Yahoo signs the first collective AI license from the ILC

A collective of publishers rebranded as the Internet Licensing Collective has announced Yahoo as the first licensee for the use of content in its AI-powered search engine, Scout. The system includes separate royalty pools, usage-based allocation rules, and governance open to publishers. An independent estimate suggests that between 15% and 35% of AI companies' revenues could be allocated for media compensation.

Over 3,000 publishers and a protocol already deployed

More than 3,000 publishers are now part of the Internet Licensing Collective (ILC), according to Leeds. Before this name change, the RSL Protocol already allowed publishers to specify machine-readable terms of service in their robots.txt files, including the option to request fees per crawl or inference. Leeds states that "thousands" of publishers are currently using this standard. RSL, which stands for Really Simple Licensing, referred both to a group of publishers working on a collective license and the organization behind this protocol. Leeds explains that a new name was desired for the nonprofit entity to reflect its collective nature and openness to all publishers.

An inaugural contract with Yahoo to power Scout

Yahoo is the first to sign a collective license with the ILC for the use of content in its AI-powered search engine, Scout. Publisher compensation depends on how frequently Scout uses their content. Yahoo pays a total licensing fee to access the collective's content, which is indexed to the revenue generated on the demand side, according to Leeds. The specific amount has not been disclosed, and Yahoo declined to respond to interview requests. RSL adopts the name Internet Licensing Collective on this occasion.

Segmented royalty pools to avoid dilution

The ILC has established a system of distinct royalty pools: each AI company signing a license contributes to a separate fund, without general pooling. Publishers only receive a fraction of the revenue generated from AI products that actually utilize their content. According to Leeds, this distribution method protects specialized publishers from revenue dilution, especially against generalist platforms like Yahoo Scout, which can aggregate content from hundreds or even thousands of publishers. In contrast, an AI product focused on healthcare might use a smaller volume of content from a few medical publishers, but with potentially much higher value for the healthcare company. The ILC aims for compensation to reflect the value of the AI product using the content, not just the quantity of content provided. Leeds cites the example of a low-volume but high-value healthcare vertical, whose compensation should not be diluted by content used by Yahoo.

Allocation rules: usage, display, and credits matter

Payments to publishers are determined by allocation rules that take into account the amount of content used, the nature of the use—fact-checking, retrieval, or display in an AI response—as well as the presence of a credit or source citation. Leeds gives the example of content used for fact-checking, which would receive a lower allocation than that of a detailed election results display explicitly crediting the Associated Press. The exact calculation formula is not disclosed. These rules aim to prevent AI companies from favoring the cheapest content and to encourage the use of sources deemed of higher quality.

Publisher steering: choice of licensees and visibility of uses

Publishers can choose which AI companies they wish to license through an ILC portal. For now, Yahoo is the only available option. When other licensees join the collective, publishers will be able to select or exclude companies to whom they want to license their content. If a publisher signs a separate licensing agreement with an AI company, they can withdraw their content from the collective. The portal is also expected to allow publishers to see how their content is being used, for example, for retrieval augmented by knowledge (RAG) or for training.

A potential 15% to 35% of AI revenues for the press

According to an estimate from Mather Economics, the collective license for AI could generate compensation for publishers equivalent to 15% to 35% of AI companies' revenues, with a potential market likely to surpass that of display advertising in the U.S. within three years. This projection is based on the current annual revenues of OpenAI and Anthropic, as well as a business model incorporating the laws of scaling AI training data. Leeds indicates that if a licensee's business grows, the amount available to compensate publishers can also increase, and the share paid progresses with revenues, allowing for the entry of new players, use cases, and products. He likens the functioning of the licensing collective to that of Spotify. On the other hand, Google and Microsoft are experimenting with value-based licensing models that compensate publishers when their content plays a significant role in the responses generated by AI, reflecting a shift from fixed-amount agreements to more flexible pricing. The ILC is in discussions with several major AI players, though names have not been specified. Leeds believes that a one-off agreement is positive but does not constitute an investment in the overall content infrastructure.