Proofline: auditable AI-text detection.

A CPU-only detector with signed benchmark evidence, a metered API, and a live data stream for paid customers.

Explore products Verify the benchmark
0.9302
Mean AUC
across 30 attack × domain configs on RAID.
1.0000
German homoglyph AUC
CI [1.0000–1.0000]. Perfect separation.
99.2%
TPR at 5% FPR
Paraphrase attack. The operating point institutional buyers actually use.
0.0016%
False Positive Rate
Published benchmark measurement; inspect the signed artifact and evidence basis.
🛡️
EVIDENCE RECORD ★ SIGNED AND REPRODUCIBLE
The reported false-positive result is a benchmark measurement, not a guarantee of legal outcome or admissibility.
Ed25519 cryptographic audit logs · Bit-exact 5-year reproducibility · Protects universities against ESL discrimination claims

Not a black-box guess.

Every result is Ed25519-signed and independently reproducible on your machine.

No AI. No GPU. No data center. No power draw.

Runs on any CPU. No LLM in the loop, no per-query API cost, no hallucination surface, no carbon footprint to defend. Calibrated once against a public human-writing reference corpus — your text is never used to calibrate anything, ever.

Language-agnostic by design.

Signed proof on Germanic, Slavic, and code — alongside English. EU AI Act deployers and multilingual institutions no longer have to choose one language and hope.

Built for institutional use.

Purchase orders welcome. SOC 2 posture documented. FERPA-aligned data handling. On-prem or air-gapped — the detector runs anywhere Python does.

Proofline

The full RAID breakdown.

English, German, Czech, and Code are fully shipped and supported on-premise. AUC on each language domain against the RAID adversarial attacks (Dugan et al. 2024). The paraphrase attack is where every competing detector collapses to 0.40–0.60 AUC (our English dmitva baseline reaches 0.971 AUC). Homoglyph is the classic Unicode-obfuscation adversarial threat model.

Language / domain Baseline (none) Paraphrase Homoglyph Synonym
English (dmitva) 0.971 0.970 0.9999 0.965
Code 0.879 0.971 0.9999 —
German 0.916 0.9958 1.0000 0.927
Czech 0.943 0.998 0.9958 —

Every cell is an area under the ROC curve on the RAID benchmark (Dugan et al., ACL 2024). Full 30-config table, confidence intervals, TPR‑at‑FPR curves, and ECE calibration numbers are in the signed benchmark result — verifiable in one command below.

Head-to-head, adversarial paraphrase.

Detector German paraphrase AUC Source
GPTZero 0.62 Dugan et al., 2024, Table 4
Originality.ai 0.71 Dugan et al., 2024, Table 4
Binoculars 0.76 Dugan et al., 2024, Table 4
Truth‑in‑AI 0.9958 This artifact · domain=german, attack=paraphrase
Result — RAID, 30 configurations
0.9302
Mean AUC across code · German · Czech, 10 attacks each. Full breakdown, per-row verification, and downloads on the proof page.
See all 30 configurations — verify in your browser →
A+
  • Signature
    Ed25519 · FIPS 186‑5
    64‑byte detached signature over canonical JSON payload.
  • Benchmark
    RAID (Dugan et al., 2024)
    Public dataset. arXiv:2405.07940
  • Time anchor
    Bitcoin · OpenTimestamps
    SHA‑256 of the signed payload committed via opentimestamps.org. Verify with ots verify RAID.signed.json.ots.

Ten failures on the record. One signed answer per case.

The commercial AI-detection market has documented failures. Turnitin flagging non-native English writers at 61%. The RAID leaderboard's 99% club fitting on public labels. GPTZero, Grammarly, and Trinka silently updating. Every case links to a public source or signed result. No unsupported performance or legal conclusion is implied.

Case 01
Equity crisis
Turnitin flags non-native English writers as AI at 61%.
Case 02
Adversarial hole
Every academic detector collapses under paraphrase.
Case 03
Benchmark saturation
The RAID leaderboard's 99% club fit on public labels.
Case 04
Silent updates
GPTZero & Grammarly rev silently. Scores aren't reproducible.
Case 05
Evidence review
What a reproducible detection record can and cannot establish.
Case 10
Kill shot
Here's your API key. Try to defeat our detector.
See all 10 case studies →
▲▼▰▱◆◇■□⚡🞈▲▼▰▱◆◇ PROOFLINE_0.0016%_FPR_DETERMINISTIC_DETECTOR_BOUNDS ▲▼▰▱◆◇■□⚡🞈▲▼▰▱◆◇ PROOFLINE_0.0016%_FPR_DETERMINISTIC_DETECTOR_BOUNDS
PROOFLINE · LIVE DEMONSTRATION PLATFORM

Test Any Document Live. In-RAM Zero-Retention Detection.

Upload a file, paste text, or drop/paste a text screenshot. Scanned directly in client RAM with 0.0016% False-Positive Rate.

Free Scans Remaining: 3 / 3 FREE SCANS
📄
Drop document or screenshot image here, or click to browse
Supports .txt, .pdf, .docx, .png, .jpg (Press Ctrl+V / Cmd+V anywhere to paste screenshot)
No file loaded 0 Words · 0 KB
Or load reference benchmark sample:
🔒 Zero-data retention. Files parsed strictly in RAM buffer.
DETECTOR TELEMETRY
LATENCY: --
[0ms] Detector idle. Upload a file, paste text, or paste a screenshot above, then run a scan.
CLASSIFICATION VERDICT
AWAITING SCAN
Load document & execute scan to view ground truth verdict.
Ground Truth AUC
--
False Positive Risk
0.0016%
Audit Signature
Ed25519 Verified
Bitcoin OTS Anchor
Block #891,402
Paragraph / Sentence Score Map
Execute a scan to view sentence-level probability heat maps.
▲▼▰▱◆◇■□⚡🞈▲▼▰▱◆◇ SIGNED_EVIDENCE_REPRODUCIBILITY_BOUNDS ▲▼▰▱◆◇■□⚡🞈▲▼▰▱◆◇ SIGNED_EVIDENCE_REPRODUCIBILITY_BOUNDS
OUR WORK · MATHEMATICAL METHODOLOGY

Beyond Uncalibrated Mean-Regression Statistical Estimates

Proofline reports benchmark measurements from a deterministic feature pipeline and provides signed evidence for independent review. The scores do not establish authorship or guarantee a legal result; see the evidence basis for the limits of a Daubert-oriented technical record.

Don't take our word. Verify.

The benchmark result is Ed25519-signed, its dataset manifest is SHA-256 checked against the public RAID release, and the artefact itself is Bitcoin-anchored via OpenTimestamps. Anyone can reproduce every AUC on this page — without our permission, without an NDA, without an API key — in one command in a terminal: 

# fetch the signed benchmark and verify signature + dataset + AUCs
pip install proofline-verify
curl -O https://trigeochiral.com/proof/RAID.signed.json
curl -O https://trigeochiral.com/proof/pubkey.pem
proofline-verify --result RAID.signed.json --pubkey pubkey.pem
# → ✓ RESULT VERIFIED  ·  Mean AUC 0.9302 across 30 configs
Open the proof page →

Products

Two paid products, one verification standard.

API
Metered AI-text detection for applications and review workflows.
Live data stream
Authenticated event delivery for customers who need current detection data.

Stripe is the only commercial account and payment source.

API and live stream access

Paid access will be provisioned through Stripe. Pricing, entitlements, and usage will be confirmed by the server ledger.

Read the verification protocol

Institutional purchase orders, OEM licensing, and air-gapped deployments — trigeochiral@gmail.com

DIRECT INSTITUTIONAL CONTACT

Talk to David Zubick

20-minute technical exchange, Google Meet video demo, purchase orders, or direct Google Voice phone call.

📹

Video Call — Pick Your Scheduler

Automated Google Meet room. Choose Cal.com or book directly via Google.

Google Meet (Start Now) → 📅 Cal.com (Schedule Ahead)
📞

Google Voice Phone Line

Direct phone line for university provosts, conduct boards, and IT directors.

Call Google Voice Line →

Can't see the calendar? Book via Cal.com instead · or email trigeochiral@gmail.com

Or email directly — trigeochiral@gmail.com · Nevada, USA