In December 2025, a publishing executive told Digiday that agentic browsers were a "three percent problem" — too small to rank among their top threats. Five months later, the number keeps surfacing.
Traffic from AI agents grew 7,851% year over year. Media sites absorb 45.6% of all agentic traffic. Perplexity's Comet and OpenAI's Atlas account for 71% of observed agent activity. Automated traffic is now growing eight times faster than human traffic. And Salon already lost a buyer because an advertiser couldn't distinguish agent visits from human ones.
Three percent was a comfort. The comfort is evaporating.
Three Protocols, Three Futures
As agentic browsers replace the visit itself — reading, summarizing, and acting on content without the user ever touching the publisher's site — three protocols have emerged to rebuild the economics. Each proposes a different relationship between publishers and the machines that consume their work.
| Protocol | Model | Publisher Role | Scale |
|---|---|---|---|
| Microsoft PCM | Pay-per-use licensing | Supplier | Terms undisclosed |
| Perplexity Revenue Share | 80/20 subscription split | Cited source | $42.5M total pool |
| Omnicom AdCP | Agent-to-agent buying | Inventory owner | Pilot phase |
Microsoft's Publisher Content Marketplace lets AI systems license individual articles on demand. Perplexity's program pays publishers when agents cite their content via Comet Plus subscriptions. Omnicom's Ad Context Protocol — built on Anthropic's MCP — lets AI agents negotiate media buys directly with publisher inventory, cutting out DSPs and SSPs.
Each is real. Omnicom executed actual client buys in Q1 2026. Perplexity says checks are going out. Microsoft has seven co-designing publishers, People Inc among them.
The Scale Problem
Now look at what they're replacing.
Perplexity's entire publisher payout is 0.014% of the U.S. digital ad market. Even generous math — assume PCM and AdCP each contribute ten times Perplexity's pool — gets you to $1.3 billion. That's less than half of one percent. The licensing consolation, restated in protocol form.
People Inc's Hedge
People Inc is positioned across all three futures. They co-designed PCM with Microsoft. They have licensing deals with OpenAI and Meta. D/Cipher, their ad platform, could plug into AdCP's agent-to-agent framework. Vogel has called Microsoft's approach "essentially a pay-per-use market" — the first publisher CEO to publicly endorse a specific protocol architecture.
This is smart hedging. It's also what you do when you can't tell which standard wins — or whether any of them will. Google's ad tech empire just lost its antitrust case, with remedies expected mid-2026 and publishers piling on private lawsuits. The forced unbundling of Google Ad Manager might reshape the ad ecosystem more than any protocol. Or the appeals could take five years and change nothing.
The old economy is collapsing faster than the new one is arriving. Agentic traffic is growing 7,851% year over year. Open web display budgets are shrinking 30%. The three protocols are real, directionally correct, and — at current scale — irrelevant to the P&L.
Three percent was a description. It's becoming a ceiling.