What an AI detection score can—and cannot—tell you

An AI detector estimates whether patterns in a passage resemble its training examples. Its result is not a record of who wrote your document. Human writing can be flagged, and generated writing can go unflagged.

Read the label before the number. One provider may report a probability for a whole passage; another may report the fraction of eligible text it flagged. Those numbers do not mean the same thing. HumanWords labels the provider and the type of score it returns.

A result from one tool cannot predict an exact result from another. Model updates, text length, language, quotations, and subject matter can affect the output. Short passages provide less evidence; HumanWords requires at least 300 characters for its detector.

Use a score to prompt review. Check whether a passage accurately expresses your reasoning, whether evidence supports its claims, and whether you can explain the argument. Do not replace sound reasoning just to move a number.

Keep your notes, source material, and drafts. If your work is questioned, those records can help you explain your writing process. Follow the rules your instructor has set for AI assistance.

Further reading: Sapling detector documentation.