AI Writing Voice Detector

Forensic audit of a piece of writing for AI-generated cadence and lexical fingerprints. Output a categorized list of every AI-ism found and a single AI-Voice score from 1 (unmistakably human) to 10 (raw LLM output).

This is not a style critique. It's a fingerprint audit. Show receipts.

Inputs

The target is whatever text the user supplies — a paragraph, a chapter, a blog post, a tweet, a README. If the user points at a file, read it. If they paste text, work from the paste. If they don't specify, ask what to audit.

Workflow

1. Read the target in full. Do not skim. Cadence tells require reading sequence.
2. Tally each AI-ism category below. Keep counts.
3. Annotate every instance. Quote the offending text with its category tag (see Annotation Format). Do not skip "minor" ones, do not summarize, do not collapse repeats — list each occurrence individually so the writer sees the full receipt.
4. Compute the score using the rubric below.
5. Output the report in the Output Format below.

Do not rewrite the prose unless the user asks. The job is detection and scoring, not editing.