On July 30, Fortnite rolled out a new experimental feature called 'conversations' that uses AI to transform how non-player characters interact with players. Unlike scripted dialogues that repeat the same lines for every user, this system allows characters to respond dynamically to a player’s behavior and choices in real time.
From Matching to Making
Until now, recommendation algorithms have mostly worked by pointing users to content that already exists. Whether it was a movie, a song, or a game level, the algorithm’s job was to connect the user with the best match. Generative AI is changing the rules. Instead of just recommending, these systems are starting to create entirely new content tailored to a specific user’s needs.
In Fortnite, developers now use simple prompts to define a character’s personality and role. The AI then uses those inputs to craft unique, real-time responses based on the player's actions. This shift moves the experience from one-size-fits-all scripting to dynamic, player-specific interactions that adapt as the user plays.
Kaon AI, a startup focused on personalized video storytelling, has taken this concept further. The company generates a unique story world for each viewer in real time, instead of pulling from a static library of content. Its success has led to a $60 million Series B funding round, with the company aiming to prove that generative AI can create new, unmatched content. Kaon AI’s CEO, Jay Dang, argues this is the next step in personalization.
Netflix and Spotify Try AI Personalization
Netflix is one company testing how generative AI can enhance personalization through advertising. The streaming giant is now using AI to generate ad content dynamically. These ads are tailored to the visual context of the show a viewer is watching. Instead of the same generic ad break for all users, the AI ensures the content aligns with what the viewer is seeing, according to a May 13 report by Adweek. Brands like DoorDash, Target, and TurboTax have tested the feature and reported significant improvements in quality and performance. Netflix also plans to expand the technology to all regions that support advertising by the end of the year.
Beyond ads, Netflix is personalizing the ad experience itself. The company is testing ad loads and frequency controls that change based on a user’s viewing behavior. PPC Land reported on May 14 that this shift could lead to a more relevant and less intrusive experience for users.
Spotify is following a similar trend in audio. The company is developing a Large Taste Model trained on 3.4 trillion signals from listener activity across music, podcasts, and audiobooks. This model allows Spotify to generate personalized audio tracks and remixes, moving beyond simply recommending existing songs. Co-CEO Gustav Söderström discussed this strategy during Spotify’s 2026 Investor Day, emphasizing a shift from access to personalization to generation.
Challenges in AI Content Generation
Disney’s attempt to integrate generative AI into its services has been more difficult than expected. In December, OpenAI announced a $1 billion equity investment from Disney, aimed at allowing Disney+ subscribers to create fan-made videos using characters from Disney, Marvel, Pixar, and Star Wars via OpenAI’s Sora tool. However, just three months later, OpenAI decided to shut down Sora entirely, pivoting its focus toward high-productivity tools and agentic systems. Deadline reported on March 24 that Disney never moved forward with the deal and withdrew before any financial arrangements were finalized.
The initial agreement included strict guardrails to manage the use of AI-generated content. Characters could be included, but the use of talent likenesses or voices was not allowed. To ensure compliance, Disney and OpenAI set up a joint steering committee tasked with monitoring user-generated content against a detailed brand policy, as reported by Axios on Dec. 11. However, generating unique content for each viewer is a far more complex challenge than pulling from a pre-existing library. It requires significant computing power, real-time moderation, and systems that can handle errors and inconsistencies at a massive scale.