Google pays about 100 publishers for content in its AI answers
A patronage scheme for roughly 100 publishers changes nothing for the millions of organisations that must still earn AI citations unpaid.
Key takeaways
- Google pays about 100 publishers for content used in AI Overviews, AI Mode, and Gemini, per The Information.
- Payments range from under $1,000 over months to more than $1 million a year, with no disclosed formula.
- This is a tiny, opaque patronage scheme, not a licensing market open to new entrants.
- The overwhelming majority of organisations, including most financial, multilateral, and industrial bodies, get nothing.
- Visibility in AI answers still has to be earned through structurally retrievable, well-corroborated content, not bought.
About 100. Out of the millions of domains Google's crawlers touch every day, roughly 100 publishers have a commercial arrangement covering content used in AI Overviews, AI Mode, and Gemini. The Information reports that payments range from under $1,000 across several months to more than $1 million a year, a spread wide enough to suggest Google has no fixed formula, or at least none it is willing to share. The Decoder, summarising that reporting, notes that some participants in the scheme don't know how their payments are calculated. Others, larger and presumably with more leverage, are holding out for better terms.
This is not a licensing market. It is a patronage system, and the distinction matters.
A genuine market would have transparent pricing, comparable deals, and some mechanism for a new entrant to negotiate terms. What Google has built instead looks like a short list of favoured suppliers, paid at rates that seem to reflect bargaining power and relationship history rather than any measurable contribution to answer quality. A publisher getting $900 over six months is not being compensated for value; it is being kept quiet. A publisher getting seven figures a year has probably hired a lawyer who threatened something uncomfortable. Neither outcome tells you anything about how AI Overviews actually select and weight sources, which remains the question that matters most to anyone trying to get cited.
The 99.9% problem
Google indexes and summarises content from a number of domains that runs into the tens of millions. A roster of about 100 paid publishers is a rounding error, not a policy. For every outlet with a contract, there are thousands whose text, data, and analysis get pulled into an AI Overview with no payment, no notification, and frequently no clickthrough. The Decoder's framing, that Google is paying "almost nothing" to "almost no" publishers, is not hyperbole. It is close to an accurate description of the base case for the open web.
For the institutions that read TCE, financial services firms, multilateral agencies, industrial groups, philanthropic and policy bodies, this should settle a strategic question that has lingered since AI Overviews launched: there is no commercial relationship coming to rescue your citation share. If you are not Reddit, not a handful of major news organisations, and not sitting on content Google considers irreplaceable, you are in the uncompensated 99.9%. The quiet payment scheme is a release valve for Google's most vocal and litigious critics, not a template for a future where content creators get paid for what AI answers borrow from them.
That has a practical consequence. Visibility in AI Overviews and Gemini responses is not going to be bought. It has to be earned through the same mechanics that govern organic citation today: structured, unambiguous content that models can lift cleanly, strong third-party corroboration (the kind that gets a claim repeated across multiple indexed sources), and enough topical authority that a model's retrieval layer treats your domain as a reliable answer to a specific question. A UN agency's statistical release or a reinsurer's risk methodology will not get a Google check in the mail. But it can still earn a citation if it is the clearest, most structurally retrievable version of that information on the web.