Mapping publisher value in the AI marketplace

AI licensing is quickly evolving from a series of one-off negotiations into a new marketplace for content. As publishers confront declining referral traffic, AI-generated summaries, and growing uncertainty around attribution, the stakes extend beyond individual contracts. The frameworks emerging now may help determine how publisher content is discovered, valued, and monetized in an AI-mediated internet.

A new report from the Open Markets Institute's Center for Journalism & Liberty examines different approaches to answering these questions. The authors argue that decisions about compensation, attribution, access, and valuation can shape publisher bargaining power for years to come. As a result, publishers may need to focus on more than individual deals. They may also need to consider the rules, standards, and market structures that develop around them.

Defining publisher value in AI licensing

The report challenges the practice of treating AI training and retrieval as separate markets. Instead, it argues that publisher content creates value across multiple layers of AI systems, including model training, fine-tuning, factual grounding, access to current information, and the credibility of AI-generated responses. The authors contend that compensation models that focus primarily on retrieval fail to account for the full range of publisher contributions.

Publishers also face a practical challenge because AI companies already use large volumes of publisher content to train their models. As a result, many negotiations focus on ongoing access and usage rather than past training activity.

Direct deals draw the most attention

The debate over publisher value is no longer theoretical. As AI companies seek access to trusted content, a nascent licensing market is emerging to establish pricing, rights, and compensation models. Although much of that activity remains confidential, direct publisher agreements have become the market's most visible expression. OpenAI has signed agreements with approximately 35 publishers, while Perplexity's publisher program includes around 20 outlets. Microsoft's Publisher Content Marketplace launched with eight invited publishers.

Several agreements involve significant payments. News Corp's reported five-year agreement with OpenAI carries an estimated value of $250 million. Thomson Reuters reported $33 million in AI licensing revenue, while People Inc. receives at least $16 million annually from OpenAI. Amazon reportedly agreed to pay The New York Times approximately $20 million per year. Most agreements remain confidential, however, which makes it difficult to compare terms across the market.

New intermediaries enter the market

Direct agreements represent only one part of the emerging ecosystem. Companies including TollBit, Sphere.ai, ScalePost, ProRata, Created by Humans, Defined.AI, and Cloudflare are also participants in the licensing market. These companies provide content marketplaces, attribution tools, crawler controls, analytics, and payment infrastructure.

Many of these services address an information gap between publishers and AI companies. Publishers often lack visibility into how AI systems access their content, how frequently they use it, and what value AI companies derive from it. Licensing platforms increasingly offer reporting tools that track crawler activity and content usage. Others allow publishers to block crawlers, redirect access through paid channels, or establish usage-based pricing models. The report notes that the AI licensing market does not follow a standard approach to pricing. Some companies charge for content access, while others charge for content usage.

Traffic continues to decline

The report examines the relationship between licensing agreements and referral traffic from AI platforms. TollBit data shows that publishers with direct licensing agreements initially received higher click-through rates than publishers without agreements. During 2025, click-through rates for publishers with agreements fell from 8.8% to 1.3%. Rates for publishers without agreements fell from 0.8% to 0.27%.

The amount of content AI companies collect far exceeds the traffic they send back to publishers. The report shows that OpenAI's crawlers scraped approximately 1,700 publisher pages for every visitor referred to a publisher site. Anthropic's ratio reached approximately 73,000 pages per referral, while Perplexity's stood at 369 pages per referral. AI-generated referrals accounted for just 0.04% of total external referral traffic. Traditional search engines continue to drive most referrals.

Most publishers remain outside the market

While several large publishers participate in licensing agreements, participation remains limited across the industry. Local newspapers, regional broadcasters, ethnic media, indigenous media, non-English language publishers, and many specialized outlets remain largely absent from the licensing market. Most agreements involve publishers with established brands and the legal resources necessary to negotiate directly with AI companies.

Compensation reaches only a small portion of the publishing ecosystem. Smaller organizations often lack the resources needed to pursue direct negotiations or participate in marketplace pilots. As a result, licensing revenue remains concentrated among a relatively small group of publishers.

The AI licensing market remains in its early stages, with no consensus around valuation, attribution, access, or compensation. Yet as the report authors point out that deal structures, pricing precedents, business models, and governance norms may become difficult to change once they become established. For publishers, the challenge is not only securing favorable terms in individual agreements but helping shape the rules that will govern how content is discovered, licensed, and monetized in an AI-mediated ecosystem.