Naomi Clarke
Naomi is an AI detection researcher with 8+ years in language-model evaluation and text classification. She treats detection scores as probabilistic signals, explains why false positives cluster around non-native, highly formulaic and heavily structured writing, and tracks how results shift as new models such as GPT, Claude and Gemini releases change what generated text looks like.
Author Snapshot
- How detection scores are produced
- False positives and uncertainty
- Model-specific detection behaviour
- Auditing published accuracy claims
About
Naomi Clarke studies how AI text classifiers behave on real-world writing rather than on the benchmark sets they were tuned against.
She explains what moves a score — length, editing, translation, genre, the model that produced the draft — and why the same paragraph can score very differently across two detectors run minutes apart.
Naomi specializes in presenting detection output as evidence with error bars, never as proof of authorship.
Areas of Expertise
- Text classification and evaluation
- False-positive analysis
- Per-model detection behaviour
- Accuracy claim review
Editorial & Review Approach
Careful language throughout: state the uncertainty explicitly, avoid definitive authorship claims, and show the exact conditions under which a result degrades.
Writing Focus
Naomi's articles are written for:
- Readers trying to interpret an AI percentage
- Writers flagged by a detector who want to understand why
- Anyone evaluating the accuracy numbers a vendor publishes