Publishers Are Finally Getting Paid for AI — Here's How Snowflake Made It Happen

For years, publishers watched Big Tech scrape their content to train AI models, with no payment and no permission required. Now a small but growing number of major news organizations are fighting back, quietly signing deals worth six figures to license their content directly to enterprise buyers.

The shift is real, but it's not coming from the AI companies themselves. It's coming from an unlikely middleman: Snowflake, the cloud data platform.

Snowflake Steps In as the Go-Between

Snowflake launched a product called Cortex Knowledge Extensions, which acts as a bridge between publishers and the companies that want to use their content inside their own AI systems. Think of it as a secure pipe. Enterprises can query a publisher's archive directly inside Snowflake's environment, without ever getting raw access to the underlying content, and without Snowflake taking a cut of what publishers earn.

So far, 17 publishers have signed on, including The Washington Post, Associated Press, People Inc., and USA Today Network. Others, including the Financial Times and The Economist, have signaled interest.

"You cannot scrape the data," said Ben Srour, principal product manager at Snowflake. "You can't steal it and use it for model training. So that's why the product has really resonated with publishers."

The Business Case Is Real, but Modest

The deals being struck are real. Publishers have signed several six-figure contracts with financial institutions, structured either as flat fees or usage-based agreements. Importantly, those payments often come out of money enterprises have already committed to spend on Snowflake, bypassing traditional procurement delays entirely.

For financial services firms, the appeal is practical. They're building internal AI tools that need continuous, trusted information feeds: supply chain updates, crisis signals, regulatory news, market commentary. A licensed feed from AP or The Washington Post is more defensible legally and more reliable in quality than scraped content.

AP's chief revenue officer Kristin Heitmann described the platform as offering "unlimited use cases" across finance, operations, crisis management, and environmental monitoring.

The Platform Fees Snowflake Is Undercutting

Not everyone competing in this space charges the same. Snowflake's zero-revenue-share approach stands out sharply against other intermediary platforms: ScalePost takes roughly 15% of publisher revenue, Cloudflare takes an estimated 30%, and ProRata.ai splits proceeds 50/50. For cash-strapped newsrooms, those fees are significant.

The Open Markets Institute's April 2026 report warned that these intermediary models risk recreating the same power imbalances publishers experienced with Google and Facebook. Snowflake's pitch is explicitly built around avoiding that trap.

A Revenue Stream, Not a Rescue Plan

The context here matters. The broader economics for publishers are brutal. Organic search traffic from Google is down 33% globally and 38% in the US over the past year, according to Chartbeat data. Google's AI Overviews have caused a 58% click decline for some publishers per antitrust filings. Publishers expect search traffic to drop another 43% by 2029.

AI licensing, the Snowflake team is careful to note, is a new revenue line, not a solution to the underlying structural crisis. The deals are incremental upside, not a replacement for the advertising and subscription revenue that AI products are quietly eating.

Still, for a publishing industry running short on options, the Snowflake model offers something meaningful: direct relationships with enterprise buyers, no middleman fees, and recurring revenue tied to how often AI agents query their content. As more enterprise teams build always-on AI tools that need live, trusted information, demand for that kind of licensed feed will likely grow.

"A year ago we were talking a lot about chatbots," Srour said. "But very quickly things are moving into this agentic, automated world. Agents are doing stuff at night all the time. What do they need? They need data. They need context. They need to know what's happening in the world."

Publishers, it turns out, still have something AI can't fully replicate: credibility.