Which streaming services ban AI music in 2026, and what does it mean for your playlists?

No major streaming service bans AI music outright, but by September 2026 every one of them treats it differently from human music: Spotify hides self-declared or detected "AI persona" artists from all recommendations, Deezer tags AI tracks and deletes the ones nobody streams, Tidal pays no royalties on fully AI-generated songs, Beatport rejects them at the door, and Bandcamp removes them on suspicion. For a listener the practical change is a set of badges and credits that did not exist a year ago; for anyone who runs a playlist it is a new due-diligence step before a track goes in. Here is the record, service by service, with dates.

Spotify: no ban, but AI personas fall out of the algorithm

Spotify's position since its September 2025 policy has three parts: unauthorised voice clones and impersonations are removed, a spam filter catches mass uploads, duplicates and the 31-second tracks built to farm royalties, and the industry's DDEX standard lets labels and distributors disclose where AI was used. Spotify said it had removed more than 75 million spammy tracks in the previous twelve months (Music Business Worldwide). Its policy head Sam Duboff described the goal as "more accurate, nuanced disclosures" rather than a binary AI label, and Spotify's Charlie Hellman said the company was "not here to punish artists for using AI authentically".

Three things followed in 2026. In April, AI Credits appeared in song credits on mobile, showing which parts of a track involved AI when the distributor disclosed it; Spotify notes that a missing credit does not mean no AI was used, because not every distributor supports the field yet. Also in April, Verified by Spotify replaced the old blue check with a green one, verified by human reviewers, and AI-generated artist profiles were excluded from it. On 11 August 2026 came the AI Persona badge: artists can declare in Spotify for Artists that their identity is AI-generated, Spotify's own review and detection can apply a "Likely AI Persona" badge, and both kinds are excluded from editorial and algorithmic recommendations by default. Listeners can still follow them. Spotify's wording is precise: "this badge is about the artist's public identity, not about how the music was made", and as of early September the appeal process for wrongly flagged artists was still unpublished (analysis by Notnoise).

The quieter change may matter more. Music marketer Andrew Southworth, who works with hundreds of independent acts, showed data from fully AI-generated projects whose algorithmic streams collapsed months before the badge existed: one profile fell from roughly 323,000 streams a day to about 1,400 from late February 2026, another from a 10,000-a-day peak to under 100 by April, and on a mixed profile only the AI release lost its algorithmic traffic while the human releases kept theirs (Spotify Secretly Killed AI Music). That is one marketer's sample, not a Spotify statement, but it matches the shape of the policy: Spotify evaluates releases, and it prefers to leave detection to distributors and step in for outright fraud (Spotify and DistroKid's new AI tagging). For scale, Universal Music's estimate in its own earnings reporting put AI artists at under 0.5 percent of streams, with about ten of them among Spotify's top 92,000 artists (Sam Duboff interview).

Deezer: tag it, hide it, delete it if nobody listens

Deezer runs the most aggressive detection in the business. Its tool has been live since January 2025, it began tagging AI tracks in June 2025, and tagged tracks are removed from algorithmic recommendations and editorial playlists (Music Business Worldwide). The numbers explain why: in April 2026 Deezer counted nearly 75,000 fully AI-generated uploads a day, 44 percent of everything arriving, with 13.4 million AI tracks tagged in 2025; they earned only 1 to 3 percent of streams, and about 85 percent of the streams they did get were flagged as fraudulent (Deezer newsroom). By July the share had passed 50 percent, peaking near 90,000 tracks a day, and Deezer announced it would remove AI tracks used for stream fraud and any AI track with no streams for six months, claiming 99.8 percent detection accuracy (Deezer newsroom). A Deezer and Ipsos survey found that 97 percent of people cannot tell AI music from human music by ear (Deezer AI detection).

Tidal, Apple, Amazon, YouTube, SoundCloud, TikTok

• Tidal went furthest on money: from 29 June 2026 a track its detection identifies as wholly AI-generated earns no royalties, with labels to follow and impersonation material removed (TechCrunch).
• Apple Music launched optional Transparency Tags in March 2026 for artwork, recording, composition and video (Music Ally) and told distributors in August that tagging becomes mandatory later in 2026 wherever AI made "a material portion of the content"; the tags were still invisible to listeners at that point (AppleInsider).
• Amazon Music has published no AI music policy of its own; it partnered with Universal in December 2024 against "unlawful AI-generated content" and put Suno inside Alexa+.
• YouTube treats audio "generated largely by AI with minimal human involvement" as potential low-value content that can be demonetised, auto-labels undisclosed photorealistic AI video since May 2026, and extended its likeness-detection tool to celebrities in April 2026 (YouTube blog).
• SoundCloud had to rewrite its terms in May 2025 after a clause allowing uploads to train AI was noticed; the new wording requires explicit opt-in before any content trains models that replicate a voice or likeness (Music Business Worldwide).
• TikTok requires a visible label on realistic AI audio and video, applies automatic detection, and has carried C2PA content credentials since January 2025 (TikTok newsroom).

The two real bans: Bandcamp and Beatport

Bandcamp announced on 13 January 2026 that music "generated wholly or in substantial part by AI" is not allowed, that impersonating artists or styles with AI is banned, and that it may remove content on suspicion alone, with a reporting tool for fans (Keeping Bandcamp human). Beatport followed on 12 August 2026: fully or majority AI-generated tracks are rejected at ingestion and the rightsholder is notified, AI-assisted tracks with a human as primary creator are accepted but tagged, and detection runs through an expanded partnership with Beatdapp (Beatportal). Beatport's own DJ survey found 60 percent would not play AI tracks and only 8 percent were open to them (DJ Mag). CEO Matt Gralen's line sums up the industry's emerging split: "There is a difference between a tool that assists human creation and a system that replaces it entirely."

What the distributors do before a track even reaches a service

DistroKid, which distributes an estimated 30 to 40 percent of new releases worldwide, moved AI self-disclosure out of beta in May 2026: uploaders say whether lyrics, music or both were AI-generated and whether the artist is a real person, and the data feeds the DDEX credits; the system relies on honesty, applies only to new uploads, and publishes no penalty for false reporting (Digital Music News). In July 2026 CVC Capital Partners agreed to take a majority stake in the company (Digital Music News). Believe and TuneCore went the other way twice: in April 2026 they began blocking tracks made on "unlicensed" AI platforms, naming Suno, then on 8 September 2026 signed a global licensing partnership under which tracks from Suno's licensed models become distributable if the artist opts in, and participating artists' music can train Suno's next models (Music Business Worldwide, Billboard). Suno's v6 models, launched the next day, add audio watermarking and download limits designed to stop mass uploads to streaming services. RouteNote's guidance is typical of the rest: fraudulent streams from bots, click farms or "deceptive third-party promotion services" get revenue blocked, releases removed and, for repeat offenders, accounts closed (RouteNote).

The enforcement has reached the courts. On 29 July 2026 a North Carolina producer became the first person in the United States sentenced for AI-driven streaming fraud after uploading hundreds of thousands of AI songs and running bot accounts that produced up to 661,440 fake streams a day; he agreed to forfeit more than eight million dollars (heise).

What a listener actually sees now

• On Spotify: an AI Persona or Likely AI Persona badge on profiles, search results and track rows from mid-September 2026; AI Credits in the song credits where disclosed; a green check only on human artists.
• On Deezer: an AI-generated tag on the track, and none of these tracks in your recommendations.
• On Apple Music: nothing visible yet, although the tags exist behind the scenes.
• On Tidal: labels are coming; the money has already stopped.
• On Bandcamp and Beatport: nothing, because the tracks are not there.

What it means if you build playlists

The platform rules are about identity and fraud, not taste, so a playlist can legally hold AI tracks everywhere except Bandcamp and Beatport. The practical risks sit elsewhere. A fully AI project without algorithmic support brings no follower growth to a playlist and, on Deezer, may disappear from it after six months of silence. Tracks tied to stream fraud can drag a playlist into a manual review. And a listener who discovers an undeclared AI vocal in a "human" playlist tends to leave. Three habits cover most of it: check the credits before adding a track, run doubtful submissions through a free detector such as SubmitHub's AI Song Checker, and prefer disclosed AI use over hidden AI use, because disclosed tracks are the ones the services have decided to tolerate. Our own playlists, from Deep House to Study Music, are curated by hand and updated weekly; a track that cannot show who made it does not get in.