The numbers from Q2 2026 aren’t subtle. According to Ozone benchmarking data reported by Digiday, tracking roughly 20 billion impressions, U.S. publisher ad request volume dropped 32–37% year-over-year in Q2, and U.S. programmatic spend fell 44% YoY across the first half of 2026. Out-stream video, the inventory type most exposed to traffic-dependent supply, was down roughly 76% YoY in June. If you’re a performance marketer planning Q3 budgets around last year’s display availability, you’re planning around a market that no longer exists.
This Isn’t a Dip — It’s the Structural Consequence of AI Search Absorbing Open-Web Traffic
The instinct is to treat a supply contraction as cyclical. Budgets shift, publishers adjust, inventory recovers. But the mechanism driving this drop isn’t a budget cycle. It’s a structural change in how users interact with information.
Danny Spears, COO of Ozone, told Digiday that platforms — Google in particular — are now stepping into the user journey and serving content in place rather than passing users through to the underlying website the way classic search did. When AI search answers a query directly — no click, no pageview, no publisher visit — the ad impression that would have been generated on that publisher’s page doesn’t exist. It’s not deferred. It’s eliminated.
Matt Barash, CCO at Nova Studio, framed the historical shift for Digiday this way: search built the pageview economy, while social keeps users engaged on-platform. The less time consumers spend navigating the open web, the scarcer — and more valuable — premium display inventory becomes.
This is the supply problem that’s actually a demand problem in disguise. Publisher ad inventory is a byproduct of human visits to publisher websites. When AI intermediaries answer the question before the user ever reaches the site, the raw material for ad impressions evaporates upstream. The open web’s challenge isn’t figuring out how to monetize traffic. It’s figuring out how to get the traffic back — and that’s a much harder problem when the platforms controlling distribution have every incentive to keep users on their own surfaces.
The Spend-to-CPM Divergence Confirms Scarcity Pricing Is Coming
The divergence between spend and CPMs tells the structural story. In the U.S., programmatic spend fell 44% YoY across the first half of 2026 while June eCPMs rose only about 7% YoY. That’s a market where demand is pulling back roughly in proportion with supply — neither side fighting hard enough to create dramatic price distortion. But that balance is temporary. In markets where demand held firmer while supply contracted, eCPMs spiked roughly 30% — the textbook inflationary pattern of a structurally shrinking inventory pool, and a preview of where U.S. pricing heads the moment demand stops retreating as fast as supply.
Web inventory spend specifically was still down 38.5% YoY in June. The only channel showing growth was in-app, up 23% — which makes sense, because app inventory isn’t dependent on search referral traffic. It operates within a walled ecosystem with its own user engagement loop.
For performance marketers, the read-through is straightforward: channels most dependent on open-web traffic patterns are the channels most exposed to AI search erosion. And that erosion is accelerating, not stabilizing.
Why Your Reach-and-Frequency Models Are Already Overstating Available Audience
A 40% supply contraction doesn’t just raise CPMs. It compresses your addressable audience. When the impression pool shrinks by that magnitude, remaining inventory concentrates into fewer publishers, fewer content verticals, and — critically — fewer unique users. Your frequency against reached individuals rises even if your caps haven’t changed, because the same people are cycling through a smaller set of sites.
That means your Q3 media plan, if built on Q4 2025 or even Q1 2026 supply assumptions, is running on inputs that overstate available reach by 30–40%. You’re not just paying more per impression. You’re reaching fewer net-new prospects per dollar, which means your effective CAC is climbing on two axes simultaneously: price and efficiency.
No amount of bid optimization or exchange-level negotiation fixes a shrinking denominator. You can’t optimize your way to incremental reach that doesn’t exist in the supply pool.
Programmatic Direct Mail Doesn’t Depend on Anyone Visiting a Website
This is where the structural case for programmatic direct mail becomes less about channel diversification and more about CAC stability.
Direct mail’s supply isn’t a derivative of platform traffic. Every mailpiece reaches a verified physical address. The impression isn’t contingent on a user clicking through a search result, visiting a publisher, or loading an ad unit. The impression is the mailpiece — deterministic, physical, and completely decoupled from the AI intermediary layer hollowing out open-web inventory.
When your digital display pool contracts 40% in a single quarter, your acquisition model needs at least one channel where the addressable universe is fixed and predictable:
- The audience is address-level, not probabilistic. You’re targeting verified individuals at verified physical addresses using first-party CRM data, Postie’s ML-built lookalike audiences, or third-party enrichment from Epsilon, Acxiom, and Experian. No impression depends on that person happening to visit a website where your ad loads.
- The supply doesn’t depreciate with algorithm updates. USPS delivers to approximately 163 million addresses. That number doesn’t shrink when Google changes how it resolves queries or when ChatGPT keeps users on its own surface.
- Matchback attribution closes the loop without view-through inference. Postie’s matchback attribution ties mail sends to conversions at the household level — no cookies, no IP mapping, no probabilistic identity layer. You measure direct mail ROAS with the same rigor you apply to your highest-performing digital channels.
- Trigger campaigns respond to real-time behavioral signals. Postie’s trigger-based direct mail activates on CRM events — cart abandonment, lapsed purchase windows, lifecycle milestones — so you’re not just replacing display prospecting. You’re running performance direct mail against the same intent signals, through a channel with stable supply.
This isn’t about replacing digital display. It’s about ensuring your acquisition mix includes a channel whose supply fundamentals aren’t eroding quarter over quarter, and whose measurement doesn’t degrade as traffic disappears upstream.
The Q3 Decision Isn’t Whether to Diversify — It’s How Fast You Shift Spend
The open web’s inventory recovery depends on solving a problem no one has a clear path to solving: convincing AI platforms to send users back to publisher websites when keeping them on-platform is the entire business model. That’s not a bet performance marketers should be making with Q3 acquisition targets on the line.
The marketers who maintain CPA and ROAS targets through this contraction will be the ones who shifted spend toward channels with stable, addressable supply before scarcity pricing fully kicked in. Programmatic direct mail delivers addressable impressions to verified audiences regardless of what happens to the open web — not because it replaces what display does, but because it provides first-party data activation and deterministic attribution in a channel where supply is structurally fixed.
If you’re stress-testing your Q3 media plan against a structurally shrinking display market, see what Postie’s programmatic direct mail performance looks like against your specific audience and acquisition targets.
