How do we interpret the results from an AI text scan?

Every AI text scan in Winston AI provides several metrics. Here is how to read each one:

Human Score — The overall probability (0–100%) that the full document reads as human-written. This is the number we recommend using for decisions.
AI Prediction Map — Sentence-by-sentence color coding:
- 0–20: AI Generated (red)
- 20–40: Likely AI (orange)
- 40–60: Uncertain (yellow)
- 60–80: Mostly Human (light green)
- 80–100: Human Written (dark green)
Sentence-level predictions are less accurate than the overall score because they analyze smaller chunks.
Readability Score — Based on the Flesch-Kincaid formula with a U.S. grade-level label. You can also check text in our readability score checker.
Sections Driving AI Score — Lists the specific sentences that influenced the prediction. Use this when reviewing documents with mixed authorship (human + AI edits).
Explain (AI Overview) — Click the Explain button to get an AI-generated narrative summarizing why the content scored the way it did and what to look for next.

Scan at least 300 words for a confident result.
Treat scores between 40–60% as inconclusive. Grab more text or use external context.
If only a few sentences are flagged, review those sections manually — they may have been rewritten or paraphrased from AI.

Was this helpful?