Follow one request through your stack. A Live Search Agent asks for your article, the content delivery network (CDN) receives it first, checks its cache, and hands back HTML in a few milliseconds. Your ad tags, analytics script and consent banner are all inside that HTML as plain text, and the agent has no use for any of them. It extracts the paragraphs it needs and leaves. Edge monetisation is the practice of acting during that exchange, in the CDN, where the request is still visible and the response can still be changed, instead of waiting for a browser stage that this visitor will never reach.
What does the CDN edge actually do?
A CDN is a distributed layer of servers that sits between visitors and a publisher's origin server, caching content close to the reader. Every request passes through it and every response passes back through it. That makes it the one place in the stack that sees all traffic, human or automated, before anything is rendered.
Modern CDNs also run code at this point. Cloudflare Workers, Akamai EdgeWorkers, Fastly Compute, and CloudFront Functions and Lambda@Edge all let a publisher run logic on each request without touching the origin. Cloudflare's documentation, for instance, describes Workers running at the location nearest the user, with a 128 MB memory limit per isolate and CPU time (the time spent executing code, not waiting on network calls) of 10 milliseconds per request on the free plan and up to five minutes on paid plans, with a 30 second default. That headroom is why a detection step, a pricing decision or a small auction can run inside the request itself.
Why can't browser-side monetisation reach AI agents?
Traditional monetisation runs after the page arrives. The browser executes the analytics tag, the header bidding auction, the affiliate script and the paywall check. AI retrieval skips that stage. The agent reads the response as text, so none of those scripts fire and the publisher earns nothing from the visit. Our guide on why AI bots do not show up in Google Analytics covers the measurement side of the same gap.
The scale is no longer marginal. Fastly's Q2 2025 Threat Insights report, published on 19 August 2025 and drawing on 6.5 trillion monthly requests across more than 130,000 applications and APIs, found that automated bots made up 37 per cent of observed traffic. AI crawlers were almost 80 per cent of AI bot traffic, and real-time fetchers, the bots that retrieve a page because a person just asked a question, made up the remaining 20 per cent or so. Fastly recorded fetcher bursts of up to 39,000 requests per minute against a single site. Arun Kumar, a senior security researcher at Fastly, said that AI bots were "introducing new complexities for digital platforms". Fetchers matter most for monetisation because each one corresponds to a live user answer, which is the closest thing to an impression that this traffic produces.
What are the two models of edge monetisation?
The phrase now covers two different commercial models, and mixing them up is the most common source of confusion.
The first is access monetisation: charge the bot at the door. The edge identifies the requester, and if it is an AI agent it returns an HTTP 402 Payment Required response with a price. Cloudflare's pay per crawl, which entered private beta on 1 July 2025, works this way. A crawler that sends no payment intent receives a 402 with a crawler-price header and can retry with an exact-price header to accept the terms. Alternatively it can send a maximum price up front. Requests must be signed with Ed25519 keys using HTTP Message Signatures so that identities cannot be spoofed. Publishers choose Allow, Charge or Block for each crawler, and Cloudflare acts as merchant of record. AWS WAF AI traffic monetization, announced on 15 June 2026 and available only on web ACLs attached to CloudFront distributions, applies the same 402 logic using the x402 stablecoin protocol. Akamai takes a partner route: in a 17 September 2025 post, Emily Lyons described edge detection of AI bot requests feeding either TollBit, which applies subscription, licensing or usage-based terms, or Skyfire, whose Know Your Agent tokens verify identity and collect payment before content is served.
The second is content monetisation: earn from what the agent reads once it is allowed in. Here the edge does not charge the bot. It classifies the retrieval, and for an eligible Live Search Agent request it can run an auction and place a relevant brand fact into the response as editorial context, with revenue paid by the advertiser rather than the AI company. blankspace works this way. It states that its auction runs in under 50 milliseconds at the edge and that revenue is attributed the same day, though those are the company's own figures, not an industry benchmark.
How do access and content monetisation fit together?
They address different questions and can run side by side. Access controls answer who may fetch the page and at what price. Content monetisation answers what the response carries once a fetch is permitted. A publisher might block low-value scrapers, charge identified training crawlers, allow Search crawlers that return referral traffic, and monetise Live Search retrievals through the content layer, all at the same edge.
Cloudflare's 15 September 2026 defaults show how quickly the access side is moving. New domains, and existing free customers who have not changed their settings, now have the Training and Agent categories blocked by default on pages that display ads, while Search stays allowed. A publisher who blocks Agent traffic at the edge is also deciding whether live retrievals can be monetised at all, so the two decisions should be made together. The detail is in our piece on Cloudflare's September 2026 AI crawler defaults.
Can you add edge monetisation without changing your CDN?
Usually, yes. AWS's own builder guidance, published on 18 September 2026 by Eitav Arditti, describes four patterns for pairing a primary CDN with CloudFront's WAF monetisation: redirect, direct, loop and gate. It lists Amazon CloudFront, Akamai, Cloudflare and Fastly as CDNs whose client IP is preserved through standard headers, and it notes that the Monetize action can work independently of AWS's own bot detection. In practice, most publishers can add edge logic beside their current setup. Content monetisation of the kind blankspace describes is deployed as a worker on the existing CDN, with no change to the CMS, the origin or the human ad stack. Whichever model you choose, check three things first: that human traffic short-circuits untouched, that bot identity is verified rather than inferred from a user agent string alone, and that the logic stays well inside the CDN's execution limits.
What can go wrong?
Edge monetisation inherits the weaknesses of bot identification. A user agent string is trivial to forge, which is why signed requests are spreading and why verification tiers matter. Classification errors cut both ways: treating a human as a bot damages the reader experience, and treating a bot as a human gives away a retrieval for free. Payment models also depend on agents being built to pay, and many are not yet. Finally, anything inserted into a response should be labelled, relevant and brand safe, a subject our guide to brand safety in AI answers covers in full.
Frequently asked questions
Why monetise at the CDN instead of in the browser?
Because AI agents read the HTML and leave without running page scripts, so browser-side tags never fire. The CDN edge is the only layer that sees every request before rendering, which makes it the practical place to measure and monetise agent traffic.
Does edge monetisation slow down my site for human readers?
It should not. A well-built edge rule identifies a human visitor and passes the request straight through, so no extra client-side JavaScript is added and the page loads as before. Test this on your own traffic before launch, because performance depends on how the logic is written and the limits of your CDN.
Is CDN-layer monetisation the same as pay per crawl?
No. Pay per crawl charges an AI crawler for access and returns a 402 if it has not paid. CDN-layer content monetisation earns from the response once the agent is allowed through. They are complementary, and a publisher can use both at the same edge.
Which CDNs support edge monetisation?
Cloudflare, Akamai, Fastly and Amazon CloudFront all run code at the edge, and each has its own bot-management or access-monetisation tooling. AWS WAF's Monetize action works only with CloudFront distributions, but AWS documents patterns for pairing it with Akamai, Cloudflare or Fastly.
Can it monetise human traffic too?
The edge sees all traffic, but this opportunity targets AI agent traffic that browser-side tools cannot reach. Human visitors continue to be monetised by the existing ad stack, which edge monetisation leaves alone.