What is verifiable AI-text detection?
Verifiable AI-text detection reports whether writing is likely AI-generated as a signed, reproducible measurement — a result you can recompute and check yourself, rather than a score you must take on trust. Proofline is a verifiable detector: every result ships with a signature and the evidence to confirm it.
How is it different from other AI detectors?
Most detectors return a black-box score: a number with no way to audit it. A verifiable detector publishes a signed result and the artifacts to reproduce it, so an independent third party can confirm the same number. Proofline also keeps no generative model in the measurement path — the result is a measurement, not another model's opinion. See the full comparison →
Can you trust an AI detector's result?
Trust should come from verification, not reputation. In July 2023 OpenAI withdrew its own AI Text Classifier for a “low rate of accuracy,” and in August 2023 Vanderbilt University disabled Turnitin’s AI detector, citing “no insight into how it works” and false-positive risk. A result you can independently recompute doesn’t ask you to trust the vendor — it lets you check.
How do you verify a Proofline result yourself?
Every result is signed (Ed25519) and tied to published artifacts. Open the verifier, load the signed result, and it re-checks the signature and recomputes the published figures in your browser. If a number was altered, verification fails. Open the proof page →
Does it work on languages other than English?
Yes. Document Review mode is language-agnostic; Quick Check is tuned for English. This matters because independent research (Stanford, 2023) found many detectors misclassify more than half of non-native English (TOEFL) essays as AI-written, while rarely flagging US student essays.
Does Proofline track how I write?
No. Proofline is not a writing-surveillance tool — no keystroke logging, no typing playback, no version-history scraping. It measures only the finished text you choose to submit. Some tools try to prove authorship by recording how you write; Proofline verifies the text itself instead.
Is it accurate?
Accuracy only means something with a named, reproducible benchmark attached. Rather than a marketing number, Proofline publishes a signed benchmark result you can recompute yourself. See the figures and re-run them on the proof page. View the signed benchmark →
- — OpenAI AI Text Classifier withdrawn, July 2023: Search Engine Land.
- — Vanderbilt disables Turnitin’s AI detector, Aug 2023: Vanderbilt University.
- — Liang et al., “GPT detectors are biased against non-native English writers,” Patterns, 2023: arXiv:2304.02819.