How Strives analysis works
Plain-language methodology and — just as importantly — its limits.
Automated analysis is a probabilistic signal, not proof of authorship or misconduct. Results can be wrong — students who write in an additional language may be affected disproportionately. Review drafts, writing history, sources, and the student's explanation before drawing conclusions. How this works.
What the signal is
AI-writing analysis estimates how closely a document's patterns resemble text commonly associated with AI-assisted writing. It produces a confidence band — Low, Moderate, or Elevated — together with the linguistic features that influenced the result. It is an automated indicator, not a determination of authorship.
What it is not
- It is not proof that a student used AI or committed misconduct.
- It cannot verify who actually wrote a document.
- It is not 100% accurate and offers no guarantees.
Known limitations
Students who write in an additional language, or who have a formal or formulaic style, may receive inaccurate results. Short texts are especially unreliable. Editing support, collaboration, and translation can all shift the patterns a model reads.
How to use it well
Treat any signal as a starting point for a conversation. Compare the submission with earlier drafts, look at the writing-process timeline, review sources and citations, and ask the student to explain their process before drawing any conclusion.
Provenance
Every result records the provider, model version, timestamp, and analysed word count, and can be re-run as the technology changes. During development the app uses clearly labelled demo data.