False positives and false negatives
A false positive flags a human-made recording as AI-generated. A false negative misses generated music. These errors concern different kinds of failure, so a single overall accuracy number cannot explain every result.
This service has no published, independently validated accuracy figure. Do not reuse another detector’s benchmark or a research paper’s result as the accuracy of Suno Checker.
Why results can change
The audio submitted, the sampled sections and the detector’s current model all matter. Re-encoding, editing and generators outside a detector’s evaluation data can change the pattern presented to it.
The cited research discusses robustness to audio manipulation and generalization to unseen models. It does not evaluate this website. Our methodology page separately documents what this tool actually sends and returns.
Read the score and the scope together
Suno Checker examines up to 30 seconds of audio. The combined score summarizes model output for those samples. A flagged-window share describes the sampled windows, not the whole composition.
Do not interpret AI detection signals 1 and 2 as isolated vocals and instruments. They are different detector outputs. Agreement between them may be useful context, but it does not create a verified probability.
A practical review sequence
First confirm that you submitted the intended recording and a supported file. Then read the overall result, each signal and the stated sampling scope. Finally compare the result with project files, source recordings, credits and the creator’s explanation.
If those sources disagree, retain the uncertainty. Avoid using a detector result alone to accuse an artist, reject a rights claim or promise that a distributor will accept a track.