One Hundred Checks From Mountain View: Inside Google's AI Contribution Pilot for Publishers

For two years, the bargain between Google and the open web has been dissolving in public. AI Overviews and AI Mode answer the question on the results page itself, clicks evaporate, and publishers watch referral traffic — the currency they traded their content for — lose value by the quarter. Google’s answer, historically, was some version of “the traffic is the compensation.”

That answer just changed. According to a report from The Information on September 30, 2026, Google has been paying approximately 100 digital publishers for their contributions to AI-powered answers in Search and other products. The payments flow through a quiet, invite-only program called the AI contribution pilot, administered through the tool publishers already know: Search Console.

What the program actually is

The mechanics, pieced together from Digiday’s earlier reporting on the pilot and this week’s scale disclosure, are simple from the publisher’s side:

• Opt-in, invite-only. Publishers are approached by Google and can join or decline; they can also opt out at any time.
• An earnings widget in Search Console. Once enrolled, a new “AI contribution” panel appears in the dashboard, displaying a monthly earnings figure with some history — and nothing else.
• “Pay-per-value,” not pay-per-use. Google only pays when it judges that a piece of content significantly contributed to an AI-generated response across Gemini, AI Overviews and AI Mode. Raw citation counts don’t earn anything; Google decides what “value” means.
• No formula disclosed. The payout calculation is entirely opaque. “It’s quite black box,” one executive with knowledge of the program told Digiday.

Google frames it as “an early-stage learning pilot to test how best to reward high-quality content,” pointing back to a June blog post about testing a new model for working with websites whose content keeps its generative AI answers fresh and accurate. The company has weekly calls with some program partners, who described the approach as “extremely collaborative.”

Why the number 100 matters

The scale is the story. When Digiday first reported on the pilot in mid-September, the participant count was murky — “at least dozens” had been approached, skewed toward small and mid-sized publishers rather than large ones, and Google wouldn’t confirm anything. A hundred enrolled publishers means the program has quietly moved past experiment territory into something resembling an early marketplace.

Notably, the roster is broader than news. Google’s earlier licensing negotiations were tied to News Showcase agreements with major news organizations; this pilot has widened to a much larger range of digital publishers. That suggests Google is testing a general-purpose compensation layer for the AI web, not a targeted peace treaty with the news industry.

The demographics of participation are telling, too. Large publishers — the ones with legal leverage and the option of suing, as several have over AI Overviews — have largely stayed out. Small and mid-sized publishers, for whom any recurring revenue for content they’re already producing is hard to refuse, are the ones inside the tent.

Peanuts, precedents, and a marketplace for inference data

The early reviews from inside the program are mixed in an instructive way. Nobody is getting rich. One source close to the situation told Digiday the early returns have delivered “peanuts” relative to participants’ advertising revenue, and one publishing executive described the numbers Google has offered as “lowball.”

And yet the publishers who spoke about the program were, on balance, glad to be in it. Their reasoning is strategic rather than financial: it is better to be inside Google’s AI licensing tent, helping to set the precedent for direct payments and data-sharing, than standing outside hoping referral economics return. “Do I wish they were more transparent? Definitely,” one program participant told Digiday. “But I’m hopeful that the fact that they’re setting a precedent for exploring paying publishers directly for content through this is meaningful.”

One executive described the pilot less as a one-off deal and more as an early test of a marketplace for inference data — a system in which publishers learn what content is valuable to AI systems, the same way Search Console taught the industry what performs in search. Google’s August launch of an AI performance report inside Search Console, which some publishers only began seeing in recent weeks, points the same direction: the feedback loop between AI surfaces and content producers is being built right now, and this pilot is its payment rail.

The strategic read

Three things make this moment significant beyond the hundred checks.

First, it is a de facto admission that “traffic as compensation” is dead as AI answers absorb the click. Google spent years defending that position in courtrooms and parliaments; a direct-payment pilot — however small — concedes the old bargain no longer holds.

Second, the black-box payout structure shifts a major pricing decision inside Google. With lump-sum licensing deals (the OpenAI-style approach), price is at least negotiated between two parties. With pay-per-value, one party both consumes the content and unilaterally determines its value. Critics quoted in earlier coverage called the scheme “a legal fig leaf” for exactly this reason: it looks like compensation while functioning as terms-of-service.

Third, it positions Google ahead of regulators and plaintiffs. Multiple lawsuits over AI Overviews’ impact on publisher traffic are pending, and the EU’s publisher-rights regime keeps tightening. A functioning (if opaque) payment program is a much stronger defense than “we send traffic” — and every month it runs, it generates data Google can use to argue the market is solving this on its own.

The open questions are the ones publishers keep asking: What is the formula? Who audits it? Can a payout that one side sets unilaterally ever be “fair”? Until Google answers them, the AI contribution pilot is best understood as the opening bid in the negotiation over what the web’s content is worth in an AI-first search economy — made to the hundred publishers least able to refuse it.