AI-generated content takes a majority of time on a major platform
Will Spotify, YouTube, TikTok, or Netflix report that AI-generated content accounted for a majority of listening or viewing time on its service in any period?
Chance by Q4 2031
2%
Cumulative probability
- Latest forecast
| Year | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| 2026 | <1% | |||
| 2027 | <1% | <1% | <1% | <1% |
| 2028 | <1% | <1% | <1% | <1% |
| 2029 | <1% | <1% | <1% | <1% |
| 2030 | <1% | <1% | 1% | 1% |
| 2031 | 1% | 2% | 2% | 2% |
Model reasoning
Content made by artificial intelligence is nowhere near half of listening or viewing time on these services. Spotify says fully machine-made music gets well under 1% of listening. All four companies are labeling or demoting such content and rarely publish these figures, so I put the chance at about 2% by 2031. The biggest change would be machine-made video becoming most of what people watch on TikTok or YouTube, and the company saying so.
Where things stand (as of 2026-10-11)
None of the four companies has published anything that comes close to resolving this, and I found nothing qualifying dated on or after 2026-09-16. Every piece of evidence I found puts the actual share of AI-generated content far below half, and all four companies are publicly working to keep it down.
- Spotify: On 2026-08-11 it said "AI Persona" artists would be left out of all editorial and algorithmic recommendations by default (newsroom). Its head of artist policy said fully AI-generated music gets "well under 1 per cent of consumption" (ABC, 2026-07-16). That was an interview, so it could not resolve the question, but it shows the level. Spotify's fan-made AI covers and remixes product is still in preview (TechCrunch, 2026-08-04), even though its Investor Day described an "era of generation" (recap). Third-quarter results come out on 2026-10-22.
- Other music services, for comparison: At Deezer, fully AI-generated tracks were more than half of daily uploads at peak, but only 1–3% of plays (Deezer newsroom, 2026-09-16). At Qobuz the figure is 0.38% of streams (TechRadar). A flood of uploads has not turned into listening.
- YouTube: The CEO's 2026 letter says AI "will remain a tool for expression, not a replacement" and makes reducing "AI slop" a priority (blog.youtube, 2026-01-21). YouTube is also cutting off ad revenue to inauthentic content (TechTimes, 2026-07-15). A Kapwing study found 21% of the first 500 Shorts shown to a new account were AI slop. But Shorts are only part of total watch time, which is dominated by long-form viewing on TVs.
- TikTok: It has labelled more than 3 billion videos as AI-generated and lets users turn AI content down in their feed with a slider (UNGA post, 2026-09; newsroom, 2025-11-18). Kapwing found 59% of a new account's first 500 For You videos were AI slop (report, 2026-07). That is an independent study, it counts videos rather than time, and it covers only a brand-new account, so it cannot resolve the question. Also, TikTok's label covers content "generated or significantly edited" by AI. That is a broad label, which would not qualify unless TikTok separately reported the share that is fully generated.
- Netflix: About 300 titles used generative AI, mostly in post-production (Variety, 2026-07-16). The question excludes that kind of production help.
Reference class and base rate
I looked for cases where a major platform published the share of total listening or viewing time that comes from AI-generated content. There is no precedent. Large platforms publish counts of labels and removals, and executives give rough figures in interviews. The only streaming service that regularly publishes an AI share of plays is Deezer, which is not one of the four, and its figure is 1–3%. So the base rate for a company publishing a majority figure is effectively zero.
Paths to YES and NO
To resolve YES, two things must both happen: (a) AI-generated content really exceeds half of total time on the whole service worldwide, and (b) the company publishes a time-share figure that meets the strict definition. My rough numbers by service:
| Service | (a) Real majority by end-2031 | (b) Qualifying report, if (a) | Combined |
|---|---|---|---|
| TikTok | ~25% (largest current AI share; strong video models such as Seedance) | ~5% (backlash, broad label) | ~1.25% |
| YouTube | ~8% (TV and long-form viewing dominate) | ~8% | ~0.6% |
| Spotify | ~1.5% (under 1% today, active suppression, label deals) | ~25% (Spotify publishes a lot in its shareholder decks) | ~0.4% |
| Netflix | negligible | — | ~0.05% |
That adds up to about 2%. The incentive to publish runs against YES: a majority-AI figure would alarm creators, advertisers, rights holders and regulators. The EU's AI Act code of practice requires labelling, not reporting how prevalent AI content is, and figures for one region would not count anyway.
Shape of the forecast
I used quarterly hazards that rise over time: 0.01% in Q4 2026, 0.02–0.03% per quarter in 2027, about 0.05–0.07% in 2028, 0.09–0.11% in 2029, 0.12–0.14% in 2030 and 0.15–0.17% in 2031. The rise reflects how long real shares would take to grow, plus the lag before anyone reports them. I computed the cumulative probability in Python as one minus the product of (1 − hazard). It comes to 0.11% by end-2027, 0.73% by end-2029 and 1.88% by end-2031.
Strongest case against
The main risk to this forecast is very fast progress in AI video. If short-form feeds become mostly AI-generated by 2029–2031 and ByteDance or Google treats that as a success to highlight, the probability could be around 5%. I give this partial weight. Even in that world, a qualifying report still needs a time-based, whole-service, worldwide figure from the company itself that separates fully generated content from AI-assisted content. Platforms have never published that, so I keep the forecast near 2%.
Resolution criteria
Question. By the last day of each quarter, has Spotify, YouTube, TikTok, or Netflix reported that AI-generated content accounted for a majority of listening or viewing time on its service in any period?
Definitions.
- Service: one of the four named services, measured worldwide across its users and content, not the combined services of its parent company. A renamed continuation of the same service counts. A country, subscription tier, demographic, channel, genre, playlist, or product surface such as YouTube Shorts does not by itself count. YouTube excludes the separate YouTube TV service; TikTok excludes Douyin.
- AI-generated content: works whose principal audible or visible content the reporting company identifies as wholly or primarily generated by generative AI. Human prompting, selection, editing, or publication does not disqualify a work. Merely using AI for recommendations, scripts, captions, translation, dubbing, mastering, restoration, isolated effects, or other production assistance does not qualify an otherwise human-created work. A broad label covering both generated and AI-assisted or altered works qualifies only if the company separately establishes the share attributable to qualifying generated works.
- Majority of listening or viewing time: strictly more than 50% of the service's total measured listening time, total measured viewing time, or combined listening/viewing time, using a consistent numerator and denominator. The metric must cover the service's entire listening or viewing activity respectively, not only music, newly uploaded works, originals, or another content category. Use the company's reported audience-time metric and its ordinary treatment of invalid traffic; do not independently estimate bot activity.
- Any period: any specified, completed measurement interval, with no minimum duration. The Observation period convention applies. Forecasts, targets, and statements about an unspecified future period do not count.
Resolves YES if a qualifying company publication reports a share above 50%, explicitly states that a majority or more than half of the relevant audience time was spent on qualifying AI-generated content, or supplies directly comparable audience-time totals that establish a share above 50% by arithmetic. Company estimates count when presented as estimates of actual usage, including estimates from representative samples of the whole service. For a reported range, its lower bound must exceed 50%; a reported point estimate above 50% counts without requiring its confidence interval to exceed 50%.
Resolves NO / Exclusions. No qualifying report by the horizon means NO. Exactly 50%, “about half,” or “up to” a majority does not count. Shares of uploads, catalogue items, plays, streams, views, impressions, revenue, or users do not substitute for shares of time. A majority among labeled content, surveyed respondents, experimental cohorts, or selected feeds does not establish a service-wide majority. Counts of AI credits or synthetic-content labels alone do not count. Independent studies and press reports cannot resolve the question unless the company itself adopts and publishes the qualifying finding.
Resolving source. Company-authored publications only: Spotify's newsroom (newsroom.spotify.com), investor relations publications (investors.spotify.com), or regulatory filings; YouTube's official blog (blog.youtube), Google/Alphabet corporate reports or filings expressly reporting YouTube data; TikTok's newsroom (newsroom.tiktok.com), Transparency Center (tiktok.com/transparency), or official corporate reports expressly reporting TikTok data; and Netflix's newsroom (about.netflix.com), investor relations publications (ir.netflix.net), regulatory filings, or What We Watched: A Netflix Engagement Report. Other official company-authored reports on their own domains also qualify. This is narrower than Public confirmation (A2): third-party confirmation alone, executive interviews, and employee posts do not count. Apply the Resolution window convention and Resolution date (A3); revisions first establishing the threshold count on their publication date, subject to the Observation period convention.