Google Priced Your Content for AI Answers at 0.1% of Ad Revenue
Google is paying roughly 100 publishers through an "AI contribution pilot" for content used in AI Overviews, AI Mode and Gemini, according to The Information. For small and midsize sites, those payments work out to less than 0.1% of their ad revenue. Meanwhile, AI Overviews cut the top organic result's click-through rate by 58%.
That 58% figure comes from Ahrefs' updated study of 300,000 keywords, and I'll come back to it, because the gap between those two numbers is basically the whole story. Google has now put a price on the content it uses to answer questions without sending the click. The price rounds to zero for most people.
What Google is actually paying (and to whom)
The details, per Search Engine Roundtable's write-up of The Information's reporting: the pilot started less than a year ago, covers about 100 small, midsize and large publishers, and pays based on how much a source "contributes" to an AI answer. Participants see accrued earnings in a widget inside Search Console.
The spread is wide. One publisher is earning more than $1 million a year. Another, which joined a few months ago, has made roughly $50,000 to $60,000 so far. Several smaller blogs and sites have earned less than $1,000 over several months, which is where the "one-tenth of 1% of ad revenue" line comes from.
And nobody seems to know how the math works. The Decoder notes that participants told The Information they don't know how Google calculates payouts, and that monthly amounts move around without explanation. Niche topics with dedicated audiences (gaming came up specifically) seem to earn more, which makes some sense if the metric is "how often you were the only good source."
Google's old position, as PYMNTS recapped it, was that it shouldn't have to pay publishers because search sends them traffic. This pilot reads like a quiet admission that the traffic argument got weaker. It just isn't an admission with much money attached.
Run the math on your own site before you get excited
Here's a rough micro-example. Say your content site makes $10,000 a month in ad revenue. At 0.1%, the AI contribution check is about $10 a month. Call it a nice lunch.
Now take one query cluster where you rank #1 and AI Overviews show up. Ahrefs found position-one CTR on those keywords was 1.6% in December 2025, against 3.9% for comparable keywords without an Overview. On 100,000 monthly impressions, that's roughly 1,600 clicks instead of 3,900. If your pages earn around $20 per thousand sessions (session RPM, not page RPM, and yours may be very different), those 2,300 missing clicks were worth about $46 a month. From one cluster.
So the trade, for a typical small publisher, looks something like losing $46 on a single cluster and getting $10 back across the whole site. The numbers are illustrative and your mileage will vary a lot. But I'd be surprised if many small sites in the pilot come out ahead.
This isn't just one vendor's dataset, either. Pew Research tracked 900 U.S. adults and found people who saw an AI summary clicked a traditional result in 8% of visits, versus 15% when no summary appeared. Clicks on links inside the summary itself happened in 1% of visits.
That 1% is the part that sticks with me.
Being cited is not the same as being visited.
The anchor problem nobody is negotiating around
My bigger worry isn't the $10. It's the anchor. Once "about 0.1% of ad revenue" exists as a real, reported number, every future licensing conversation starts from there, whether that's Google's next program or some other AI company deciding what content is worth. Opening offers tend to stick. It's a bit like a landlord publishing the rent before anyone has seen the apartment.
Some larger publishers seem to understand this. The Decoder reports that several are holding out of the pilot to keep pressure on Google, and that sits alongside the legal fights: Penske Media's lawsuit, the EU complaint from independent publishers, and the European Commission's antitrust probe. If you run a brand blog or a mid-size content site, you are not in that negotiation. Nobody is going to call you.
Which, honestly, is kind of freeing. You can stop waiting for a compensation model to fix the traffic problem and just plan around it.
My prediction: by the end of 2027, the median payout for small and midsize sites in whatever this program becomes stays below 0.5% of their ad revenue. Maybe the big names get meaningful deals. I'd bet against the long tail seeing real money.
Measure the leak per query cluster, not per site
Site-level traffic charts hide this. You need the cluster view. Here's a version you can do in about an hour:
- Export 16 months of Search Console query data (queries, impressions, clicks, average position). The API or a Looker Studio connector is easier than the UI export, which caps rows.
- Group queries into 10 to 20 topic clusters. Rough regex is fine. Don't overthink it.
- Flag which clusters trigger AI Overviews. Spot-check manually or use a rank tracker that records SERP features.
- Compare CTR year over year where average position held steady (within about one spot). If CTR fell more than 40% while position didn't move, that cluster is probably being answered on the SERP.
- Put a dollar value on the lost clicks using your own session RPM or conversion value, same math as above.
We went through a related version of this when Shopping ad CTR rose because AI Overviews absorbed the low-intent queries. The pattern keeps repeating: the easy, fully answerable queries leave first.
Build pages the Overview can't finish for the reader
Once you know which clusters are leaking, shift effort toward pages with a click reason. A summary can explain what a CPM benchmark is. It can't hand you the spreadsheet, the calculator, the downloadable template, or the dataset you ran yourself.
From what I've seen, the pages that still earn clicks after being cited tend to fall into a few buckets:
- Tools and calculators (the answer requires the user's own inputs)
- Original data people want to verify or reuse
- Templates, checklists, swipe files
- Opinionated comparisons where the "it depends" is the actual value
Pure definitions and "what is X" explainers are the opposite. Those are exactly what AI Overviews were built to eat, and at this point I'd stop commissioning new ones unless they feed a bigger page. If you want the engine-by-engine version of this, we broke down how getting cited by ChatGPT, Gemini and Perplexity differ, and it's a different job from ranking.
One messy thought I keep circling back to: the pilot rewards being the source Google leans on most, which in theory means better content gets paid more. In practice, you can't see the formula, the payments swing month to month, and the dollars are tiny. It's hard to optimize for a number you can't see. USA TODAY reorganized around owned audiences for more or less this reason, and they have a lot more pull with Google than most of us.
Treat the check as a receipt
If you get into the pilot, sure, take the money. It's better than nothing, and the Search Console widget might tell you something useful about which pages Google relies on. I just wouldn't let it change the plan much.
The number that actually runs your business is still clicks and what they're worth. Google just told you, pretty clearly, what it thinks your answers are worth without them. I'd take that seriously, and maybe a little personally too.
