Teens are net positive across every category; ages 18–24 are net negative and lead in lapsed use at 19%.
For platforms and studios, the strategic priority is shifting from production savings to trust, rights and disclosure.
The mainstreaming of GenAI does not automatically translate into demand for AI-made content. More than half of U.S. consumers report regular or occasional use, yet engagement falls when books, news, music or video are known to involve AI.
The sharpest warning comes from ages 18–24: they are experienced users, but they have the highest lapsed-use rate and are net negative toward AI-assisted media in every category measured.
Jung Han
AI use is mainstream—but ages 18–24 lead in defection
In Luminate’s U.S. Entertainment 365 survey, 22% of respondents used GenAI regularly and 30% used it occasionally. The technology has moved beyond early adopters into everyday search, writing and media creation.
Among ages 13–17, 33% reported regular use and 36% occasional use. The oldest cohort stood at 8% and 29%, respectively. The most consequential group, however, is ages 18–24: 19% said they had tried GenAI but no longer used it, the highest lapsed-use rate of any age group.
This makes it difficult to explain young-adult skepticism as unfamiliarity. The data point instead to a post-trial challenge: product sameness, inconsistent quality, concerns about labor and copyright, or fatigue from AI saturation may matter. The survey does not establish the cause, so retention, lapse and reasons for lapse should become core industry metrics.
Figure 1. GenAI awareness and use status by age
Table 1. Key GenAI use-status indicators
Source: Luminate U.S. Entertainment 365, Wave 17, U.S. ages 13+, N=2,000. Sums use displayed values and may reflect rounding.
The 58-point fault line between teens and young adults
Net interest—share more interested minus share less interested—splits sharply by age. Ages 13–17 are positive for AI-scripted film and TV (+32), AI-produced music (+19), video games containing AI assets (+31), and AI-made social media imagery or video (+29).
Ages 18–24 move in the opposite direction: film and TV –26, music –29, games –10, and social media –29. The gap between the two cohorts reaches 58 percentage points in film and TV and in social media.
Treating ages 13–24 as one digitally native segment can therefore reverse a demand forecast. Interactive experimentation may work for teens; young adults may first require proof of human authorship, lawful data use and creator compensation.
Figure 2. Net interest in AI-assisted content by age
Table 2. Net interest in AI-assisted content
Note: Net interest = more interested minus less interested. Source: Luminate U.S. Entertainment 365, Wave 17.
Books at 71%, news at 65%: consumers defect when AI touches authorship
Bain’s May 2025 survey shows the highest negative response in books: 71% would not engage or would be less likely to engage. News and magazines follow at 65%, music at 62%, social media at 61%, audio at 60%, and video at 59%.
Text and dialogue-based media expose AI to the strongest expectations of authorship, factual responsibility and authentic voice. By contrast, video games have the lowest negative share at 40%, the largest neutral share at 44%, and the largest positive share at 16%. AI may be easier to accept when it expands an experience—dynamic dialogue, personalization and world generation—rather than replacing the perceived author.
Figure 3. Engagement likelihood after disclosure of AI generation
Table 3. Engagement likelihood by media type
Source: Bain Media Consumption Survey, May 2025, U.S. general population N=5,089. Unlabeled 2–3% segments derived to total 100%.
Style replication is already mainstream: images 38%, teens 49%
Among AI-aware consumers, 38% had used GenAI to create images in the style of a specific artist or author, followed by writing at 34%, video at 27%, and music at 24%. Among ages 13–17, the shares rise to 49% for images and 43% for video.
The behavior creates a double-edged market. Unauthorized style replication can dilute creator identity and compete with original work. The same demand can support licensed style models, authorized voices and characters, and fan-remix products with tracked usage and revenue sharing. The strategic question is shifting from whether imitation can be stopped to how consented, attributable and compensated creation can be productized.
Figure 4. GenAI creation in the style of a specific artist or author
Table 4. Creation in a specific artist’s style by age
Source: Luminate U.S. Entertainment 365, Wave 17.
Industry impact: calculate demand destruction before production savings
GenAI business cases have centered on lower cost and faster production. The surveys show that savings can be offset by audience defection, creator disputes and brand erosion. Invisible assistance—previsualization, localization, search and recommendation—must be evaluated separately from uses that replace the identity-bearing core of a product, such as scripts, reporting, songs, faces and voices.
For platforms, the relevant measures are post-disclosure click-through, completion and churn—not only unit production cost. Studios and labels must account for rights clearance and talent relationships; publishers for trust and correction liability; game companies for real-time safety and consistency.
Table 5. Industry opportunities, risks and priority responses
Source: K-EnterTech Hub analysis based on Luminate and Bain survey findings.
Platforms move from labels to detection and rights control
YouTube requires disclosure when realistic people, places or events are meaningfully altered or synthetically generated. In May 2026 it moved labels to more prominent positions and introduced automatic detection signals. A label alone does not reduce recommendations or monetization eligibility, while C2PA metadata indicating fully generated content can make the disclosure persistent.
YouTube expanded likeness detection to entertainment in April 2026 with agencies including CAA, UTA and WME. Spotify said unauthorized vocal impersonation is allowed only with artist authorization, reported removing more than 75 million spammy tracks in the prior 12 months, and backed DDEX-based credits that specify AI contributions. The U.S. Copyright Office has recommended federal protection against unauthorized digital replicas, while C2PA is building interoperable provenance infrastructure.
The direction is clear: responsible AI is becoming a rights-and-metadata stack, not a binary label. Delivery contracts will increasingly need provenance records, likeness permissions, process-level credits and machine-readable usage terms.
The next market: official AI licenses and machine-readable rights
An official license can let fans use an authorized voice, visual style, character or story world within defined boundaries, with different prices for private creation, public sharing and commercial use. This converts uncontrolled demand into a governed product.
The infrastructure must identify training and generation inputs, permissions, output identifiers and revenue allocation. For content companies, a machine-readable rights graph may become as important as the size of the IP library itself.
Platforms can turn compliance into a feature by displaying verified rights badges, original-creator links, permitted remix scopes and payment records. Transparency can become a demand-enabling product rather than a warning label.
K-content strategy: sell human authorship and authorized expansion together
Korean studios, labels, webtoon companies and game publishers should prepare two products at once: premium originals with provable human authorship, and authorized AI extensions that let fans and partners participate within licensed boundaries.
International sales packages should include the production steps using AI, model and data provenance, performer and creator consent, the final human approver, and downstream reuse rights. This reduces contract friction and improves auditability for global platforms.
Segment the message. For teens, emphasize interactive characters, story participation and licensed remixing. For ages 18–24, lead with creator compensation, lawful data and quality control. For older audiences, explain the scope of AI and human review in simple terms.
The winning capability is not content volume but trusted choice: consumers can distinguish human originals, AI-assisted works and licensed remixes, while creators can verify consent and compensation.
Methodological cautions
Luminate and Bain use different samples and question wording; their figures should not be combined as one population trend. Net interest does not show the size of the neutral group. Displayed shares may not total 100% due to rounding. The association between lapsed use and negative content interest does not establish causality. Style-generation questions refer specifically to prompting in the style of a named artist or author, not to all forms of AI-assisted creation.
Sources
Bain Media Consumption Survey, May 2025, U.S. general population N=5,089, as cited in the supplied source graphic.
Luminate, “How U.S. Audiences Feel About Gen AI in Entertainment Content”
Luminate, “Generative AI in Music, Film & TV 2026”
YouTube, “Improving AI labels for viewers and creators”
YouTube, “Expanding likeness detection to the entertainment industry”
Spotify, “Spotify Strengthens AI Protections for Artists, Songwriters, and Producers”
Spotify, “Artist-First AI Music Products”
U.S. Copyright Office, “Copyright and Artificial Intelligence”