Funnels to Hourglasses – Where Ads Go with AI

Thank you for all the feedback on my last article on Advertising in the Age of LLMs. I continue to believe that advertising on AI platforms will not need to conform to prior models, and that automated intelligence creates space to evolve advertising into something more akin to agentic commerce.

Yet a question that comes up often in these debates is the direction of travel for media — or said differently, where will the ads go? As someone who has worked adjacent to media for more than a decade, and believes in a vibrant media ecosystem for the good of society, this question hits close to home.

The Story So Far

To situate an answer, it’s worth briefly recapping the shifts underway in media over the last few decades. (This is U.S.-focused, but many of these trends are global.)

1. Collapse of geographic monopolies: Media was once dominated by organizations that held geographic monopolies over their audiences — local TV, newspapers, and radio. These monopolies gave advertisers efficient reach: buying ads from the top outlets secured access to most consumers in an area. The internet blew up those boundaries. While it took decades to play out, this undermined a core value proposition media owners offered advertisers, while simultaneously exposing just how duplicative much their content really was.
2. Democratization of publishing: The blogging revolution – and later, user-generated video, audio, and social media – made everyone a publisher. New, digital-first publishers emerged. But the abundance of content began shifting value to aggregators (search engines, social platforms, video platforms) that helped users navigate the flood.
3. Digital adaptation: Traditional media adapted by moving content online, building digital ad businesses, and even acquiring web-first brands. Many publisher brands still carry weight with advertisers. With help from ad-tech providers, media companies tried to close the gap with aggregators by adopting interactive ad formats, targeting, and access to programmatic ad buying.
4. Rise of the attention economy: Media fragmentation and content abundance turned attention into a primary currency, concentrated on social and video (including podcasts). Consumer behavior followed, with exposure-focused advertising budgets shifting toward these channels, further fueling the aggregators’ advertising engines.

We are likely in the late stage of this evolution. Anecdotally, the share of digital ad spend going to web-focused publishers has declined. Publishers became reliant on a set of fragile techno-economic underpinnings — third-party cookies, Apple’s ATT policies, and even Meta’s and Google’s patronage of journalism. Once those props were kicked out, the revenues underpinning much of web media fell. Add in post-pandemic pressures on efficiency, and it’s no surprise that advertisers and agencies consolidated spend with a smaller set of partners – namely, the aggregators.

I raise these points not to be a downer on traditional/web media, but to paint a clearer backdrop against which LLMs arrived.

Where Do the Ads Go?

Ads follow attention (the spice must flow). The rise of LLMs is sharpening a split I foresee for the internet into three, distinct, AI-mediated, spheres: Information, Entertainment, and Services:

1. Information (traditional and web media) is most at risk of being displaced of advertising. These publishers were already vulnerable given reliance on search engine traffic. “Arts and Entertainment” queries alone account for ~17% of Google’s volume. That context makes Penske Media’s recent Google lawsuit more understandable. AI also is exposing how much duplication exists here; I find it interesting that AI platforms seem to be pursuing content licensing agreements from only a handful of sources (NYT, Reddit, etc.).
2. Entertainment (particularly streaming video) is safer, at least for now. Cord-cutting and the migration of live sports online give advertisers continuity in proven brand-advertising formats. But disruption is possible as generative video progresses (consider Runway Aleph, DeepMind Veo3, Bytedance Seedance). And YouTube continues to dominate streaming distribution.
3. Services weren’t traditionally seen as media, until Amazon Ads crystallized retail media into a scalable model. Others – Target, Best Buy, DoorDash, Instacart, Costco, Expedia, Intuit, and more – have followed with varying success. Some lack strong “owned and operated” surfaces, which may push them toward more agentic, AI-driven approaches. The space between e-commerce and AI agents is likely the next battleground for attention (and advertising).

From Funnel to Hourglass

Advertising has long been about persuasion, especially in brand building. Those dollars follow attention, which increasingly sits in fragmented, discovery-oriented channels. Marketers know this and are finding that consumers arrive better informed at the point of decision. (Over 50% of U.S. consumers now start shopping searches directly on Amazon.)

OpenAI’s recent media kit paper “How People Use ChatGPT” suggests that 21% of conversations are about “seeking information,” and ~2% involve specific products. My hunch is that AI today contributes more to product discovery (attributes, comparisons) than product exposure (awareness a product even exists), but that may shift as AI platforms push toward proactive experiences.

What this opens up is space for advertising beyond the click.

In a recent discussion, Deloitte’s Kasey Lobaugh argued that AI economics will push businesses towards becoming hyper-relevant service organizations. Will the marketers who thrive in the AI era be those who move beyond persuasion and into orchestration? SKU-level personalization should be table stakes. Imagine these journeys tuned end-to-end: a custom unboxing, a generated celebrity endorsement, or branded intelligence delivered via expert agents (e.g., a GM agent walking you through how the Ultium EV platform differs from the Rivian Skateboard platform). In this framing, the funnel flattens in the middle but widens again at the bottom – an hourglass model, if you will, where discovery is fragmented, persuasion is compressed, and value accrues in rich, post-transaction experiences.

Conclusion

Advertising has always followed attention, but never in a straight line. The collapse of geographic monopolies gave rise to aggregators; the rise of AI may push us into a world where discovery, persuasion, and commerce no longer fit neatly in a funnel. Instead, they compress, expand, and shift across zones of information, entertainment, and services. For marketers, the challenge isn’t just finding the audience – it’s designing for the new end states of attention: agentic, contextual, and transactional. That, I suspect, is where the ads will go.