Publishers have been waiting for another round of licensing deals with artificial intelligence companies. The first few agreements were small but encouraging. They suggested the industry might finally get paid for the journalism that underpins so much of the output of AI large language models.

That optimism won’t survive 2026.

As long as Google’s search crawler and its AI-training crawler are functioning as a single system, the market for licensing journalism into AI models is effectively stalled.

The Search–AI bundle eliminates the market

Google’s power on the open web has always come from the same place: If you want anyone to be able to find you, you have to participate in its index. Until recently, publishers could assume that Google’s search indexing and its AI model training were separate processes governed by separate rules.

They aren’t.

Google has positioned access for model training as a condition, formal or informal, of maintaining your visibility in Search. The practical message to publishers is clear: If you want traffic, your content must be available for AI ingestion.

Every other AI company sees the same incentive structure. Why pay for something your largest competitor gets for free?

This is the reality shaping 2026. It has nothing to do with publishers being too accommodating, too fragmented, or not negotiating hard enough. As long as participation in search implies participation in training, the price for data is zero.

Robots.txt no longer works as a line in the sand

For years, publishers treated the robots.txt protocol as a boundary. It wasn’t perfect, but it let them believe they could draw a technical distinction between “indexing for discovery” and “scraping for training.”

With the workflows intertwined, blocking one means blocking both. The choice becomes binary: total visibility or total exclusion. Most publishers can’t afford the latter. So the wall they assumed they could put up is gone.

This removal of friction destroys the scarcity upon which a licensing market depends.

The small deals won’t add up

We will continue to see a few pilot agreements and modest partnerships between publishers and AI companies. They will generate headlines and offer some temporary relief to the publishers who enter into them. But they won’t make a liquid market. They won’t scale. And they won’t reset the economics across the industry.

They will be more like public relations exercises than true economic signals.

2026 is the year publishers stop waiting

Without a crawler split, there is no price discovery, no competitive tension, and no incentive for major AI companies to write large checks for recurring licensing fees. Regulators may eventually force a separation, but not on a timeline that affects this year. As a result, publishers will see no meaningful licensing revenue in 2026.

The companies that adapt fastest will be the ones that stop expecting that money.

Instead, they’ll invest in the parts of their business over which they still have control: direct audiences, owned platforms, first-party data, events and, increasingly, structured content feeds built for agentic AI systems, should those systems eventually require licensed inputs.

For now, there is only one question that matters to the future of AI licensing:

Has Google split its crawlers?

In 2026, the answer remains no.