Crypto could fix AI’s broken business model by rebuilding, on-chain, the traffic-for-content deal that generative AI is currently dismantling — that’s the case a16z General Partner Chris Dixon laid out alongside growth investor David George at the firm’s LP Summit, in a conversation published May 31, 2026. Their argument: AI chatbots now answer questions directly instead of sending readers to the websites that created the underlying information, and crypto’s ability to move value and enforce rules automatically is one of the few plausible ways to pay creators once that traffic stops flowing.

Key takeaways

  • Chris Dixon’s “internet pact” is breaking. For 20 years, search and social sent traffic to content sites in exchange for an implicit deal: show a snippet, deliver a click. Generative AI increasingly gives the full answer and keeps the user, breaking that exchange.
  • Google faces a textbook innovator’s dilemma. David George says a16z’s internal data shows AI-native query volume rising as Google search volume falls in the same categories — starting with low-monetizing knowledge lookups, not yet its highest-value commercial searches.
  • Crypto’s core value-add isn’t intelligence, it’s coordination. Dixon frames AI as a technology for building intelligent systems and crypto as a technology for solving coordination problems — how money moves, how rules get enforced, how creators get paid at scale.
  • Network effects, not brand, drive winner-take-all outcomes. Dixon says only two forces reliably produce durable competitive moats: network effects and enterprise sales — and George says most consumer AI apps have neither yet, despite reaching roughly a billion monthly active users faster than Google, Facebook or TikTok did.
  • Stablecoins are the proof of concept. Dixon points to stablecoin transfer volume now exceeding Visa’s monthly volume as evidence that crypto-native payment rails already work at institutional scale — the same rails he thinks an AI content economy would need.

Why AI is breaking the internet’s traffic-for-content deal

Dixon’s argument starts with what he calls, in his book, a “new covenant” — the unwritten rule that has governed the web for two decades. A search engine or social platform shows a snippet of a website’s content and, in exchange, sends the user a click. Wikipedia, news outlets, recipe sites and Stack Overflow all built their economics on that exchange.

Generative AI breaks it. Dixon describes tools like ChatGPT as effectively “one-boxing the whole internet” — giving a complete answer inline, the way Google already does with Wikipedia snippets, so there’s no reason left to click through. He cites Chegg, the homework-help company, as the starkest example: a publicly traded business that was “hit by a ton of bricks” once AI models could answer the same questions the site was built to monetize.

The risk, in Dixon’s telling, isn’t abstract. If the ad and subscription revenue that funded the original content dries up, there’s no obvious mechanism left to fund the next generation of it. “If you kill the business model that first created Stack Overflow, you’re not going to get future Stack Overflows,” he said — a problem for the AI models too, since they were trained on exactly that kind of human-generated content.

Google’s innovator’s dilemma, in real data

David George’s contribution is the data point behind the theory. He described an internal a16z chart plotting AI-native query volume against Google search volume in the same categories, moving in opposite directions — concentrated, for now, in knowledge retrieval, which he calls “the lowest monetizing forms of Google’s revenue model.” High-value commercial searches — insurance, travel, shopping — haven’t shifted yet.

That gives Google a classic innovator’s dilemma: it owns Gemini, one of the leading frontier model families, but its ~$100-billion-plus search business depends on delivering a link, not an answer. George argues the constraint isn’t technical — Google “actually [has] the best models right now” — it’s that the business model itself is “predicated on… getting the link and not the actual product.” Committing fully to AI-native answers would cannibalize the ad business that funds the company.

Why Dixon thinks crypto — not AI — is the fix

Dixon’s framing is that AI and crypto solve different problems. AI, in his words, is about building intelligent systems; crypto is about “collective action” and “coordination” — getting large numbers of independent parties to agree on rules and move value without a single intermediary controlling the exchange. Payments are the clearest example, but Dixon extends the logic to questions a content economy needs answered: how money flows to creators, how AI-copyright royalties get tracked, and how a fragmented web of publishers gets compensated automatically instead of through one-off licensing deals.

He points to stablecoins as evidence the coordination layer already works at scale: stablecoin transfer volume, he says, is now higher than Visa’s per month, up from payment rails that cost roughly $10 per international transfer a few years ago to under a penny today — cheap and fast enough, in theory, to support automated micropayments to millions of individual content creators, something legacy card networks were never built to do economically. Digital Asset Radar has covered that shift in more detail in our analysis of whether stablecoins are replacing banks and how the x402 payment protocol lets AI agents pay directly in stablecoins.

Dixon is explicit that this is a hypothesis, not a settled answer: “I don’t have a full answer to it, but I feel like this is the kind of thing we should be having discussions about.” His clearer prediction is about form, drawing an analogy to photography: cameras first replicated painting, then film became a genuinely new medium once photographic technology matured. He expects generative AI to follow the same arc — first replicating existing media, then producing “native” formats that couldn’t have existed before, each needing its own, likely crypto-native, way of paying the people who make it.

Network effects, moats and who actually wins

The conversation’s second thread is about market structure: what makes a technology business defensible once the current AI land grab settles down. Dixon says a16z’s investing thesis rests on two forces that reliably produce durable moats — network effects and enterprise sales — and that crypto networks like Ethereum and Solana already have the first once they reach scale. Reed’s Law, the idea that group-forming networks can create value that scales faster than simple user-to-user connections, is one framework the firm uses to reason about that kind of compounding; we go deeper on it in our explainer on Reed’s Law and the exponential age of crypto.

AI applications, George says, are a harder case. Consumer AI products have reached roughly a billion monthly active users faster than Google, Facebook or TikTok — and did it almost entirely through organic growth rather than paid acquisition — but “there is no network effect” yet locking users in. Dixon and George both describe technology markets as tending toward “winner take all,” using the Glengarry Glen Ross line — first prize a Cadillac, second prize steak knives, third prize you’re fired — to describe how thin the reward for being the number-two player usually is. Their conclusion for where AI value settles: chips and end-user applications capture the most value, while the model layer in between gets commoditized as cloud providers and app builders both push to drive its cost down.

Frequently asked questions

Can crypto fix AI’s broken business model?

Chris Dixon argues it could, because crypto is built for exactly the coordination problem AI content creates: moving money and enforcing payment rules among large numbers of independent parties without a central intermediary. He points to stablecoin volume already exceeding Visa’s monthly transfer volume as evidence the rails exist, but stresses this is an open question he’s trying to prompt discussion on, not a solved problem.

Why is Google’s search business threatened by AI?

David George says a16z’s internal data shows AI-native query volume rising as Google search volume falls in the same categories, starting with low-value knowledge lookups rather than Google’s highest-monetizing commercial searches. Google faces a genuine innovator’s dilemma: it has leading AI models in Gemini, but its roughly $100-billion-plus search business depends on delivering a link and ad impressions, not a direct answer.

What does Chris Dixon mean by crypto being a “second-order effect” of social media?

Dixon argues social media’s first-order effect was giving anyone a voice online; its second-order effect, decades later, was crypto — because a project like Bitcoin needed platforms where communities could organize and evangelize it, something that didn’t exist before social media. He uses the same framework to ask what AI’s own second-order effects will be, beyond the obvious first-order effect of generating content.

Do network effects apply to AI companies the way they do to crypto networks?

Not yet, according to David George. Consumer AI applications have reached roughly a billion monthly active users faster than Google, Facebook or TikTok, largely through organic growth, but he says there is currently no clear network effect locking users to one AI product the way there is with established crypto networks like Ethereum or Solana once they scale.

Sources

The original source video, plus the independent sources this article’s key claims were checked against: