A Miracle of Efficiency, Catastrophe of Incentives 

By Bryan Quinn, President and Co-Founder, Shopsense

Institutions don’t typically collapse from the outside. They do so from within, one small compromise at a time. Then suddenly one day, you look up and find yourself standing in a building with no walls.

That’s what’s been happening to the open web. And the instrument of its demise has a name: CPM.

Cost per thousand impressions. It’s the metric that built an industry, funded a revolution in global publishing, and, I’d argue, corrupted the very ecosystem it was designed to sustain.

The miracle and the curse

Let’s give the CPM its due. In the early days of the internet, it was a genuinely elegant solution to an otherwise paralyzing problem: how do you monetize an infinite, frictionless medium? You can’t charge for every page read. You can’t build a tollbooth in front of a global library. So, you sell attention. In bulk, at scale, with predictable pricing. Publishers could fund journalism, advertisers could reach audiences, and a whole economy of freely available information became possible.

CPM was simple and scalable. But, it so happens, a catastrophe of incentives.

Because here’s what the CPM actually taught publishers: more pages equals more money. More ads per page equals more money. More page refreshes, more ad freshes, more auto playing video, more ad tech scripts loading in the background while you’re trying to read a recipe. All of it equaled more money.

And so, following the entirely rational logic of the incentive structure ad tech created for them, that’s exactly what publishers built.

The race to the bottom had a GPS

Clickbait wasn’t a lapse of editorial judgment, but a business strategy perfectly calibrated to a CPM-dependent world. Outrage drives clicks. Clicks generate pageviews. Pageviews generate ad impressions. Publishers sell impressions. The math was impeccable, but the product degraded.

Users responded, as they always do. They installed ad blockers by the millions. They migrated to social platforms that felt, at least at first, less hostile to their attention. They retreated into walled gardens, where the experience was curated, fast, and provided ad experiences largely based on user engagement (CPC, CPA, CPI), thus ensuring some level of relevancy.

And those walled gardens, in turn, consolidated the data, the ad serving, and the measurement. They could offer advertisers something the open web struggled to: a controlled environment where outcomes, not just impressions, could be demonstrated.

Meanwhile, programmatic intermediaries multiplied like invasive bamboo, and somewhere in that tangle of pipes, a substantial portion of the advertiser’s dollar simply… evaporated before it reached the publisher who created the content.

So, publishers needed more impressions to compensate for lower CPMs. Which meant more pages, ads, and ultimately clutter. A flat spin.

AI pulls the e-brake

I want to be direct about what artificial intelligence does to an impression-dependent business model: It renders it structurally insolvent.

When someone asks an AI system for a product recommendation, no pageview occurs. When a summary surfaces an answer without requiring a click, the impression inventory disappears entirely. The CPM model assumed that users would perpetually move through the open web generating countable moments of attention. AI is decoupling information retrieval from that motion altogether.

Publishers who spent years optimizing for pageview volume now face another pressing question: What’s their strategy for getting their content into the AI ecosystem? Jonathan Roberts of People Inc. said it well: “the best AI needs the best inputs.”

Publishers who treat AI training and licensing as an afterthought will find themselves providing the raw material for systems that deliver little reciprocity. Those who negotiate deliberately, who choose which content to make available, on what terms, and in service of which relationships will have a genuine opportunity to shape how their editorial voice persists in the next era of information. The previous era rewarded volume. The AI era will reward intentionality.

The CPM didn’t cause AI, of course. But it leaves flat-footed publishers poorly positioned to survive it. A model built on volume has no resilience when that volume disappears.

What we owe the web we built

The choices we make right now about how we structure incentives and what we decide to reward will determine whether what comes next is better — or simply worse than what we’re leaving behind.

Performance-based models that compensate publishers for engaged time, real conversions, and demonstrated user value are not a niche experiment. They’re the architecture of a web that respects their audience’s attention. When publisher incentives align with user value rather than raw volume, the rational strategy changes and the end-to-end user experience improves.

First-party relationships in the form of newsletters, communities, direct connections, etc. should be celebrated. They’re a foundation. Publishers who know their audiences, who have a direct line to them unmediated by third-party gatekeepers, have genuine and, more importantly, durable connections.

Advertisers have a role here, too. Every dollar that flows toward measured outcomes rather than purchased impressions is a vote for a different kind of web. The industry has the leverage to demand better infrastructure, more transparent supply chains, and formats that are relevant to content the users are consuming.

The web is a choice we keep making

The internet’s original promise was a free, open, global exchange of information. It’s being eroded by an incentive structure that rewards the wrong things, compounded over decades, until the cumulative damage becomes the prevailing state we all move through without noticing.

The CPM was a perfectly rational response to an actual problem. It just happened to be flawed. Not wrong in intent but wrong in consequence, which is the most expensive kind of wrong there is.

We can build a web where creating genuine value for users is the most direct path to sustainable revenue. We know what that looks like. We have the tools, the models, and, frankly, the historical education in what happens when you don’t.

What we need now is the will to choose it.