Different detectors disagree because each one draws its AI-versus-human line in a different place.
AI Detectors puts a sensitivity selector first, so you can see how lenient, balanced and strict thresholds change the same passage's label.
How to use AI Detectors
- Choose lenient, balanced or strict sensitivity before scoring.
- Paste an English passage of at least 40 words and analyze it.
- Compare how the label shifts as you change the sensitivity.
Use cases
- Seeing how threshold choice changes an AI-likelihood call.
- Checking whether a passage lands near a decision boundary.
- Explaining why two detectors can disagree on the same text.
Good to know
AI Detectors computes one statistical score and then applies your chosen sensitivity to label it: lenient flags only the most uniform text, strict flags more. That is why detectors disagree in general, since each tool picks its own threshold. The underlying estimate is an English-only heuristic and indicative only; it needs about 40 or more words and analyzes the first 40,000 characters of long text with a note.
Frequently asked questions
What does the sensitivity setting actually change?
It moves the threshold that turns the score into a label, so strict flags more passages as AI-leaning and lenient flags fewer, on the same score.
Why do AI detectors disagree with each other?
Each detector measures slightly different signals and picks its own cut-off, so the same passage can be labelled differently by different tools.
Which sensitivity should I trust?
None blindly. Use balanced as a default and treat the label as indicative, then read the passage yourself before drawing any conclusion.
Is the score affected by text length?
Very short text is unreliable, so the tool asks for about 40 or more words before it gives an estimate.