AI is now everywhere in music: in the code behind new plugins, inside the plugins themselves, and in whole songs made from one line of text. Some of it makes you better. Some of it only makes you think you are. Here is the difference, and why the streaming platforms have started to care about it.
Written by M III D O · 30 October 2026
When someone says "AI plugin", they can mean two very different things. It is worth keeping them apart, because the pros and the risks are not the same.
The developer used an AI model to help write the code. You never see the AI. You get a normal EQ, compressor or synth, and it is only as good as the testing and the ears behind it.
The plugin itself uses machine learning while you work: it listens to your audio and suggests settings, cleans a vocal, or turns a sample into a synth patch. You see the AI every time you press the button.
Writing a plugin used to take years of learning. Now people describe what they want in plain words and an AI writes the code. Sound On Sound calls this "vibe coding" and walks through building a small step sequencer for Reaper this way.
The good: more people can try their own ideas, small and unusual tools get made that no big company would bother with, and a developer who already knows audio can test an idea in hours instead of weeks.
The bad: the same Sound On Sound article is honest about the limits. Some features in the author's own test simply did not work, the same request can give different code each time, and detailed, professional instruments are beyond what he expects this to do. A developer writing for Production Expert adds more risks for buyers: code that is not tuned for real-time audio (clicks, crashes, high CPU), new home-made licence systems, plugins that all look and work the same, and developers who disappear after the first release.
Based on the checklist in the Production Expert article. These questions work for any plugin, AI-made or not.
A plugin's reputation does not come from how it was made. It comes from how it sounds, how stable it is, and whether the developer keeps showing up.
Analog Obsession is a good example. Developer Rıdvan Küçük gives his plugins away for free and is supported by donations on Patreon, as Bedroom Producers Blog reported when the whole range went donationware in 2020. MusicTech praised his "trademark classy GUI style" and wrote that the model built him a dedicated audience. No big brand, no big marketing: the name came from the work.
There is a technical side to trust too. Many free plugin developers ship their files without a paid code-signing certificate, and a lot of producers use those plugins every day. Here is what a signature does and does not tell you:
So a signature tells you who made the file and that nobody changed it. It tells you nothing about the sound. And a good name does not make a file safe if you got it from a random mirror site. Download from the developer's own page, every time.
Some of the most-used plugins already have machine learning built in. In each case, the AI makes a suggestion and you make the decision:
None of these write your music. They save time on a job you already understand. If you don't understand why the assistant moved the EQ, it can't teach you anything, and you can't tell when it's wrong.
There is a healthy way to use AI and an unhealthy one.
This is my opinion, and I hold it strongly: no singer, songwriter, producer, or artist in any field becomes a real artist without their own ideas and real talent. You have to work on your abilities, your skills and your ideas, and you have to learn from your mistakes. A tool can help with all of that. It cannot do it for you.
Use AI to feed your knowledge, not to replace it.
Platforms are now separating artists who use AI as a tool from uploads where AI did everything.
I think this is fair. If all your music is AI output and you never try to learn or grow as a person in this industry, it isn't fair to be listed as a producer or an artist next to people who spent years building their craft.
No. It can make you a faster one, if you are already doing the work. Learn the basics, use the assistants to save time, listen to everything they do, and keep the ideas yours.
The tool is new. The work is the same: talent, ideas, practice, and learning from your mistakes.
Sound On Sound: Benjamin McCulloch, Vibe Coding: How Anyone Can Create Plug-ins (January 2026) · Production Expert: Deivids Hvostovs, Vibe-coded plugins: a developer's take and the risks worth knowing (24 August 2026) · Bedroom Producers Blog: All Analog Obsession VST plugins are now free (February 2020) · MusicTech: LOADES is the new free de-esser from Analog Obsession (16 February 2022) · Microsoft Learn: Microsoft Defender SmartScreen overview · Apple Support: Safely open apps on your Mac · iZotope: Shahan Nercessian, iZotope and Assistive Audio Technology (19 July 2018) · sonible: smart:EQ 4 · Waves: Clarity Vx · Sonic Charge: Synplant · Spotify newsroom (25 September 2025) · Spotify newsroom: One Year In, Putting Human Artistry First (20 October 2026) · Deezer newsroom (21 July 2026) · Bandcamp: Keeping Bandcamp Human (13 January 2026). The opinions in this guide are my own.