Parag Agrawal, the former CEO of Twitter, is back with another shot at reshaping how the web works. His company, Parallel Web Systems, has unveiled Index, a platform designed to give publishers and content creators real visibility into how AI agents are consuming their work โ€” and a new mechanism to get paid for it. The launch partners read like a who's who of serious publishing: The Atlantic, Fortune, PR Newswire, PitchBook, Enigma, RocketReach, ZoomInfo, Alex Heath's Sources, Packy McCormick's Not Boring, and Mario Gabriele's The Generalist all signed on early.

Why This Matters Now

AI agents don't browse the web like humans. They pull from dozens or hundreds of sources simultaneously to complete tasks for users โ€” a behavior pattern that breaks every existing model for compensating content creators. "The core thesis of the company was that agents will use the web a lot more than humans, and as a result of that, everything about the web will change, both the technology and the business models," Agrawal told Fortune. Parallel already sells web access infrastructure to AI companies including Harvey, Notion, and Opendoor. Index is its attempt to build an economic layer on top of that infrastructure.

The Shapley Value Bet

Index's compensation model relies on a concept called Shapley value, borrowed from game theory. Rather than charging flat licensing fees or paying per citation, the system attempts to estimate how much any given source contributed to an AI agent's completed task and the overall value generated. A piece of content that is more unique or used in higher-value tasks should theoretically receive more compensation. It's an ambitious framework โ€” and one that depends on accurately measuring attribution across highly complex, multi-source outputs. Parallel intends to start by applying this model internally, within agents using its own tools, before expanding to external agent ecosystems.

Publishers Are Looking for Alternatives

The timing isn't coincidental. The New York Times sued OpenAI and Microsoft in 2023 over alleged unauthorized use of millions of articles in AI training. Dow Jones and the New York Post have sued Perplexity over similar claims. Last year, Anthropic agreed to a major settlement with authors in a copyright case. Against that backdrop, publishers are desperate for frameworks that don't require litigation to resolve. "AI agents are becoming the next major interface for accessing information, but the economics of the web have not caught up with that reality," said Nicholas Thompson, CEO of The Atlantic, in a prepared statement.

How This Differs From Existing Licensing Deals

OpenAI has pursued fixed-fee licensing agreements with publishers including the Associated Press, Axel Springer, and News Corp. Agrawal argues those deals won't scale to an agent-driven web. "If only a few large companies have access to the premium content and no one else does, how will anyone compete?" he said. Fixed-price arrangements risk locking out smaller publishers and smaller AI startups alike โ€” creating exactly the kind of concentrated power structure that undermines the open web's value in the first place.

The Competitive Landscape

Parallel isn't alone in trying to build new economic models for AI-web interactions. Cloudflare launched Pay Per Crawl in private beta last year, letting publishers charge AI crawlers a flat per-request price across their domain. The key difference: Cloudflare charges for crawl access, while Index ties compensation to the estimated value of the source's contribution and the agent's completed work. Both approaches aim to move beyond the binary choice of blocking AI crawlers or allowing free access โ€” but they solve different parts of the problem.

Key Takeaways

  • Parallel Web Systems' Index platform uses Shapley value from game theory to estimate how much each content source contributed to an AI agent's output and compensate accordingly
  • Launch partners include The Atlantic, Fortune, PitchBook, ZoomInfo, and independent creators like Alex Heath's Sources and Packy McCormick's Not Boring
  • Parallel has raised $100M Series B at a $2B valuation in April 2026, five months after its $100M Series A at a $740M valuation
  • The platform will initially apply within Parallel's own agent ecosystem before expanding to external agents

The Bottom Line

Index is an intriguing attempt to solve a real problem โ€” but the devil is entirely in the implementation. Shapley value sounds elegant on paper; measuring it accurately across unpredictable, multi-source AI outputs is a different beast entirely. Publishers who've been burned by AI companies once already will be watching closely. If Agrawal can deliver on even a fraction of this promise, though, he's built something that every AI developer and content creator will need to pay attention to.