time librarian focused on the academic corpus.
Meet everyoneWant to be part of where this goes next?
Try the free tier, follow the engine on GitHub, join the Discord, or reach out about Enterprise. We answer everything.