Pick a side of Yahoo's own podium and you will find a different answer to the same question. Stand with the CEO and AI micropayments are a dead end that will never pay a publisher's bills. Stand with the licensing standard Yahoo helped found and AI micropayments are the whole plan. Nobody at Yahoo seems to have told the other room.
That is not a gotcha. It is the clearest live example of a split running through every publisher currently deciding whether to bet real resources on AI content licensing deals: the people negotiating the deals believe in them, and the people running the business the deals are supposed to fund are hedging in public, on stage, in their own words.
What Yahoo's CEO Actually Said On Stage
Jim Lanzone argued at HumanX in Amsterdam on September 23 that AI micropayments cannot fund journalism at any scale, and that advertising, not per-use payments, should carry the open web's costs instead.
His framing was blunt. The large language models, he said, "effectively photocopied the entire internet to build their answers, without linking back." But his prescription for fixing that was not a bigger licensing check. It was Scout, Yahoo's own AI answer engine, which he says will keep surfacing links to publishers as a block of blue results next to the generated answer, the same way classic search did. His stated reasons: users need to verify an AI engine that can be wrong, and sending traffic back is how the web has always worked.
“"It will never be enough, even for someone like Reddit," Lanzone said of AI licensing payments, arguing that referral traffic and search advertising are the only revenue base that scales.”
Reddit is not a random example. Reddit has one of the largest publicly known AI data licensing deals in the market, reportedly worth real money from both Google and OpenAI, and its CEO has separately said Google referral traffic is now falling faster than the licensing checks can offset. Lanzone picked the single best-funded publisher in the licensing economy and said even that is not enough. That is a specific, pointed claim about the ceiling on this entire business model, from the CEO of a company that is simultaneously trying to sell it.
What Yahoo Already Signed Up To Sell
Yahoo is a named, founding backer of the RSL Collective, a nonprofit built around the Really Simple Licensing standard, whose entire purpose is getting AI companies to pay publishers for content under exactly the per-use model Lanzone just waved off.
RSL adds licensing terms directly to robots.txt, the file that already tells crawlers what they can access, so a publisher can attach a price instead of a blanket block. The standard offers five license types: free, attribution, subscription, pay-per-crawl, and pay-per-inference. Pay-per-inference is the one Lanzone's "never enough" comment lands on hardest: a publisher gets paid only when an AI company's model actually uses that publisher's content to generate a specific answer, not simply for crawling the page.
The group is run by Doug Leeds, former CEO of Ask.com, and Eckart Walther, a co-creator of RSS, who frame it explicitly as an ASCAP or BMI for the AI era: a collective licensing body that sets terms once and negotiates on behalf of every member, rather than each publisher cutting its own one-off deal. Reddit, People Inc., Internet Brands, Ziff Davis, wikiHow, O'Reilly Media, Medium, The Daily Beast, Raptive, Ranker, Evolve Media, and Yahoo are named backers. Since enforcement is the hard part, since RSL alone cannot block a noncompliant crawler, the Collective is partnered with the CDN Fastly to act as the actual gatekeeper. Leeds has described Fastly's role bluntly, as "the bouncer at the door to the club."
Leeds also points to the $1.5 billion Anthropic settlement, finally approved by a federal judge on July 20, 2026, as evidence that AI companies will in fact pay when the legal exposure is real. That deal covers roughly 482,000 pirated books at around $3,000 each, with 91 percent already claimed. It is a real number. It is also a settlement over training-era piracy, decided in court, not a voluntary per-use license anyone agreed to pay going forward. Citing it as proof of appetite for pay-per-inference is an inference of its own, and a more fragile one than the Collective's pitch lets on.
Pay-Per-Crawl, Pay-Per-Inference, And Ads: What Each Model Actually Promises
Three different revenue models are now live in the same market at once, each with its own trigger for payment, and a publisher's AI content licensing deal only pays out under the specific trigger its deal actually uses.
| MODEL | WHO'S SELLING IT | WHAT TRIGGERS PAYMENT | WHERE YAHOO STANDS ON IT |
|---|---|---|---|
| Pay-per-inference (RSL) | RSL Collective: Reddit, Yahoo, Ziff Davis, People Inc., and others | Paid only when a model cites or draws on the specific content in a generated answer | Yahoo is a named founding backer |
| Pay-per-crawl (Cloudflare) | Cloudflare, acting as a bot-level toll both | Paid per crawl attempt, regardless of whether the content is ever used | Not Yahoo's model; see our prior [read on whether pay-per-crawl actually pays](/blog/cloudflare-pay-per-crawl-does-it-pay) |
| Bespoke direct deals | OpenAI, Google, and others, roughly 91 confirmed agreements across 314 domains | Negotiated privately, terms almost never disclosed | Reddit and the largest media groups hold most of these; Yahoo is not confirmed among them |
| Ads against the AI answer | Yahoo, through Scout | No publisher payment at all; publisher gets a referral click, same as classic search | Lanzone's stated preference, publicly, over any of the above |
Put those four rows next to each other and the contradiction stops looking like a slip and starts looking like a company that has not decided what it actually believes, and is running two incompatible bets on the same balance sheet while it waits to find out which one wins.
What a licensing deal already buys the 91 publishers who have one, from the June 2026 Press Ranger / OtterlyAI citation study
That data, which we broke down in full when the study came out, is the strongest evidence either side of Yahoo's split has. The premium is real, 48% more citations per page on ChatGPT for the 91 publishers who already have a deal. But the club capturing it is tiny, concentrated, and dominated by five media groups. RSL's entire pitch is turning that narrow, bespoke premium into something any publisher can opt into through a standard rather than a negotiation. Lanzone's position is that the premium will never be large enough to matter even for the biggest winner in the club. Both readings can be correct about different publishers at the same time, which is exactly why nobody should take either one as settled.
Why The Contradiction Matters If You Don't Have A Seat At The Table
Almost nobody reading this runs a content business inside the 91-deal universe, which means the Yahoo split is not really a licensing story for most brands, it is a signal about how much weight to put on licensing revenue when it shows up in someone else's pitch.
Vendors selling AI visibility tools, agencies selling citation audits, and publishers pitching advertisers have all started leaning on licensing deals as proof that AI citations carry real economic value. That is directionally true. It is not evidence that a specific brand without Reddit's negotiating leverage or Yahoo's seat on the RSL board is about to get paid for anything. Google's own scheme for compensating sources is narrower still, and nobody outside a short list of publishers has seen a dollar from it yet.
The honest reading of Lanzone's comment is that he is talking his own book, too. Scout runs on ads and referral traffic, so of course its CEO prefers a world where that model wins the argument. RSL's backers are talking their own book in the other direction. Neither side is a neutral authority on what AI content licensing deals are actually worth, and a brand's content strategy should not be built on either one's forecast.
What To Do About AI Content Licensing Deals Right Now
The useful move is the one available regardless of who wins the Yahoo argument: build the kind of coverage and citation profile that earns attention on merit, because that is the only lever a brand without a licensing seat actually controls.
That means treating earned media, digital PR, and classic link building services as one coordinated effort aimed at the same target, rather than three separate budgets with three separate goals. The outlets and formats that already earn AI citations without a licensing deal behind them, the specific process for turning digital PR work into AI citations, are available to any brand today, no RSL membership and no Fastly handshake required.
Watch the RSL Collective's enforcement record over the next two quarters rather than its backer list. A standard twelve publishers signed is not the same as a standard AI companies actually pay into. If Fastly starts blocking noncompliant crawlers and AI companies route payment through it anyway, pay-per-inference becomes a real line item. If it stalls the way early robots.txt-based AI opt-outs mostly did, Lanzone's bet on ads and referral traffic will have been the correct read, and the standard will have been a negotiating position dressed up as infrastructure.
Either way, do not wait to find out before building the thing that works under both outcomes: content and coverage good enough to get cited without anyone having to pay for it.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.