Who Pays for the Agentic Web — the Agent or the Advertiser?

Same traffic, two customers

Every one of these posts started from the same shift: the visitor is increasingly not a human but an agent. What I didn’t fully appreciate until I’d built both sides is that this single shift spawns two entirely different business models, and they point in opposite directions.

• The agent as customer. The agent pays the publisher for access. This is the x402 / paid-access series [4]: an HTTP 402 quotes a price, the agent settles a micropayment, the edge verifies, the content flows [1]. Incentive: scarcity. The content is worth gating because it’s worth paying for.
• The agent as audience. An advertiser pays the publisher to influence the answer the agent gives. This is the Ads for Agents series: sponsored content written for the crawler, dropped into the markdown an AI ingests, measurably tilting the recommendation [2]. Incentive: reach. The content is worth giving away because the answer it shapes reaches everyone who asks.

Same bot knocking on the same door. One publisher wants to charge it; another wants to pay to reach it. That’s not a contradiction. It’s two products.

The two rails, side by side

The mirror is exact. Paid access makes the agent a paying customer and treats each fetch as revenue. Agent ads make the agent a distribution channel and treat each answer as an impression. The first wants the content scarce; the second wants it everywhere.

Why segmentation, not a winner

My bet — labeled as analysis, because no one has the revenue data yet — is that publishers won’t pick one rail. They’ll segment by content type, the way they always have.

• Tolls fit must-have and timely content. Breaking news, proprietary data, market-moving analysis: content an agent needs right now and can’t substitute. Scarcity pricing works because the buyer has no alternative. This is where the x402 rail earns real money.
• Ads fit evergreen and authority content. Explainers, reviews, “best X for Y” roundups, content where being the answer the agent gives is worth more than a per-fetch toll. Reach pricing works because the value is in influence, not access.

There’s a precedent that makes this feel less speculative: broadcast television runs carriage fees (cable operators pay to carry the channel) and ad sales, on the same programming, for decades. Publishers are institutionally comfortable running two revenue models over one catalog. The agentic web just gives them a new pair.

The uncomfortable part: ads erode the thing tolls sell

The two rails aren’t neutral neighbors. In the Ads series I showed that even labeled sponsored content bends a retrieval agent’s answer, and that the effect is larger on the cheaper models a free tier would actually run [2]. The market has already felt this: a major answer engine that experimented with ads found that the ads eroded trust in the answer, which is precisely what makes a clean, unbiased answer something a user would pay for.

So the two rails feed each other. The more ads pollute the free tier’s answers, the more valuable the paid, unbiased answer becomes, and the more sense the x402 toll makes for the content where trust is the product. Bias on one rail is a sales pitch for the other. That’s the closing loop of the whole arc: brands pay to bias agents; users pay for the answer that isn’t biased [3].

What this means if you’re building on AWS

The optimistic read, and the one I actually believe, is that this is about optionality, not a forced choice. A publisher on AWS can run both rails on the same origin:

• Tolls: AWS WAF AI traffic monetization returns the 402 and settles the x402 payment at the edge [1] (Cloudflare’s Monetization Gateway ships the same pattern [5]): the paid-access rail, no origin code to maintain.
• Ads / agent representation: serve the agent-facing version of the page (the one an AI crawler ingests) from the same distribution, with sponsored units clearly labeled: the ad-funded rail.
• Transparency as a product. The most durable position across both rails is the same one: be explicit about what’s paid, what’s sponsored, and what’s neutral. When ads erode trust, the publisher who labels everything and offers a clean paid tier owns the high-value end of the market.

Two rails, one catalog, segmented by content type. The agent pays for what it needs; the advertiser pays to reach it through what it gives away. Both will exist, because they’re not competing for the same content — they’re pricing different content the way it deserves.

Which rail would your content run on, and would you be honest with readers about which one they’re on? I’d genuinely like to know.

The series, in one place

Paid access (the agent pays)

• HTTP 402: The 30-Year Placeholder
• I Built the Agent That Pays
• Getting Paid by Agents
• Why Doesn’t the Web Just Use HTTP?
• Terms in the 402

Agent ads (the advertiser pays)

• Your Website’s Second Audience Just Got Its Own Ads
• When Ads Bias Your Agent
• Who Pays to Un-Bias Your Agent?

Sources

• [1] AWS WAF AI traffic monetization — the Monetize action returns HTTP 402 and settles the x402 payment at the edge.
• [2] When Ads Bias Your Agent — my experiment showing sponsored content measurably tilts a retrieval agent’s answer, more on cheaper models.
• [3] Who Pays to Un-Bias Your Agent? — the two-tier agent economy; even labeled ads erode answer trust.
• [4] Why Doesn’t the Web Just Use HTTP? — the layer map for the paid-access rail.
• [5] Cloudflare Monetization Gateway — x402 charging at the edge, x402 Foundation under the Linux Foundation.

About the Author

Stefan Christoph is a Principal Solutions Architect at AWS, focused on agentic AI, media & entertainment, and helping builders move from demo to production. He writes about AI architecture, developer productivity, and the future of software.

This is a personal blog. Opinions expressed here are my own and do not represent the views or positions of my employer.

Learn more →

• AIAgents
• AgenticWeb
• ContentMonetization
• Advertising
• Strategy
• AmazonBedrock