Google's AI contribution pilot pays for answers, not clicks

Digiday reported on 14 September 2026 that Google has been quietly rolling out an "AI contribution pilot": a Search Console programme that pays some publishers when their content shapes an answer in Gemini, AI Overviews or AI Mode. By 17 September, 9to5Google had confirmed the detail that matters most to anyone running an AI search programme: the payout has nothing to do with whether a link ever appears.

That single design choice breaks an assumption sitting underneath every AI search measurement framework we know of, including the one in our own AI Search Attribution Handbook, published five days before Digiday's story ran. We built that model around a citation producing a link, and a link producing a referral row in GA4. Google's own pilot shows a case where neither of those things has to happen for the content to be worth money.

What Google actually confirmed

Google pays for meaningful contribution to an AI answer, not for appearing next to one, and it says so plainly in how the pilot is scoped.

Publishers accepted into the pilot see an "AI contribution" panel inside Google Search Console showing a monthly earnings figure, with some payment history, according to Digiday's reporting. Google has approached at least dozens of publishers, weighted toward small and mid-sized sites rather than large ones, and the scope runs well beyond news publishers, per 9to5Google's coverage on 17 September 2026.

Google has not published how it decides what counts as a significant contribution or how a payout is calculated. Digiday's reporting describes some participating publishers as finding the early payments small next to what the same content earns in advertising revenue.

The line Google drew: generation, not citation

Content earns a payout only if it materially shapes the answer while Gemini is drafting it, not if it is merely referenced once the answer already exists.

9to5Google's account of the mechanics, also covered the same week by Search Engine Journal on 15 September 2026, is specific: content that shapes how a response is generated can accrue earnings, while content that only confirms a fact the model has already stated does not qualify, and neither does a link added next to the answer after it has been generated. In plain terms, the moment that earns money is upstream of the moment that earns a citation.

Why this breaks the referral-traffic measurement model

A page can generate real value inside an AI answer and show zero referral traffic, zero citations and zero rows in a GA4 AI-referral report, because the payout and the click were never the same event.

Every GEO measurement approach in wide use, ours included, tracks visibility through the artefacts a citation leaves behind: a link a crawler followed, a click a visitor made, a session GA4's AI channel grouping picked up. The AI Search Attribution Handbook we published on 25 September 2026 is built on exactly that chain. It is a defensible model for the traffic economy. It has no way to see a page that earned money by shaping an answer nobody ever clicked through from.

That is not a flaw specific to our handbook. It is a gap in the whole measurement category, and naming it is worth more to a reader than pretending the model we shipped last week already covers it.

What to check on your own site this week

Run your existing GEO audit against the generation stage specifically, not just the citation stage, because the pilot rewards a different part of the pipeline than the one most audits check.

• Confirm your highest-value pages are readable by AI search crawlers at all. Our step by step GEO audit checklist starts here for a reason: a page that returns the wrong content type or sits behind a blocked crawler never reaches the generation stage, paid or not. Also check which crawlers you are actually letting through, covered in our AI crawler access playbook.
• Look for the pages an AI engine already paraphrases without linking. Search a distinctive sentence or figure from your own site inside ChatGPT, Gemini or Perplexity and see whether the substance comes back with no citation attached. That is the shape of content this pilot is built to reward, and it is a different signal from the one we describe in why llms.txt does not earn citations.
• Do not restructure a content programme around chasing this pilot specifically. It is invitation-only, the payout formula is undisclosed, and Digiday's reporting suggests some participants see it as a small addition next to referral advertising, not a replacement for it.
• Add a second measurement column alongside referral traffic: not "how many clicks did this page send us" but "how often does this page's own wording show up inside an AI answer, with or without a link." The first question is what most GA4 dashboards, including the AI referral view we describe in the attribution handbook, already answer. The second question has no dashboard yet, at Google or anywhere else, and it is the one this pilot is quietly paying for.

Our take

Google deserves credit for testing a real payment mechanism instead of leaving publishers with nothing but the hope that AI Overviews sends traffic eventually, a shift we first flagged when AI Overviews started building interfaces rather than just answers. Our view is that the pilot is honest about a trade-off the rest of the industry has been slow to say out loud: being useful to an AI answer and being visible in your own analytics are now two different achievements, and a site can win one without the other. Build for both. Do not assume your attribution dashboard, ours included, is already counting the one that pays.