Keeping AI Music Detection Current: Suno v6 and ACE-Step 1.5

AI music generators do not stand still. New versions can arrive before the industry has even agreed on how to describe their sound, which makes a static detector less useful with every release.

That means there is nothing to install, migrate or reconfigure. The same upload flow now uses the newly validated model generation automatically.

What changed

We expanded the validation coverage behind ArtifactNet to include recent output from Suno v6 and ACE-Step 1.5. New coverage is evaluated before it becomes part of the model update, so a generator name is not simply added to a product page; it becomes part of the evidence used to judge whether the detector is ready to ship.

The release also updates the main detection model and its matched review layer together. Keeping those pieces aligned matters because an AI-music result should be judged by the full system, not by one isolated score.

What the benchmark showed at release

For the September 15 release we re-ran our ArtifactBench evaluation against the preceding production model. The new model improved both detection quality and the rate at which ordinary music is incorrectly identified as AI-generated.

The production model has been updated again since; this table records that release, not today’s model. Those numbers are one checkpoint, not a finish line. The goal is to keep validation current as generator families and distribution patterns change, then make the improved model available without asking customers to change how they work.

Why current coverage matters

For a label, distributor or creator, a detection result is useful only if it reflects the audio that is actually arriving today. Adding a new generation to the validation range helps keep reviews, catalog checks and internal research from being anchored to older examples of AI music.

ArtifactNet is designed for that continuing work: assess the tracks in front of you now, update the evidence as the field moves, and keep the product interface stable while the model improves underneath it.

How to check whether a song was made with Suno

1. Upload the track to the free detector. Open demo.intrect.io and add the audio file (MP3, WAV, FLAC and other common formats, up to 50 MB). No account is needed to start.
2. Read the probability and the line. The result gives a probability of AI and says whether it sits at or above the 20% line for AI. A score near the line deserves a second listen and, if you have one, a lossless copy of the file.
3. Treat the result as a signal, not proof. The score tells you whether the audio carries the patterns AI generators leave. Combine it with what you know about the track before you act on it.

FAQ

Can you tell if a song was made with Suno?

ArtifactNet checks whether a track carries the patterns AI music generators leave in the audio, and its validation range includes Suno v6. Upload the track at demo.intrect.io: the result shows a probability of AI and which side of the 20% line it falls on. A score is a signal to review, not proof of authorship.

Is the Suno AI detector free?

Yes, within limits. Without an account the demo analyses the first 2 minutes of each track, up to 6 minutes in total, and a free account adds 10 more minutes. No card is needed.

Can the result be wrong?

Yes. No detector is perfect: lossy encoding can move the score, and a human-made track can score as AI. Use a lossless file when you can, and treat a result close to the line as a reason to listen again rather than as a verdict.

Related reading

• How to Detect AI-Generated Music
• AI Music Detection for Labels and Distributors
• Audio Provenance vs AI Detection vs Watermarking
• Free AI music detector — check a track before you clean it.