Insights
Spotify Wants Every AI Agent to Play Your Song. Make It Easy to Ask For.
MusicPromoToday Staff · October 23, 2026
Share this article
Copy the link or spread it on social.

On October 22, Spotify told investors that the next front door to music isn't a search box. It's a sentence. Co-CEO Gustav Söderström described Spotify as "an agent that you can not only click on in our app, but talk to, ask questions of, and instruct," and said the goal is to be the media service people use "from any agent." It's already wired into ChatGPT, Claude and Meta's Muse. When a listener asks one of them for a driving playlist or for artists like someone they love, Spotify picks the songs.
That changes the job of a release campaign. You aren't only competing for a search result or a playlist slot anymore. You're competing to be the answer to a request nobody has typed yet: "something moody for a 6 a.m. run," "who produced this," "more like this but newer." Songs that are easy to describe, and backed by the listening signals Spotify reasons over, get picked. Vague songs don't.
What Spotify actually said about AI agents
From Spotify's Q3 2026 results and the earnings call:
- Scale: 788 million monthly active users and 306 million Premium subscribers at the end of September.
- Where Spotify lives now: it was the first service inside ChatGPT, it works with Meta Ray-Bans and OpenAI's Dots, and it was the only media service in Meta's Muse at launch. It also plugs into Claude.
- Who decides what plays: "We are not sending the music or anything to any of these agents." The outside agent "remote controls Spotify in the background," and Spotify does "the reasoning over everything we know about you and then decide[s] what to play."
- What listeners ask for: Söderström's example was sorting years of favorites by key and BPM and mixing them into a running playlist "at your exact cadence," then asking "exactly who wrote and performed what on each track."
- The signals: almost 40 million Premium users now listen to mixed playlists, mix consumption grew 90% quarter over quarter, and those mixes tell Spotify "exactly how and when" songs fit together. Spotify's large taste model is driving "double-digit growth in discoveries."
- The behavior: "tens of millions of people" are already talking to Spotify in plain English.
Two things follow. The listener's interface is moving to language. And the decision about what plays still sits with Spotify's own systems, fed by real listening behavior. Agents don't replace the fundamentals. They make every gap in them more expensive.
A prompt is a brief. Can your song answer it?
When Spotify launched inside ChatGPT, it told users that for best results they should "add details like genre, mood, or artist." Claude's Spotify connector suggests "Find artists similar to Khruangbin that I haven't saved yet." Spotify's Muse examples include a driving playlist "inspired by your destination" and a workout playlist that starts at 7 a.m.
Read those as creative briefs. Every one mixes four things:
- A moment: run, drive, focus, dinner, 2 a.m. after the club.
- A feel: mood, energy, tempo.
- A reference point: another artist, an era, a scene.
- A fact: who wrote it, who's on it, when it came out.
A song that fans, curators and writers can place on all four has somewhere to land. Most release plans answer one. "New single out Friday" is a fact. It isn't a moment, a feel or a reference point.
So before the campaign starts, write the song's brief in plain language, one line each:
- Where it lives: the two or three moments it genuinely fits.
- How it feels: mood and energy in words a listener would actually use, plus the honest tempo.
- Who it sits next to: three to five artists your listeners already play. Not your dream comparisons. The real neighbors in your audience data.
- What's true about it: writer, producer, featured players, the story behind it.
Then use that language everywhere the campaign touches: pitch, press, bio, creator briefs, ads.
Layer one: the signals Spotify reasons over
Spotify said plainly that it does the choosing. That makes the work inside the app matter more, not less.
Credits and metadata. If listeners can ask who wrote and performed what, the answer comes from the credits you delivered. Check songwriter, producer and performer credits with your distributor before release, not after. A missing credit is a missing answer.
The editorial pitch. The Spotify for Artists pitch form asks for genre, mood and culture tags and says "more detail gives your song a better chance." Pitch at least seven days ahead so the track lands in your followers' Release Radar. Use your brief's language, so what editors read matches what everyone else says about the song.
Context playlisting by real listeners. Playlists have always taught Spotify which songs belong together; mixes now teach it how and when. Your job is to get the song into real people's run, drive and focus playlists, not only the big editorial ones. Ask for it in fan comms: "add it to your running playlist" is a better call to action than "stream the new single." Saves and adds are the signal you're after.
Taste neighbors. If you want to be the answer to "more like X," your listeners need to overlap with X's. That's a targeting decision for your Spotify promotion and paid plan, not a hope. Point budget at fans of your real neighbors and keep at it until the overlap shows up in your audience data.
This is the same direction we flagged when Spotify let listeners see and steer their Taste Profile. Agents push it further: the listener says what they want, and Spotify matches it against everything it knows.
Layer two: the conversation before Spotify
This part can't be measured yet. A request to an agent often starts in the middle of another conversation: planning a road trip, building a training routine, asking about a scene or an era. The agent works with that context before it asks Spotify for anything. Spotify hasn't published how those handoffs work, and nobody outside these companies knows how much weight any input carries.
What you can control is whether the open web describes your music clearly. That's the same work that makes press and creator coverage useful anyway:
- PR that describes the sound. A feature or review that says what the record sounds like, who it sits next to and where it fits gives readers, and any machine reading along, a usable sentence. "An exciting new release" helps nobody. Brief writers with your feel and your neighbors. That's what good music PR should deliver now.
- Bio and release copy. Your artist bio, press release and website should use the same descriptors as your pitch. Specific, factual, consistent.
- Creator captions. In an influencer and UGC campaign, ask creators to name the song and the moment in the caption ("my 6 a.m. run song"), not just to use the sound.
None of this is a guarantee. It makes your music easy to describe, which is the job either way. We made the same case when YouTube Music taught fans to ask instead of search.
Wiring it into the release plan
Four to six weeks out: write the brief. Audit credits and metadata with your distributor. Build press angles around the sound and the story, not just the date.
One to two weeks out: pitch through Spotify for Artists at least seven days ahead, using the brief. Brief creators on the moments. Set paid targeting to your neighbor artists' audiences.
Release week: fan comms ask for adds to specific moment playlists. Creators post with the moment in the caption. Publish your own context playlist with Playlist Notes explaining why the songs sit together.
Weeks two to six: keep paid running against neighbors, feed fresh UGC into it, and give press a second angle: the making-of, a collaborator, the live version.
Segments: run separate creative for three groups. Existing listeners get asked for saves and adds. Fans of your neighbors get the reference point up front. Moment audiences (runners, drivers, late-night listeners) get the moment front and center. Our guide to Meta ad creative testing covers how to test those variants at volume.
How to measure it
There's no line in your dashboard that says "came from ChatGPT." Read it through proxies:
- In Spotify for Artists, streams from listeners' own playlists and library, other listeners' playlists and algorithmic playlists, before and after the campaign.
- Saves and playlist adds per listener.
- "Fans also like": are the neighbors you targeted starting to show up?
- The language: what words do fans, writers and creators use about the song? If they echo your brief, the brief is working.
Don't promise anyone agent-driven numbers. Nobody can attribute them cleanly yet.
What to do this quarter
- Write a one-page brief for your next release: moments, feel, neighbors, facts.
- Audit credits on your last three releases and fix the gaps with your distributor.
- Pitch every release at least seven days ahead, in the brief's language.
- Choose three to five real neighbor artists and point paid and creator work at their fans.
- Change your fan call to action from "stream it" to "add it to your [moment] playlist."
- Brief press and creators with the same descriptors.
- Build one context playlist of your own and use Playlist Notes.
- Track the proxies monthly and drop what doesn't move.
Agents won't make an undefined song discoverable. They'll make a well-defined one easier to find. MusicPromoToday has run music campaigns since 2012, with more than 3,000 campaigns across 40+ markets, for every genre and every territory. MusicPromoToday builds the PR, creator, playlist and paid plan around how your song is actually heard, so it has a clear answer when a listener asks. When you're ready, Plan My Release with the MusicPromoToday team.
FAQ
Can artists pay to be recommended by ChatGPT, Claude or Muse on Spotify?
No such product has been announced. Spotify says outside agents remote-control Spotify and that its own systems decide what to play.
Does Spotify send my music to these AI agents?
Spotify says it isn't sending music to the agents; playback happens in Spotify. When it launched in ChatGPT, it also said it would not share music or other audio and video content with OpenAI for training.
What should I fix first for AI-driven music discovery?
Your credits and your description. Make sure writer, producer and performer credits are complete, then write a plain-language brief (moments, feel, neighbor artists, facts) and use it in your pitch, press, bio and creator briefs.