Why Google's AI payment pilot produces tiny checks

Google has been running a small payment pilot for publishers whose content shows up in its AI-generated answers, and the checks are so small that some participating executives say the offers are not worth accepting. The deeper problem is not the dollar amount. The pilot is built on top of the same search index that ranks and indexes the web for ordinary results, and that index was never designed to measure who contributed what to a given answer. The result is a payments layer that cannot, by construction, tell a publisher what it is being paid for.

The pilot has been described in The Verge and Ars Technica as a small set of participating sites paid per piece of content that surfaces in AI Overviews, the natural-language summaries Google now places above its traditional search results. The payments are not tied to ad clicks or to traffic the AI answer drives back to a publisher; they are tied to Google's own assessment of whether the content meaningfully grounded an answer. That makes the program more like a discretionary license fee than a market-priced syndication deal.

The mechanics, as Digiday has documented, sit inside Search Console. Digiday saw screenshots of a new AI-contribution panel that lists which pieces of a publisher's content fed into an AI answer and a monthly payout figure. The panel does not show how the payout was calculated, only the total. Some participating executives told Digiday they valued the program for the relationship it gave them with Google's product team. Others said the offer was too small to bother with, and a few said the lack of an underlying formula made planning for the next quarter impossible.

Google's own public-policy post describes the program as a pilot partnering with websites whose content contributes to freshness and factuality through grounding, a process that uses external sources to anchor a model answer in verifiable information. The post says payment depends on Google's assessment of meaningful contribution, not on raw usage. That framing is the trouble. "Meaningful contribution" is not defined for publishers, and the assessment is not auditable, so the per-piece check looks closer to an honor-system tip jar than a metered license. A publisher who wants to grow next quarter's payout has no metric to optimize against, only a relationship to maintain.

Publishers who want their work out of Google's AI answers can use existing search controls, but those controls also pull a site out of regular Google search, which is the distribution most publishers still depend on. Several publishers told Digiday they had no clean way to exclude content from AI Overviews without losing organic traffic, and that constraint is the lever the regulators are now pulling on.

The New York Times has an active copyright suit over AI training data. Penske Media has sued Google over lost referral traffic tied to AI Overview display. The UK government has ordered Google to provide an AI opt-out that does not penalize organic search. The European Commission has an open antitrust probe into whether Google is fairly compensating publishers for content that surfaces in AI answers.

What regulators are pointing at, and what publishers are running into, is the same structural gap. A payment layer grafted onto a search index that ranks, de-ranks, and indexes the web for free will keep producing tiny, opaque checks, because the index it sits on does not know how to count contribution. What would move the needle is a separate AI indexing layer, auditable contribution metrics, and an opt-out that does not double as a search-engine ejection. The pilot is still live, the UK and EU actions are still open, and the lawsuits are still active. That is the window in which the design choice gets made.