Detector validation

We will publish weekly detector data,
once the pipeline is instrumented.

Detector pass-rate benchmarks are easy to fake. We’re not going to publish numbers that haven’t been independently verified. This page is what we’re committing to, and what it will look like when it goes live.

What we will commit to publishing

  • A weekly run of our 200-essay corpus through 5 detector APIs (GPTZero, Originality.ai, Copyleaks, ZeroGPT, Sapling) every Monday at 06:00 UTC.
  • The corpus rotation schedule, so no essay is used for more than 13 weeks.
  • Per-run logs with timestamps, hashes, and per-essay verdicts.
  • The full methodology, downloadable as a PDF.

What we will not publish

  • Marketing-grade pass-rate screenshots not backed by verifiable runs.
  • Comparisons to competitors run from anything other than their paid customer tier.
  • Numbers that we cannot reproduce on demand.

Why this page is empty today

The pipeline is being built. We will not publish a single pass-rate number on this page until the cron job is live, the corpus is rotated, and an outside reviewer has audited a run. Until then, the integrity-first thing to do is to say so directly.

Subscribe below to be emailed the first time we run a verified test.

Get notified when detector data goes live

In the meantime

Our actual differentiation isn’t the detector pass-rate. It’s in the pipeline: voice fingerprint, citation-safe extraction, deadline-aware orchestration, and the signed Disclosure Bundle you can attach to your submission. That’s what the product is actually for.