Original insights by Barry Schwartz
30-second rundown
Key learnings:
- Google is testing payment for meaningful AI input: A small publisher pilot pays when content materially helps generate an AI answer, but not when it merely confirms a fact or appears as a link afterwards.
- The payout formula remains hidden: Search Console can show monthly earnings and payments, while publishers still cannot see which articles earned money or how Google priced their contribution.
- Original reporting may become a licensable asset: Businesses should strengthen first-hand research, evidence, and expertise now because generic summaries are unlikely to command value in an AI content market.
Barry Schwartz has spotted Google passing envelopes under the table, except the table is inside Search Console and the recipients are publishers whose work helps feed AI answers. Google is testing direct payment when publisher content contributes meaningfully to an AI response, a sharp break from the old bargain of free crawling in exchange for possible traffic.
The scheme is called the AI contribution pilot. It covers AI Overviews, AI Mode, Gemini, and other Google AI experiences, but it is early and limited.
1. Google Is Paying for Contribution, Not Mere Appearance
Under the reported terms, earnings accrue when a publisher’s content contributes significantly to generating an AI response. A page does not qualify merely because it confirms a fact already assembled elsewhere or appears later as a supporting link. Payment depends on meaningful contribution, not raw usage or citation.
That distinction matters because it quietly assigns different values to different kinds of content. A copied definition or reheated product summary may help confirm what Google already knows, while an original investigation, specialist explanation, or proprietary dataset can change the answer itself. The first is another tin can in the desert. The second may become part of the engine.
Publishers can accept the terms, track earnings, update payment details, or cancel through Search Console. Reporting cited by Schwartz says Google approached dozens of publishers, including smaller organisations and businesses outside traditional news.
Do not mistake that for an open programme. The number is unconfirmed, and Schwartz did not have access himself. His article relies on programme material and publisher accounts reported by Digiday, so this is an early test, not a finished commercial system.
2. The Dashboard Shows Money but Hides the Price List
Participating publishers reportedly see monthly accrued earnings and payment records in Search Console. The figures can differ because of tax deductions and payment thresholds.
The extraordinary part is what the dashboard does not reveal. Publishers cannot see which content triggered a payment or how Google calculated its value. Google reportedly bases payouts on its own estimate of contribution rather than a simple count of uses, leaving the price formula locked in a back room with the lights off.
This makes the pilot impossible to optimise with confidence. A publisher may see money but not know whether it came from one investigation or twenty explainers. Without article-level evidence, any editorial reaction risks becoming superstition with a spreadsheet.
The opt-out matters too. Publishers can reportedly leave, which creates at least a narrow commercial choice. Yet a choice made without a visible rate card is still a casino decision, and Google owns the chips, cameras, carpet, and arithmetic.
3. The Real Prize May Be a Market for Inference Data
Some participants reportedly see more value in the precedent than the early payments. They describe weekly conversations with Google and a possible marketplace for inference data, meaning content used while an AI system constructs an answer.
That possibility changes the strategic question. Publishers spent years treating traffic as the payment for content. AI answers weaken that exchange by consuming information while keeping the user inside the platform, so direct compensation could become a second route for value.
The durable asset is original knowledge that an AI answer cannot manufacture cheaply elsewhere. For a small business, that can mean first-hand test results, clear pricing evidence, documented customer outcomes, expert analysis, or useful comparisons grounded in real work. It does not require a newsroom. It requires publishing something that survives the question, “Could fifty generic websites say this without doing the work?”
Google has finally acknowledged, in the language it understands best, that some words feeding the oracle may be worth money. The envelope is real enough to notice, but still too thin and too mysterious to count.
Build evidence no one else owns. If the AI economy eventually pays for nourishment instead of merely raiding the refrigerator, you will have something worth putting on the menu.
