Editorial AI-QA & Disclosure Kit

Run AI checks your editors can defend — without betting a writer's reputation on a detector score

Detector subscriptions charge $12.95–$99.99 per month and still hand you a score nobody can fully defend: OpenAI retired its own AI-text classifier over its “low rate of accuracy”, a major university disabled Turnitin's detector over ~750 wrongly flagged papers a year, and Stanford measured a 61.22% false-flag rate against non-native English writers. What agencies and publishers actually buy is the process: the policy, the QA workflow, the multi-detector consensus rules, the per-article evidence pack and the reader disclosure — the part that survives a client asking “how do you know?”. One payment of $42 — no subscription.

Get the kit — $42
One-time payment · ≈ THB 1,490 · free updates for v1.x · instant download

What's inside the kit

One zip, 12 files: a policy, a workflow, checklists, logs, disclosure texts and one small audit tool — plain Markdown and CSV you can read in any editor. No software to install, nothing to subscribe to.

Policy

A 3-tier AI-use policy, ready to adopt

None, assisted or generated — with a content-type matrix saying where each tier is accepted, writer and editor obligations, a client-facing summary for your onboarding docs, and enforcement steps. Anchored to APA and MLA credit guidance.

QA workflow

The pre-publish QA SOP

When to run a full check and when a light one, the 7 steps from file freeze to sign-off, the decision table, the AI-SUSPECTED conversation protocol — and the INCONCLUSIVE protocol that resolves on draft history, not on another score.

Consensus

Multi-detector consensus checklist

How to combine two or three detectors into one defensible reading — official APIs only, no scraping. A panel table of four evaluated vendors with date-stamped pricing, credit math and the quarterly re-check routine.

Evidence

Evidence pack + risk log + audit tool

One CSV row per article: scores, report IDs, plagiarism, fact-check, named reviewer, outcome, decision, disclosure. A risk log for judgment calls, and a small audit script that blocks incomplete records from being archived.

Disclosure

Disclosure statements that match reality

Reader-facing text per tier with a placement guide (on-article, site policy page, delivery note, colophon) — English plus Spanish, Brazilian Portuguese, German, Japanese and Thai. Two client-email templates included.

Handbook

The style-guide insert

The page that drops into your team handbook: four duties for writers, four for editors, three standing rules for the publication — English in full, core rules in five more languages.

What this kit will never tell you

No accuracy numbers. Not ours — this kit scans nothing — and none invented for the vendors either. The category's own history is the reason: OpenAI retired its own classifier in 2023 citing its “low rate of accuracy”, and the strongest commercial detector vendor runs a standing bounty for its own false positives. This kit sells process and records: the part that still holds up in a dispute.

Not a detector, not an engine. There is deliberately no scanning engine in the box. Your checks run through the official APIs of detectors you choose (the checklist evaluates four), combined by consensus rules — and a single score is context in the record, never proof against a person.

The process works in every language. Detector scores on non-English text or non-native English writers default to INCONCLUSIVE — Stanford measured 61.22% of human-written TOEFL essays flagged as AI — and the SOP resolves those on draft history and a conversation. Schools and universities are explicitly out of scope: this kit is for agencies and publishers reviewing commissioned work.

How you'll use it

From zip to a defensible QA record on a real article in about 90 minutes.

1
Day 1 — adopt

Policy + disclosure placements

Adopt the 3-tier policy, pick your disclosure placements, paste the style-guide insert into your handbook. The defaults are complete — you only tune the content-type matrix.

2
Week 1 — instrument

Detector panel + logs

Open accounts with two or three detectors through their official APIs, copy the evidence and risk CSVs to your editorial drive, and run the audit tool on the worked examples.

3
Every article — run it

QA → evidence → publish

Freeze the file, run the panel, screen plagiarism and facts, apply the decision table, record the outcome, disclose per tier, sign off. Quarterly: refresh the detector table.

One price. The process is yours.

No monthly plan, no credits to babysit, no per-seat fee — detector subscriptions already charge enough of those. Keep yours if you have one; this is the layer they leave out.

$42 one-time payment

≈ THB 1,490

Buy the kit — $42
  • Policy, QA SOP, consensus checklist, evidence + risk logs (English)
  • Disclosure statements: EN + ES / PT-BR / DE / JA / TH
  • Style-guide insert + evidence-audit tool + honest-limitations brief
  • Free updates for v1.x · 14-day full refund

Secure checkout via Stripe · card payments · taxes calculated at checkout

Frequently asked questions

Is this an AI detector?

No — on purpose. The detection market sells scores on subscription ($12.95–$99.99+ per month as of September 2026), and the history of those scores is rough: OpenAI retired its own classifier over low accuracy, a major university disabled Turnitin's AI detector, and Stanford measured a 61.22% false-flag rate against non-native English writers. What agencies and publishers lack is the defensible layer around whatever detectors they use — policy, workflow, consensus rules, evidence, disclosure. That is this kit.

Which detectors does it work with?

Any vendor you choose, through their official API — the kit enforces official channels (no scraping, no unofficial wrappers, no resold keys). The consensus checklist evaluates four vendors (Originality.ai, Copyleaks, GPTZero, Winston AI) with pricing captured 2026-09-20 plus re-check instructions, and stays vendor-neutral: the kit has no affiliate relationship with any detector company.

We publish in several languages. Still useful?

That is exactly the buyer it is built for. The process is language-neutral: policy, evidence, disclosure and sign-off work the same everywhere. Detector scores themselves are treated with suspicion outside English — non-English and non-native-English work starts at INCONCLUSIVE and resolves on draft history. The disclosure pack's key statements ship in EN, ES, PT-BR, DE, JA and TH.

Can a school use this on student work?

No — and the kit says so in its license and its documentation. It is built for agencies and publishers reviewing commissioned editorial work before payment and before publishing. Academic-integrity decisions are a different domain with different duties of care.

What exactly do I get?

A zip with 12 files: the 3-tier AI-use policy, the pre-publish QA SOP, the multi-detector consensus checklist, the honest-limitations brief with primary sources, the evidence-pack and risk-log CSV templates with worked examples, the disclosure statements in six languages, the style-guide insert, a small evidence-audit script (macOS/Linux) and the license. Plain Markdown and CSV except one shell script.

What is the refund policy?

If the kit doesn't fit how you work, write to us within 14 days of purchase and we refund in full — the files are yours to keep or delete. One refund per purchase; we only ask for optional feedback.

Do I get updates?

Yes — every v1.x update (detector-table refreshes, template fixes, policy additions) is free. Re-download any time from your download page with your email and password.

Judge the process, not a score.

$42 one-time · ≈ THB 1,490 · free updates v1.x · instant download

Get the Editorial AI-QA Kit — $42