Google is testing paying publishers. AI overview tracking still beats waiting
Google is quietly testing whether to pay a handful of publishers when their content shows up inside an AI answer, and the honest reaction is not excitement, it is a reason to get serious about ai overview tracking before anyone changes strategy.
What Google's AI Contribution pilot actually is, and is not#
A fox does not change its whole hunting ground because it heard one rustle in the hedgerow. Google is reportedly testing a Search Console panel called AI Contribution, showing a small group of mostly small and mid-sized publishers dollar figures tied to content that "meaningfully contributes" to AI Overviews, AI Mode and Gemini answers, per 9to5Google's report. Worth stating plainly: this is trade press reporting on a discovered feature, not a Google announcement. There is no Google Search Central page or help-centre document describing AI Contribution by name, and the payout formula has not been disclosed.
That gap between what is confirmed and what is merely spotted matters more than the headline. folkfox has said publicly, on its own SEO and GEO pages, that no special AI-specific markup or file is required to appear in Google's generative features, a position Google itself confirms in its own guidance: "make pages for your audience, not just for generative AI search," per Google's AI optimisation guide. A quiet pilot testing payment for citation does not change that; it is a monetisation experiment layered on top of the same visibility question, not a new set of rules to chase.
A second, confirmed shift landed the same week#
Separately, and with an actual effective date behind it, Cloudflare's AI-crawler policy switched to blocking Training and Agent crawler categories by default on ad-monetised pages for new domains from 15 September 2026, while Search crawlers stay allowed, per Cloudflare's own Content Independence Day announcement. Because a crawler like Googlebot serves more than one purpose, a site that blocks Training without a carve-out risks blocking Google's own AI-training access even while leaving Search crawling untouched, which is exactly the kind of thing decent ai overview tracking should catch before a client notices a drop.
Google is testing an "AI Contribution" pilot in Search Console, paying publishers when their content is used in AI Mode, AI Overviews and Gemini.
That reaction, translated, frames the pilot as the start of citation quality becoming its own revenue line, separate from organic traffic. It is a reasonable early read. It is still a read of an unconfirmed pilot, not a policy, and betting a strategy on it is a thicket nobody needs to walk into yet.
The ai overview optimization question nobody needed a pilot to answer#
While the industry debates a payment pilot that barely exists yet, the more useful, already-measured story is sitting in plain sight: most of the AI-specific groundwork sites have already done is not being read by anything. Ahrefs studied 137,210 domains in May 2026 and found 28% publish an llms.txt file, the proposed AI-crawler instruction standard, while 72% do not, and among the domains that did publish one, 97% received zero requests for it the entire month, per Ahrefs' own study: "Of the ~38,000 domains with a valid file, 97% saw no requests for it whatsoever in May. No bots. No humans. Nothing."
| Item | Value |
|---|---|
| Publish llms.txt | 28% |
| Do not publish one | 72% |
| Item | Value |
|---|---|
| 97% of published llms.txt files received | 97% of published llms.txt files received |
| zero requests in May 2026 | zero requests in May 2026 |
Google's own position matches the data rather than contradicting it: such files "neither harm nor help your site," and a separate server-log study, covered by ezy.ai's research, found only one major AI company, Meta, reading llms.txt files at any meaningful scale, with OpenAI, Anthropic and Perplexity all near zero.
A parallel Ahrefs study tracked 1,885 pages that added JSON-LD schema markup specifically to chase AI citations and found "no major uplift in citations on any platform," per Ahrefs' schema study. Schema stays technical hygiene. It is not a lever for ai overview optimization, however often it gets sold as one, and the fox that keeps digging the same empty burrow is not being thorough, it is being stubborn.
What real ai overview tracking should actually measure#
If markup is not the lever, what is? Google's own AI features guidance points at the same fundamentals it has pointed at for years: unique information, first-hand experience and named evidence, tracked over time rather than assumed from a single audit. Two genuinely new surfaces are worth watching specifically, both confirmed this month. AI Overviews now dynamically expand for some queries without a user clicking "Show more," pushing organic results further down the page, per Search Engine Land's coverage, and Google is rolling out link carousels for developing topics inside AI Mode, the first genuinely new link surface to appear in months, per Search Engine Land's separate report.
None of this requires waiting on Google's payment pilot to resolve. The best answer engine optimization tools, whichever vendor supplies them, are only as good as the discipline behind them: measuring cited queries as a named metric, monthly, against a baseline set before you touch a single page.
Five moves that beat waiting on the AI Contribution pilot#
These five moves work whether Google's payment pilot ever leaves the test group or quietly disappears, because none of them depend on it. A pilot with no public criteria is not a plan a client can budget against, and folkfox would rather hand over five things a team can start doing this afternoon than a forecast built on a Search Console screenshot nobody outside the test group has actually seen.
Set up ai overview tracking, manual or automated, that records which queries cite you inside an AI answer, separate from ranking position.
Publish one if you like for completeness, but do not expect it to change what gets crawled or cited; the measured data says it will not.
If your site sits on Cloudflare, check whether the new default blocking of Training crawlers is accidentally catching Google's own AI-training access alongside unwanted ones.
A generative system lifts a sentence, not a section, so write claims that survive being quoted alone, with the source named at the point of the claim.
Snapshot impressions, clicks and citation rate for your most commercially important queries now, so a future change, whether from you or from Google, is measurable.
| Check | Where to look | Good direction |
|---|---|---|
| Cited queries | Manual AI Overview and AI Mode spot checks | Rising |
| Impressions vs clicks | Search Console performance report | Impressions hold, clicks may fall |
| llms.txt requests | Server logs, if published at all | Not worth chasing per Ahrefs' data |
| Site reputation standing | Search Console manual actions report | No active action, especially in the EEA |
- Cited queriesManual AI Overview and AI Mode spot checksRising
- Impressions vs clicksSearch Console performance reportImpressions hold, clicks may fall
- llms.txt requestsServer logs, if published at allNot worth chasing per Ahrefs' data
- Site reputation standingSearch Console manual actions reportNo active action, especially in the EEA
How to get cited in AI search, stripped of the pilot-chasing noise, is really just this list done consistently, month after month, on the queries that pay a client's bills rather than the ones that make an easy screenshot.
Reporting ai overview optimization results the way a client will actually believe#
Report the gap honestly rather than the pilot breathlessly. A client asking about Google's AI Contribution deserves the accurate answer: it is a small, undisclosed test with no public criteria, not a monetisation strategy to plan a quarter around. What they can plan around is the measured, sourced work of tracking citation the same way this piece has laid out, applied consistently across the queries that actually drive enquiries rather than the ones that make a tidy screenshot for a slide.
Chasing the citation matters more than chasing the file format nobody reads.
The honest measure of progress here is not whether Google pays you. It is whether your own ai overview tracking shows the same cited queries climbing month over month, built on evidence rather than a file format that 97% of readers, human or otherwise, never open. How to check ai citations reliably is simply to keep asking the same question on the same schedule: did this page get quoted this month, and can I prove it.
If your team wants this measurement layer built properly, rather than assembled from whichever tool happened to be free, folkfox's SEO and GEO services track exactly this, alongside the same evidence-first discipline behind folkfox's AI consultancy and content marketing work.
Frequently asked questions#
What is Google's AI Contribution pilot?
It is a reported Search Console panel testing whether Google should pay a small group of publishers when their content meaningfully contributes to AI Overviews, AI Mode or Gemini answers. It has not been publicly confirmed by Google, has no disclosed payout formula, and covers only a limited test group as of September 2026.
What is ai overview tracking?
AI overview tracking means measuring whether your content is actually cited inside AI-generated answers, such as AI Overviews or AI Mode, rather than only measuring its organic ranking position. It tracks citation as its own metric, separate from clicks or ranking.
How to get cited in AI search?
Write self-contained sentences with specific, sourced facts, since a generative system lifts a sentence rather than a whole page. Google's own guidance points to unique information, first-hand experience and named evidence as the durable factors, not special markup or an llms.txt file.
How to check ai citations for my site?
Manually search your priority queries inside AI Overviews and AI Mode and note whether your page is cited, or use one of the available answer engine optimization tools that automate the same check. Cross-reference with Search Console's performance report: if impressions hold while clicks fall, you are likely being summarised rather than demoted.
Does llms.txt help a site get cited in AI answers?
The evidence says no. Ahrefs studied 137,210 domains and found that among the 28% publishing an llms.txt file, 97% received zero requests for it in a month. Google itself states such files neither help nor harm a site's visibility.
Does adding schema markup improve ai overview optimization?
Ahrefs tracked 1,885 pages that added JSON-LD schema specifically to chase AI citations and found no major uplift on any platform. Schema remains useful technical hygiene but is not a proven lever for ai overview optimization.
Read more on this topic#
Generative engine optimization services have a click problem
The click-versus-visibility distinction this piece leans on, explored through a different AI Overviews experiment.
Read the pieceSEO AND GEOAI search visibility experiments need an earned citation trail
Two recent tests that make the same evidence-first case this piece makes about citation tracking.
Read the pieceSEO AND GEOThe content marketing ROI report Google never wanted to build
A related Search Console reporting change, forced by regulation rather than piloted quietly.
Read the pieceReady to measure AI visibility properly, pilot or no pilot?
folkfox builds ai overview tracking that survives whatever Google does next, measuring citation as its own evidenced metric rather than hoping for a payout.
Want folkfox in your Google results and AI answers? Set folkfox as a preferred source.