AI is dismantling the open web's ad-funded business model — what survives and what doesn't

Summary

AI is collapsing the economic foundation of the open web by severing the connection between publishers and advertiser-funded attention. As users shift from search engines to AI answer engines like ChatGPT and Perplexity, they bypass the ad-supported websites that once monetized traffic. And as AI agents handle increasingly complex tasks—research, appointments, purchasing—users will spend less time on content altogether, draining the ad pool that built the modern internet.

This is the core argument from Michael Mignano at Lightspeed Venture Partners. What matters is not just that Google search traffic will decline, but that the entire transactional model breaks. When someone asks ChatGPT "best places to travel in Spain," no publisher gets the click. No ad gets served. The relationship between advertiser and attention is severed.

What fails under this pressure

Content businesses that rely on organic search rank and commodity information face the most immediate threat. Travel sites, recipe blogs, niche publishers—anything built on the assumption that Google would funnel attention to their pages. That game has been running for a decade (Google itself has been pulling hotel bookings, shopping results, and flight purchases directly into search results), but AI accelerates the endgame. These publishers can't afford to exist if traffic goes to zero.

The calculus differs by business type. Entertainment platforms like Meta and YouTube remain relatively insulated as long as they hold attention. The ad model works because users stay in the app; what matters is eyeballs, not referral traffic. Google's search dominance is less threatened than content publishers think, because Google still owns the moment when users want to actually buy something—the "toll booth" of commerce. Perplexity and other AI agents are now gunning for that same position.

What survives and consolidates

Premium, paywalled content becomes scarcer and more valuable. As the supply of free, ad-supported content shrinks, demand for high-quality journalism and analysis increases. Stratechery, Semi-Analysis, and niche reporting that operates behind paywalls face less pressure because they're already decoupled from the ad model. The best content will consolidate more attention and more ad dollars. Some will also double down on direct monetization—Reddit's deal to license its content to Google for training data offers a template: sell your data to AI companies at premium rates while continuing to serve ads to users who still visit.

The reinvention layer

AI platforms will attempt to rebuild advertising within their own interfaces. Perplexity is already experimenting with ads. OpenAI is rumored to be exploring them. Google has successfully embedded ads into its language model interactions. The question is whether those dollars will flow from advertisers and platforms back to publishers, or whether the new gatekeepers—OpenAI, Perplexity, Google—will capture the full value. A company pitching an ad network for LLMs makes intuitive sense, but the distribution likely runs through the model owners themselves, not independent networks.

Publishers too small to cut their own licensing deals (like OpenAI's agreements with News Corp and Axel Springer) may rely on third-party services that block AI scrapers and charge per-access fees. As AI agents move beyond content ingestion into task execution, APIs become chargeable events. The model shifts from "you own the traffic" to "you charge per interaction."

The harder question

What emerges depends on whether publishers can participate in the new model at all. If OpenAI and Perplexity keep the bulk of ad revenue and don't route enough back to creators, publishers will have no incentive to publish, and the models will have no content to index. Alternatively, if agentic products argue they're just accessing information on behalf of users, paywalls become a game of friction. A user hits a Stratechery paywall; an AI agent hits a paywall; the agent goes elsewhere. Paywalls only work if the content has enough perceived value that users (or the agents acting on their behalf) convert.

The outcome remains unresolved. One possibility floated: creators publish freely to the open web and receive direct payouts when AI platforms generate revenue by serving their facts or takes to millions of users—a YouTube-like model for LLM-generated queries. Another: the open web fragments into paywalled premium zones and commodity-content wastelands, with AI platforms as the new intermediaries.