C4Text analysisStylisticLayer 1 (Deterministic)

Anchoring Lack

Detects the absence of concrete references like dates, place names, specific measurements, and named individuals, details that ground text in reality and are hard for AI to fabricate convincingly.

Technical description

Counts and classifies concrete anchoring elements: temporal references (dates, years, time periods), spatial references (locations, institutions, laboratories), quantitative anchors (specific measurements with units, exact sample sizes), and named entities (researchers, instruments, reagents). Computes an anchoring density per 100 words and flags sections with zero anchors over extended spans.

How it works

Layer 1 (deterministic): Uses pattern matching patterns to detect dates, measurements with units, institution names, and proper nouns. Counts anchoring elements per paragraph. Identifies anchor deserts (spans of 100+ words with zero concrete references). Computes overall anchor density score.

Why this matters

Genuine research papers are anchored in specific experimental contexts, particular labs, equipment, dates of data collection, and named collaborators. AI-generated text lacks this grounding because the model cannot invent consistent, verifiable specifics. The absence of anchoring details across multiple paragraphs strongly suggests synthetic generation.

Score thresholds

0-1: Well-anchored text with frequent specific references
2-3: Moderate anchoring with some abstract passages
4-5: Almost entirely lacking concrete, verifiable references

Limitations

Theoretical and review papers naturally have fewer concrete anchors than empirical studies. Introduction and discussion sections may legitimately be less anchored than methods and results. The indicator may not recognize domain-specific anchors outside its dictionary.