For fifteen years, the deal in digital marketing was simple and generous: there was a huge, cheap supply of high-intent clicks sitting at the bottom of the funnel, and if you were disciplined about capturing them, you could build a business on it. Someone searched, an ad appeared, they clicked, they bought. Performance marketing was, for a long time, the closest thing to free money in the discipline.
That era is ending, not because anyone mismanaged it, but because the inventory it ran on is being absorbed. AI search is quietly eating the clickable placements performance was built on, and the strategic consequence is bigger than most media plans have priced in.
The capture layer is physically shrinking
Start with what’s actually happening on the results page. By early 2026, clickstream analyses put zero-click Google searches near 68%, up from about 60% two years earlier, the fastest acceleration of that trend in a decade. AI Overviews now appear on a large and growing share of searches and cut click-through by roughly 60% where they show up. Google’s AI Mode passed a billion monthly users. Combined outbound clicks, organic and paid together, are down more than a fifth in two years.
Read that as a paid media problem, because it is one. Every one of those absorbed clicks was a placement performance marketing could have monetized. The cheap, abundant, high-intent inventory that made demand capture feel easy is contracting, and it’s contracting fastest exactly where it was densest: informational and commercial-research queries, the on-ramp to the buying journey.
The demand didn’t disappear. The click did. And the click was the thing paid media was harvesting.
The reflex is to chase new capture surfaces
The industry’s instinct is already visible, and it’s predictable: find new places to run the same play. ChatGPT has shopping ads now, with entry costs that dropped from six figures to five. Retail media is booming, Amazon alone is set to take tens of billions in US ad spend in 2026. The pitch is “the capture surface moved, so move your capture spend with it.”
That’s not wrong, exactly. Those surfaces are real and worth testing. But it’s re-fighting the last war. Every one of those channels will commoditize the way search did, bid density rises, costs climb, the edge erodes, and when it does, the advertiser with the stronger brand will capture more cheaply on them, just like they always did on Google. Chasing the new capture surface is a treadmill. It buys you a quarter, not a moat.
When the cheap capture inventory shrinks, bidding harder for what’s left is not a strategy, it’s a bidding war you enter from a weaker position every year. The way out is to stop competing only for interception.
The durable move is upstream
Here’s the strategic conclusion, and it’s the one I’m now planning budgets around: if the capture layer is shrinking and commoditizing, the way to keep the funnel full is to create more of the demand that flows into it. That means investing more in brand and awareness, the upper-funnel, demand-generation work that performance marketers spent a decade treating as the soft, unaccountable part of the plan.
The logic is direct. You can only capture as much demand as something upstream created, and the capture surfaces are getting more expensive and more crowded. So the leverage moves to being the brand a buyer already wants before they ever open a search box or an AI assistant, because when they do, they’ll ask for you by name, and branded demand is the one kind of capture that stays cheap. This is the demand-generation vs demand-capture balance being forced by the market: for years the plan drifted toward capture because it was cheap and measurable. AI search is removing the “cheap” half of that, which changes the math.
It’s the same shift SEO is living through
None of this is separate from what’s happening on the organic side. GEO and AEO, being the brand an AI engine cites and recommends, are the organic-side response to the exact same collapse of the neutral, clickable middle. Brand-led paid is the paid-side response. Both are answers to one fact: when the engine answers instead of routing, being genuinely known and preferred beats being merely present and cheap.
That’s why I keep arguing paid and SEO are one demand engine. They’re converging on the same strategy from two directions, create demand, earn preference, become the brand the AI names and the buyer wants, because the era of cheaply intercepting demand you didn’t build is closing on both of them at once.
The catch: you have to be able to measure it
There’s a reason this shift is hard, and it isn’t strategic disagreement. It’s measurement. Upper-funnel brand investment cannot defend itself on a last-click report, it never could, which is precisely why it kept losing budget to capture. If you move money upstream and keep grading it on the old scoreboard, the first efficiency review will claw it right back.
So the shift to brand only survives if the measurement moves with it: geo holdouts, matched-market tests, and media mix modeling that can credit demand generation for the demand it actually creates. Make that argument with incrementality, not attribution, or the strategy dies the way every under-measured brand budget has always died.
What this means for the plan
Take the pressure on the capture layer as a signal, not a threat. The cheap-acquisition era rewarded whoever was most disciplined at harvesting. The next one rewards whoever is best at creating demand and earning preference, because interception is getting expensive and being wanted is getting decisive. Test the new capture surfaces, sure. But put the real weight upstream, into brand and demand generation, and stand up the measurement that lets you defend it. The advertisers who move first, while upper-funnel is still underpriced, will own the demand everyone else is about to start bidding for.