AI Licensing and the Open Web: Can Small Publishers Benefit?

AI licensing could give small publishers a way to earn money when their work is used to train or answer questions with AI systems, but it cannot save the open web on its own. Its value depends on fair terms, meaningful participation, transparent use, and whether licensing income can offset the loss of readers and advertising.

Why AI licensing is part of the debate

Publishers have long made material available online in exchange for attention, subscriptions, advertising, or referrals. Search engines typically sent visitors to source pages; generative AI tools can instead summarize information in a response, sometimes reducing the reason to click through. The extent of that effect varies by product, query, and publisher, and the long-term impact is still uncertain.

AI companies need large collections of text, images, code, and other material. Some content is openly accessible, some is governed by licences, and some may be subject to copyright claims or disputes. Whether a particular use requires permission depends on the material, use, jurisdiction, and applicable law. A licensing agreement is one way to secure permission and compensation, but it is not a universal settlement of those questions.

For smaller outlets, the concern is not just whether a model learned from an article. It is whether businesses can capture the value of reporting without sending readers to the people who paid to produce it. Licensing could add a revenue stream while leaving that larger problem unresolved.

How licensing can work

A publisher or rights holder grants an AI company permission to use specified material under agreed conditions. A deal might cover a defined archive, new articles, images, or particular uses such as model training, retrieval, or displaying excerpts in generated answers. These activities are not identical: training uses material to develop a model, while retrieval systems may fetch and present material when responding to a user.

Contracts can set limits on duration, territory, attribution, data retention, and onward sharing. They may also specify whether content can appear in answers, whether the publisher receives usage reports, and how payment is calculated. Payment could be a fixed fee, a recurring amount, or a formula tied to a measure such as usage. The details are contract-specific; there is no standard rate or universal licensing model.

Some arrangements are negotiated directly with individual publishers; others involve a collective or intermediary representing multiple rights holders. Collective negotiation may reduce the administrative burden for small organizations, but a publisher should understand who controls the rights, how payments are allocated, what costs are deducted, and whether members can leave.

What small publishers could gain

• Revenue beyond ads and subscriptions: A licence could compensate publishers for uses that do not produce a conventional page visit.
• A seat at the table: Negotiation can establish conditions for attribution, reporting, and permitted uses rather than leaving those matters entirely to a platform’s default practices.
• Shared negotiating capacity: A trade group or collective may help smaller publishers participate in deals that would be difficult to negotiate one at a time.
• Support for original work: If payments are material and predictable, they could help fund reporting, specialist coverage, archives, and other costly work.

These are possibilities, not guarantees. A deal that pays a modest sum but permits broad reuse may be a poor trade for a publication with valuable, distinctive work. Conversely, a limited licence with clear reporting might be worthwhile even if it does not solve the publisher’s entire business problem.

What licensing cannot fix

Licensing does not automatically restore referrals. If an AI answer satisfies a reader without a click, a publisher may receive a fee but still lose visits, newsletter sign-ups, subscriptions, or advertising opportunities. A contract can require links or attribution, but those provisions do not ensure that users will follow them or that the placement will be prominent.

Nor does payment necessarily reach the people who created the material. A publisher may own or control some rights, while freelancers, photographers, wire services, or other contributors retain rights or have contractual claims. The chain of ownership needs to be checked before granting permission. A publisher also cannot license material it does not have authority to license.

There are broader issues that a private deal cannot settle: whether models reproduce inaccurate claims, how to identify sources, whether users can distinguish generated text from reporting, and how to handle material that should not be reused. Licensing can establish rules between parties, but it cannot replace public policy, competition, or accountability measures.

Questions to ask before signing

A small publisher should compare the expected payment with the value and scope of the rights being granted. Review the agreement with appropriate legal advice when possible, especially if it covers an archive, multiple formats, or work contributed by others. Useful questions include:

• Exactly which works and rights are covered, and are future works included automatically?
• Does permission cover training, retrieval, generated answers, or all of these?
• How are payments calculated, reported, audited, and distributed among rights holders?
• Can the publisher withdraw, exclude certain content, or prevent sublicensing?
• What attribution, links, correction processes, and limits on data retention are promised?
• Does the agreement restrict licensing to other companies or affect existing contracts?

It is also important to distinguish a licence from a platform’s terms of service or a technical opt-out. Blocking a crawler may affect future collection, but it does not necessarily remove material already obtained or settle questions about past use. The practical effect depends on the system and the relevant agreement or law.

Other ways to support the open web

AI licensing is one tool among several. Publishers can continue investing in subscriptions, memberships, donations, events, syndication, and direct relationships such as email newsletters. These approaches do not depend on a single AI company, although each requires effort and may suit different audiences.

Industry groups can develop shared licensing terms and advocate for better disclosure and source attribution. Technical measures, including machine-readable preferences, may communicate a publisher’s choices to some services, but adoption is not universal and such signals are not a substitute for enforceable agreements. Public-interest funding and reader support can also help sustain reporting whose value is not captured by clicks alone.

The strongest strategy is likely to be diversified: pursue reasonable licensing where rights and terms make sense, while maintaining ways for readers to find, trust, and support the original publication. A healthy open web needs both sustainable publishers and routes from information back to its sources.

Frequently Asked Questions

Can small publishers license content directly to AI companies?
Yes, if they control the relevant rights and a company is willing to negotiate. A publisher should verify contributor agreements and any existing syndication or archive contracts before offering material.

Does an AI licence guarantee that readers will visit the source?
No. A contract may require attribution or links, but it cannot guarantee clicks, prominent placement, or that readers will choose to open the source page.

Can publishers opt out of AI training?
Some services recognize technical instructions or allow publishers to make requests, but practices differ. An opt-out may not cover every use or undo collection that has already occurred.

Is licensing the same as saying AI training is legal?
No. A licence is permission for uses within its scope. It does not determine the legality of other uses or resolve disputes involving material outside the agreement.

Will AI licensing save local journalism?
Not by itself. It could provide additional income, but local journalism also depends on dependable funding, audience relationships, and a continuing ability to reach readers.

Related

• Google Project Magi: What It Was and Its Bing Challenge
• Best Alternatives to Agent Skills for Portable Workflows
• K8sGPT Kubernetes Troubleshooting Without Unsafe Fixes