Independent resource on AI content authenticity — detection, standards & policy
Community Forum →
AI Content Authenticity — Independent Resource

The signal in
a world of synthetic
content

Benchmarks, standards, and research on AI content detection — from perplexity scoring to C2PA provenance. Built for researchers, publishers, and policy teams.

Key Detection Signals
Perplexity
High
Burstiness
High
Vocab Diversity
Medium
Structural Patterns
High
N-gram Entropy
Medium
⚠ Policy Watch

The EU AI Act requires transparency labeling for AI-generated content as of August 2026. The US Executive Order on AI Safety mandates watermarking research for government-facing AI systems. C2PA adoption is accelerating across major platforms. See our standards coverage →

Benchmark Snapshot

AI Detection Tools Compared

Updated March 2026 · 2,400 samples · human-written and AI-generated text across 5 content categories.

Originality.aioriginality.ai
Accuracy
91%
False Pos.
7%
Latency
420ms
Type
text
GPTZerogptzero.me
Accuracy
87%
False Pos.
10%
Latency
380ms
Type
text
Writer.comwriter.com
Accuracy
84%
False Pos.
8%
Latency
290ms
Type
text
Copyleakscopyleaks.com
Accuracy
79%
False Pos.
12%
Latency
510ms
Type
text
ToolAccuracyFalse PositiveLatencyType
Originality.aioriginality.ai 91% 7% 420ms text
GPTZerogptzero.me 87% 10% 380ms text
Writer.comwriter.com 84% 8% 290ms text
Copyleakscopyleaks.com 79% 12% 510ms text
Full comparison & API docs →
Authenticity Standards

The protocols shaping content provenance

From cryptographic signing to spectral watermarking — the technical standards defining how authentic content is identified.

C2PA — Content Credentials

The Coalition for Content Provenance and Authenticity defines how metadata is cryptographically bound to media files. Adopted by Adobe, Microsoft, Google, BBC, and the New York Times.

Active
2021–present

AI Watermarking (NIST)

NIST AI 100-4 defines voluntary guidelines for watermarking AI-generated content. Mandated for government AI use under the 2023 Executive Order.

Guidance
2023–present

EU AI Act — Article 50

Transparency obligations for AI-generated text and deepfakes. Generative AI providers must label synthetic content and disclose training data summaries.

Enforcement
Aug 2026
Full standards coverage →
Glossary

Key terms

The vocabulary of AI content detection — from statistical signals to provenance protocols.

Perplexity PP
A measure of how "surprised" a language model is by a given text. AI-generated text has characteristically low perplexity — it follows statistically likely token paths.
Burstiness
The variance in per-sentence perplexity across a document. Human writing is bursty; AI text has unnaturally uniform sentence-level surprise values.
C2PA Standard
Coalition for Content Provenance and Authenticity. Defines how cryptographic manifests are embedded in media to prove origin and edit history.
Watermarking
Embedding imperceptible signals in AI output that survive editing and compression, enabling detection of AI-generated content even after humanization.
False Positive Rate FPR
The rate at which a detector incorrectly flags human-written text as AI-generated. A critical metric for academic and publishing deployments.
Deepfake Audio/Video
Synthetic media in which a person's voice or likeness is generated or substituted using AI. Detection uses spectral analysis and liveness signals.
Full glossary (30 terms) →
Research

Original studies

AI Humanizer Bypass Rates: 2025 Annual Survey

14 AI humanizer tools tested against 6 major detectors using 4,200 standardized samples. Bypass rates ranged 23–91% depending on tool pairing.

March 20264,200 samplesBenchmark

Domain-Specific False Positive Rates in AI Detection

STEM writing has naturally lower perplexity. Detectors calibrated on general corpora produce false positive rates as high as 34% on technical content.

Feb 2026890 samplesAcademic

C2PA Adoption Across Major Platforms: 2025 Audit

Audit of C2PA Content Credentials implementation across 12 major platforms. Adoption is fragmented — only 4 of 12 implement full manifest verification at display time.

Jan 2026Platform AuditStandards
All research →