How AI Revenue-Sharing Platforms Are Finally Paying Creators

The narrative around generative AI and creators has spent the last three years dominated by lawsuits, scraping scandals, and existential dread. But in 2026, a fundamentally different ecosystem is taking root.[1]

Rather than fighting the tide of artificial intelligence, a growing coalition of artists, publishers, and tech startups are building the infrastructure for a "licensed web." This movement aims to prove that generative AI can thrive without relying on unauthorized data scraping.[1]

At the center of this shift is the concept of "fractional attribution"—a mechanism that tracks exactly which pieces of human-created content were used to generate an AI's response, and pays the original creators accordingly.[2]

Leading this charge is ProRata.ai, a platform founded by tech veteran Bill Gross. ProRata's consumer-facing search engine, Gist.ai, operates entirely on licensed content, ensuring that every answer is backed by compensated human journalism and artistry.[2]

How does fractional attribution actually work? When a user asks an AI a question, the system doesn't just hallucinate an answer from a black-box training run. Instead, it uses Retrieval-Augmented Generation (RAG) to pull relevant information from a licensed database.[1]

ProRata's proprietary algorithm then analyzes the final output, measures the proportional value of each contributing source, and distributes revenue on a per-use basis. If a response relies heavily on a specific magazine article, the ad revenue or subscription fee generated by that query is split to favor that publisher.[2]

The model is gaining massive traction. As of mid-2025, ProRata had secured deals with over 500 publishers, including major media conglomerates and specialized outlets like Fast Company and The Boston Globe.[2]

The momentum has even reached national levels. In late 2025, the Danish Press Publications' Collective Management Organisation (DPCMO)—representing 99 percent of Denmark's news industry—signed a landmark collective licensing agreement with ProRata.[3]

This Danish partnership establishes a replicable framework for media associations worldwide, allowing publishers to opt into AI answer engines on a non-exclusive basis while maintaining the right to remove their content at any time.[3]

But real-time attribution is only half the equation. For visual artists and musicians whose work is used in the foundational training of models, companies are pioneering retroactive compensation structures.[1]

Adobe's Firefly model represents the most prominent example of this approach. Since 2023, Adobe has been issuing annual "Contributor Bonus" payouts to artists whose stock photos, vectors, and illustrations were used to train its generative AI.[5]

The 2025 and 2026 Adobe payouts calculate compensation based on two factors: the all-time total of an artist's approved images used in training, and the number of traditional licenses those images generated over the previous 12 months. Adobe even lowered its minimum payout threshold to $1 to ensure smaller creators received their share.[5]

To ensure these ethical standards aren't just empty corporate promises, independent watchdogs have stepped in to verify compliance. The non-profit Fairly Trained, founded by former Stability AI executive Ed Newton-Rex, has become the industry's gold standard for ethical data sourcing.[4]

Fairly Trained issues a "Licensed Model" certification—often colloquially referred to by developers as a "vegan model"—to AI systems that explicitly avoid using unlicensed, copyrighted data.[4][6]

To earn the badge, companies must prove their training data was explicitly provided by creators for AI use, licensed openly, or firmly in the public domain.[4]

While these platforms offer a hopeful blueprint, the transition is not complete. The industry remains bifurcated between these new ethical upstarts and legacy AI giants. Yet, the financial viability of platforms like ProRata and the popularity of Fairly Trained certifications suggest that consumer and enterprise demand is shifting toward AI tools that respect creators' rights.[1]