What NoParrot Scores Mean
Every claim in a NoParrot result is scored by confidence level. Here's how to read the colors, percentages, and overall score.
Confidence Colors
Each sentence in a NoParrot answer is highlighted with one of three colors based on how many AI models agree on the underlying claim.
Verified
80–100%All models agree on this claim. High confidence — the fact is consistent across independent AI sources.
"The speed of light in a vacuum is approximately 299,792 km/s."
Uncertain
1–79%Partial agreement. Some models confirm this claim, others don't mention it or phrase it differently. Worth double-checking.
"Vitamin D supplementation above 2000 IU daily may reduce cancer risk."
Disputed
0–49%Models actively disagree. At least two models contradict each other on this claim. Treat with caution.
"The Great Wall of China is visible from space with the naked eye."
NoParrot Score (0–100)
The NoParrot Score is a single number that summarizes how reliable an AI answer is, based on all the claims it contains.
How it's calculated
Each factual claim gets a confidence value and a color. The NoParrot Score is the weighted average of all claim confidences: green claims contribute fully, yellow claims contribute half, and red claims contribute zero. Opinion claims are excluded from the calculation.
How to interpret it
High confidence. Most claims are verified across models. The answer is broadly reliable.
Mixed confidence. Some claims are verified, others are uncertain. Read the highlights to see which parts to trust.
Low confidence. Significant disagreement between models. Cross-check with authoritative sources before relying on this answer.
Limitations
NoParrot is a powerful tool, but it has boundaries. Understanding them helps you use it well.
Agreement ≠ truth
If all four models agree on something, it's more likely to be correct — but not guaranteed. Models can share the same training biases. A green score means consensus, not absolute truth.
Use as one tool among many
NoParrot helps you identify which parts of an AI answer are most and least reliable. For critical decisions — medical, legal, financial — always consult qualified professionals and authoritative sources.
Rapidly changing facts
AI models have knowledge cutoffs. For very recent events or fast-changing data (stock prices, election results), models may agree on outdated information.
Subjective topics
Questions about opinions, preferences, or values will naturally produce disagreement. A red score on "what's the best programming language?" reflects genuine subjectivity, not unreliability.
See it in action
Check your own AI answers and see the scores for yourself — no signup required.