Detection that fits inside an editorial workflow.

The editorial problem

Academic journals, preprint servers, and editorial publishers face an AI-detection problem that is distinct from the classroom one. Submissions arrive from authors the editors don't know personally; the stakes are a published record that's hard to retract; the volume is moderate (dozens to hundreds of submissions per month at a mid-tier journal); and the category is still shaking out the ethics of disclosed AI assistance. A tool built for classroom use often gets the weighting wrong.

What we offer journals

• Citation-aware scoring. Reference lists and inline citations are excluded from the AI-likelihood calculation. The research-paper detector exposes this as a named preset; the same treatment is also available via the main detector with the research-paper profile selected.
• Section-level heatmap. Per-section scoring so the methods-section formal register doesn't pull the paper's overall score.
• Bulk API. Integrates with editorial systems (Editorial Manager, ScholarOne) via a CSV round-trip today; native integration on request.
• COPE-aligned workflow guidance. We don't dictate policy, but we provide a recommended editorial workflow that treats detection as one signal in a process that includes authorship-contribution declarations and an appeal procedure.
• Editorial-board briefing. A 45-minute briefing session with your editorial board on what the tool does, where it fails, and how to interpret outputs. Included for institutional journal subscribers.

Recommended editorial workflow

1. Author discloses AI involvement at submission, what was used, for which parts, and how. Most reputable journals now require this. Our tool works best when paired with a real disclosure requirement.
2. Detector runs on submission. Result is available to editorial staff; not shown to peer reviewers.
3. Editor reviews the heatmap on flagged papers. A uniform low-level signal likely indicates formal genre; a section hotspot is the case worth investigating.
4. Author is given a chance to respond before any formal action. This is the appeal procedure equivalent to what classrooms need.
5. If the paper moves to misconduct review, detection is one input to a COPE-aligned process, not the sole evidence.

What we won't do

• Provide a "is this AI-written: yes/no" verdict suitable for retraction decisions.
• Share one journal's submission data with another.
• Integrate with systems that would surface detection scores to peer reviewers; reviewers should assess work on its merits.
• Compete with Turnitin's similarity-matching product. We do detection; similarity is a separate problem.

Pricing

Journal subscriptions are quoted per submission volume, with volume brackets. Editorial-board briefings are included at the subscriber tier. Request a quote via /contact (institutional sales handles journal accounts).