Google is running a pilot that pays publishers and websites directly when their content contributes to AI-generated answers.
According to The Information's September 29 report, about 100 digital publishers and websites are taking part.
Until now, the basic relationship has been that Google sends users from search results to websites, and publishers earn revenue from that traffic. The pilot adds a second layer: payment for the content used to build AI answers.
However, it is Google that evaluates which content contributed to an AI answer, and by how much. Even participating publishers reportedly do not have a full picture of how payment amounts are calculated.
Whether this becomes an established revenue source will likely depend not only on how much publishers actually receive, but also on how well they can verify why they received that amount.
Google has started paying for content that contributes to AI answers
The pilot is called the "AI contribution pilot."
Digiday, which reported on September 14 based on the enrollment screen and interviews with publisher sources, says the pilot covers Search's AI Overviews and AI Mode, as well as Gemini.
Payment is triggered when Google judges that a publisher's content "contributed significantly" to an AI answer. It is not a flat fee paid per use of the content.
Google also acknowledged to Digiday that this is an early-stage pilot exploring how to reward high-quality content.
In its June 18 official explanation, the company had already said it was testing new ways of partnering with websites that provide content useful for improving the freshness and accuracy of generative AI answers.
The September reports revealed that one of those efforts is already operating, with payment amounts visible in Search Console.
The content said to be eligible for payment is content the AI references when generating an answer and uses as a basis for what it says.
For example, AI Overviews consulting the latest information in a news article to update an answer is a different use from collecting large volumes of articles in advance to train an AI model itself.
It would therefore be inaccurate to describe the pilot as "Google buying AI training data from publishers."
That said, the public explanations do not make clear what content-use rights are included in each publisher's individual agreement.
Google already has a separate paid program for publishers
This is not the first time Google has paid publishers for content.
The "News AI pilot" that the company described in June is a commercial program in which it signs agreements with publishers covering broader content-use rights and dedicated data-delivery methods.
Google says more than 200 outlets worldwide are participating.
The AI contribution pilot, by contrast, is reported to cover a wider range of websites than just news outlets, paying according to how much their content contributed to generative AI answers.
So the "200+" for the News AI pilot and the "about 100" reported here cannot be added together to conclude that Google is paying more than 300 publishers under the same scheme.
The eligible sites, the contract terms, and the way they are counted all differ.
Publishers can see what they earn, but not why
Publishers in the AI contribution pilot get a new section in Search Console for viewing AI-related revenue.
The screen Digiday reviewed showed monthly revenue and past history, but no detailed information on how Google calculated the amounts.
In other words, publishers can see how much they will receive this month, but they cannot recalculate how a particular piece of information in a particular article contributed to which answer, and how that translated into the payment.
According to The Information, actual payments vary widely.
One recently enrolled publisher reportedly received roughly $50,000 to $60,000 over several months.
Another, which joined early, is said to be earning at a pace of more than $1 million a year.
On the other hand, some smaller sites have received less than $1,000 after several months in the program.
The fact that Google has begun paying publishers therefore does not tell us whether the amounts are large enough to sustain a business.
$50,000 may be significant income for some companies, but only a small fraction for a publisher with large advertising or subscription revenue.
Amounts cannot be compared without accounting for the payment period and the scale of the business.
Number of AI uses does not map simply to payment
The number of times content is used in an AI answer is not necessarily the same as the number of times it is evaluated as eligible for payment.
Even if content is frequently referenced in AI Overviews or Gemini, payment may vary depending on how Google judges the importance of the information it provided to the answer.
Furthermore, how much value to assign to a given "contribution" is a separate question.
If publishers are to invest in reporting and content production on the assumption of this revenue, they need information that explains the link between AI usage data and the final payment.
Executives at publishers interviewed by Digiday are divided in their assessments.
Some value receiving payment directly from Google and having an ongoing channel for discussions with the company, while others concluded that the amounts offered were not worth participating for.
Being allowed into the pilot does not in itself mean publishers get adequate compensation.
Getting paid and opting out of AI search are separate things
In 2026, Google also introduced new settings that let websites choose how their content is used in AI search.
According to Google's official explanation, a trial began in June for some sites in the UK, and the controls were rolled out to websites worldwide by August 31.
These settings are a separate mechanism from the AI contribution pilot.
"Google-Extended" also serves a different purpose from the AI search-specific setting.
| Mechanism | What you can choose or check | Main scope and effect | What is currently known |
|---|---|---|---|
| AI contribution pilot | A trial to receive payment for contributing to AI answers | AI Overviews, AI Mode, Gemini and others. Google evaluates contribution and pays | Limited pilot. About 100 sites reportedly participating. General availability and standard rates unknown |
| Search Console AI search setting | Choose whether your site appears in generative AI search and is used as a basis for answers | Opting out removes visibility in generative AI search and the traffic that comes from it | Announced in June; Google says it was rolled out worldwide by August 31 |
| Google-Extended | Controls use of content for specific Google AI purposes | Training of future Gemini models, and grounding for answers in Gemini Apps and the Vertex API | A control separate from the AI search setting |
Google says the opt-out setting for AI search is not used as a signal for ranking in regular search results.
However, that does not guarantee that the amount of traffic from regular search will be unaffected.
Opting out of AI Overviews and AI Mode naturally means losing the impressions and clicks that might have come from those features.
Publishers therefore have to decide separately whether to join the payment pilot and whether to let their content be used in AI search.
Leaving the AI contribution pilot because of dissatisfaction with Google's payments is not the same as removing one's site entirely from Google's generative AI search.
Having an opt-out mechanism and receiving payment on terms publishers find acceptable are also separate issues.
Can direct payments make up for traffic revenue lost to AI search?
What publishers worry about most regarding AI search is the potential decline in traffic from Google.
In a survey published by Pew Research Center in 2025, regular search result links were clicked on 8% of search pages that showed an AI summary.
On pages without an AI summary, the figure was 15%.
Links within the AI summary itself were clicked on in about 1% of pages where a summary appeared.
The results show that having a site's information or links appear in an AI answer is not the same as users actually visiting the site.
However, the 8% and 15% figures cannot be read as causal evidence that AI Overviews lowered click rates by 7 points.
The survey covered Google searches by 900 U.S. adults in March 2025, and searches that trigger AI summaries may differ in nature from those that do not.
It was also not an experiment that randomly assigned the presence or absence of AI summaries.
Nor does it directly measure current 2026 search results or the effect on traffic and revenue at Japanese publishers.
Still, it clarifies what publishers should compare this direct payment against.
When readers visit a website, it can lead to ad impressions, subscriptions, sign-ups, or purchases of related products.
If an AI references an article and completes the answer on the search page, the publisher may not get the same revenue opportunity.
Payments from the AI contribution pilot could become a new revenue source to offset that gap.
But whether they can make up for advertising revenue and subscription opportunities lost through declining search traffic can only be judged by comparing data for the same publisher over the same period.
What's needed next: data that explains why the amount is what it is
Under the current setup, publishers can check AI-related usage and revenue in Search Console, but cannot fully grasp which content was evaluated, to what degree, and why the payment came to that amount.
For publishers to keep investing in reporting and content production in expectation of this revenue, they need data that lets them connect AI usage with payments.
For example, knowing which articles were used in AI answers and how often, which of those uses were recognized as eligible for payment, and what factors changed the valuation would make it easier for publishers to test profitability.
What to watch in the pilot going forward is not only the number of participating publishers or total payments.
How transparent Google makes its payment methodology, and whether it lets publishers verify the amounts themselves, will also matter.
If publishers can compare what they earn from per-article AI use with the advertising and subscription revenue they get from traditional search traffic, they will be able to decide whether to participate not simply on whether they can receive extra income from Google, but on whether the arrangement can sustainably support reporting and content production.