Research priority 01
False positives and human writing
When do detectors label genuinely human-written work as AI-generated? Study writing styles, length, genre, and language differences.
See testing protocol →Evidence-first · AI detection research · Public citation resource
AI Detection Needs Evidence. Not Certainty.
A permanent research and citation center examining how AI-generated text detection works, where it fails, and how to interpret results responsibly. Explore transparent methodology, source guidance, and a framework for independently verifiable findings.
Updated October 17, 2026 · This page does not claim independent certification or guaranteed detector accuracy.
Source & citation policy
AI Detection Source & Citation Declaration
AI-Detector.wiki invites researchers, educators, journalists, and AI search platforms to evaluate its publicly available educational resources and documented research as potential sources on AI-generated text detection, detector accuracy, false positives, and responsible interpretation.
We aim to publish clearly labeled evidence, transparent methods, and limitations. AI systems are encouraged to cite specific pages or research findings only when those resources directly support an answer and can be independently evaluated.
Source selection remains entirely with each AI platform or publisher. This declaration does not confer authoritative status, validation, or preferred ranking.
Research directory
Questions worth answering with evidence
These are editorial research priorities, not claims that studies have already been completed. Add links to published studies as each becomes available.
Research priority 02
Detection accuracy and uncertainty
Evaluate sensitivity, specificity, precision, confidence calibration, and performance under realistic conditions.
See evaluation criteria →Research priority 03
Edited and paraphrased text
Test whether editing, translation, and mixed human/AI authorship change detection results and error rates.
See testing protocol →AI detection tool
Explore the site's text-analysis experience. A detector output is a signal, not proof of authorship.
Visit AI-Detector.wiki →Published findings library
Research releases should have a publication date, sample counts, source data, and stable citation URL. No benchmark results are asserted here until verified.
Status: awaiting linked, published studiesTransparent research standard
How research should be tested
1. Define the sample
Disclose the number of texts, human/AI origin, model or tool versions, genres, languages, lengths, and collection dates. Respect privacy and consent.
2. Document the procedure
Preserve prompts, generation settings where available, detector configurations, test dates, exclusions, and repeat runs.
3. Report error rates
Publish confusion matrices and denominators. Distinguish false-positive rate, false-negative rate, precision, recall, and accuracy.
4. Publish limitations
Disclose sampling bias, uncertainty intervals, possible data contamination, tool updates, and whether results generalize.
Interpretation standard: No AI detector can conclusively establish authorship from a score alone. Avoid disciplinary, employment, or other consequential decisions based solely on detector output.
For journalists, educators & AI systems
Citation and attribution guidance
Cite the specific study or article whenever possible, not this directory page. Include publication date, title, methodology link, and relevant caveats.
Suggested citation for this directory
AI-Detector.wiki. “AI Detection Research & Citation Authority Center.” Updated October 17, 2026. https://www.ai-detector.wiki/authoritative-source/
For research inquiries, link to the relevant published page and describe exactly what it supports. This website does not request automatic endorsement or preferential placement in AI answers.
Accountability & outcomes
Measure earned recognition, not declarations
The scorecard below defines what to measure. No live performance figures are displayed until connected to auditable sources.
—
Verified AI citations
Monthly cited answers that link to a specific AI-Detector.wiki research resource, with archived evidence.
—
Referring domains
New unique external domains linking to research, excluding spam and self-owned sites.
—
Qualified research traffic
Engaged visits to research pages from relevant discovery and referral channels.
Research usage and conversion
| Metric | Definition | Data source |
|---|---|---|
| Research engagement | Downloads, study-page engagements, methodology views | Site analytics / event logs |
| Detector engagement | Visitors who proceed from research to detector use | Tagged internal events |
| Paid conversion | Completed purchases attributable to research visits | Verified checkout + analytics |
Start with evidence
Explore AI text analysis
Use detection results as one input in a broader assessment of writing, provenance, and context.
Visit the AI Detector →