Methodology and limitations

SolDetector evaluates writing-pattern signals at document and sentence level. Depending on the selected runtime, analysis may use a local heuristic method or an explicitly labeled experimental server-side model.

How to interpret results

• Elevated signals indicate stronger similarity to patterns learned or defined by the active method.
• Mixed signals indicate uncertainty or conflicting evidence.
• Lower signals indicate less detected evidence, not confirmed human authorship.

Known limitations

Formal prose, reference material, translated text, technical language, short samples, templates, and heavily edited writing may be misclassified. Generated text may also receive a lower signal after editing or paraphrasing.

Validation standard

No detector is presented as universally accurate. Production claims require independent, provenance-tracked evaluation, useful AI recall, controlled human false-positive rates, cross-domain evidence, and acceptable deployment cost.