223: I Scored 33/100 on Cloudflare's New Agent-Readiness Scanner

I ran the Cloudflare Agent Readiness scanner on nohacks.co, got 33 out of 100, and felt insulted. One toggle later the score was 67. That gap is where this episode starts. My case: the web is splitting into two economies. A retailer-friendly agentic web where AI-referred traffic now converts 42 percent better than human traffic. And a page-view-driven publisher web losing 20 to 90 percent of its Google traffic in a year. This split is structural, not a rebalancing.

Chapters:

• 00:00 Opening
• 00:41 The 33, then 67 score on nohacks.co
• 02:46 The thesis: the web is becoming two webs
• 03:55 Cloudflare Agents Week: 28 shipments in five days
• 05:35 The data stack: Adobe 393%, the conversion inversion, publisher declines
• 10:08 The mobile-first parallel
• 11:21 Why this split is permanent
• 13:12 The "publishers will adapt" and licensing-deal rebuttals
• 17:02 Three sector-specific moves
• 19:34 Closing: run the scanner, reply with your score

Key Numbers

• Cloudflare shipped 28 agent-infrastructure pieces during Agents Week (April 13 to 17, 2026)
• AI traffic to US retailers grew 393 percent year over year in Q1 2026 (Adobe)
• AI-referred traffic converted 42 percent better than non-AI in March 2026, after converting 38 percent worse in March 2025
• Google traffic to publishers down 33 percent globally; some local publishers down 25 to 90 percent
• Automated traffic growing 8× faster than human traffic year over year

Three Takeaways

1. The split is structural, not a rebalancing. Retailers fit the agentic web because agents complete the same action the website wants. Page-view publishing does not, because the agent summarizes instead of sending a human to see an ad.
2. The composite score is a trigger, not a target. Ignore the number. Read the per-check list and fix the signals that ship against real agent runtimes today.
3. Watch the infrastructure before the mainstream catches up. The mobile-first rebuild shipped years before the indexing caught up. This is the same gap, different shape.

What to Do

• Run isitagentready.com on your website. Toggle the category, re-run, compare.
• Transaction-driven revenue: agent readiness is tied to revenue. Fix the checks that ship against real agent runtimes now.
• Page-view-driven revenue: model your P&L with 30 percent less traffic this quarter. If it breaks the business, start diversifying today.
• Reply to the newsletter with your score. I read every reply.

Sources

• Cloudflare Agent Readiness Score
• Adobe Q1 2026 AI Traffic Report via TechCrunch
• Publisher traffic drop (Press Gazette)
• What is the Agentic Web (nohacks.co)

No Hacks is a podcast about web performance, technical SEO, and the agentic web. Hosted by Slobodan "Sani" Manic.

Transcript

[00:00:00] Hey there. Before I start this week's episode, I wanna share something with you. I looked at a podcast analytics tool and I noticed that number one, no hacks, has a lot more ratings than I thought it would. That was a pleasant surprise. And number two, they're all five star ratings, which I'm just shocked.

[00:00:16] So if you have rated the podcast in your favorite podcast platform. Thank you so much. If you haven't, please consider leaving a rating or review whatever works for you. 'cause that really helps. Onto the episode.

[00:00:41] Last weekend, I was catching up on everything that CloudFlare has announced in their. Agent weak and I saw a tool that they call Agent Readiness Checker. It's at, is it agent ready.com? Which smart. Had to test No hacks.co This podcast website, the one that you are on, if you ever read anything, I write it, it scanned it, it was fast, it gave me a score.

[00:01:08] It was, uh, 33 out of a hundred. A website. I spent a lot of time building and making AI and machine readable that hurt. It was level two, it was bought aware or something silly like that. So, it's a number. So let me, let me have a look. Lemme see what actually got checked. Robots D xt. Structured data marked down on request M, CCP endpoint, lms, dxt, LLMs, dxt.

[00:01:37] A lot of payment and authentication signals. And just some these, my website passes. Some of it it doesn't, and some of them like the payment checks are off checks. MCPN pointed. Why? 'cause no hacks.co is a blog. It's a content website with podcast episodes embedded. I don't sell things. I don't run agent workloads.

[00:01:59] I just looked at the agent readiness thing. Once again, I see there's a tiny little thing where you can select a website category. It's almost invisible, so default all checks. I switch it to content site and run the same scan once again. I got 67 out of a hundred. Okay. I can live with that.

[00:02:19] I'm not thrilled. I need a hundred of course. But you know, you flip one toggle and the score just doubles. The interesting number here is not 33, it's not 67. It's the gap between them because a tool like this did not exist two years ago, and now he does. And I'm sitting in front of a scoreboard for a web that did not exist two years ago.

[00:02:46] The web is becoming two webs, and I'm going to say that a bunch of times. , This episode. I wanna walk you through what I mean by that and where I think the obvious take is wrong and what that means for whoever is listening to this episode right now.

[00:03:02] And I wanna be clear about something. This episode is going to talk a lot about something that CloudFlare has shipped this month in their agents week. But this is not a CloudFlare episode. What they shipped is the evidence of something bigger. The web is being split into, and that split is not neutral.

[00:03:20] And you could say it's not fair.

[00:03:39] This scanner exists because somebody believes there's something worth measuring here. Two years ago, nobody believed that. Nobody thought about that. That shift is what I want to walk you through.

[00:03:55] So let's look at what happened. CloudFlare agents week ran from April 13th to April 17th, 2026. They shipped 28 things in five days. CloudFlare handles roughly 20% of all global internet request traffic. That is a lot, and that is why this matters. So what does this pace tell you?

[00:04:21] It doesn't say CloudFlare is winning.

[00:04:24] No, I'm, I'm not going to say that. The real read here is this is what infrastructure looks like when a vendor has decided to use emerging tech to be its next traffic tier infrastructure does not ship at dis space for speculation purposes.

[00:04:46] So the obvious question is what are they building? What's there to build? And for anyone in this audience, what does any of that have to do with your website?

[00:05:00] I've been watching this pattern since late 2024. Agent traffic started showing up in logs that used to be. Almost purely human plus Google Bot and some random crawlers vendor APIs are getting an agent facing site. The shift from optimize for search engines to optimize for the thing that tells search engines what to say is happening.

[00:05:24] The CloudFlare week is the week that all of that just became undeniable for the broader public.

[00:05:33] 

[00:05:35] Here's some specific data that made me commit to this two waves idea. There are three numbers in three different sectors all within a month. Adobe Analytics published their Q1 2026 AI traffic report on April 16th last week. The number that headlines it is that AI traffic to us retailers is up 393% year over year.

[00:06:04] I'll have to say something about this report before I continue pulling those numbers and, and sharing another number that you want to be aware of as. Adobe is releasing these numbers, and I believe they're correct because those are hard facts that if they're making this up, they could really get in trouble, but at the same time, they're recommending a fix, which is their Adobe LLM optimizer.

[00:06:28] So the data might be a hundred percent true. It's still wrapped in marketing. And thank you to my friend El Arts for pointing this out.

[00:06:39] Now the second fact that they shared and that you should be aware of. AI referred traffic converts 42% better. The non-AI referred traffic in March, 2026 compared to March, 2025 when it was 38% worse. And Adobe's line from this report clearly says that many websites are not entirely readable by machines.

[00:07:04] And for an AI to refer a user to you, it needs to be able to read you that. That's just how it works.

[00:07:15] Let's look at something that came from Google. Google Shared Google, uh, publisher traffic.

[00:07:22] Now let's look at another industry. Let's look at publishing, web publishing to be specific. According to a press Gazette report from November, 2025, I think the traffic Google sends to publishers is down 33% globally. Local publishers. It's down 25 to 50% and some individual publishers reporting 20 to 90% traffic drop year over year.

[00:07:52] So it's the same shift, it's the same timeline. It's two completely, completely different economies.

[00:08:01] And that's what I mean when I say two webs. This is not a metaphor. It's two different economies of traffic that are converting at different rates that are going in completely opposite directions

[00:08:14] and the thing that breaks my brain a little is that this is a year over year comparison. That means 12 months ago, the thing we're measuring right now did not show up as two webs. It showed up as one web with kinda some weird noise at the edges. 12 months ago, agents converted worse than humans and it was something that might be worthwhile or worth paying attention to in the future.

[00:08:41] 38% worse than. Humans today in retail, what agents sends you converts better by 42%. I don't think inversions like this can on invert.

[00:08:57] And let's look at two more facts. Number one, automated traffic is growing at a rate that is eight times faster than human traffic year over year. This is 2026 data according to A-C-M-B-C report, which is a weird source, and this is why these 28 shipments in five days by CloudFlare make more sense.

[00:09:26] Agents and ai, they're a new visitor class. The Agentic web is currently severely underbuilt. CloudFlare is trying to fill that layer. The readiness scanner, the browser run with web MCP, the managed authentication for access email service for agents agent memory. By the way, some of those I wrote about at the No Hacks blog, so you just go to no hacks.co and you can read those articles there.

[00:09:54] All those are pieces of something that. It has to exist for that 393% growth year over year to keep going up.

[00:10:05] 

[00:10:08] And I watched this shape before. Remember mobile. Remember Steve Jobs standing on that? Stage and announcing the iPhone and everything that happened in the years and even decades after we got responsive web design. Then the browser engines kind of caught up, and then Google shipped to mobile first indexing as the default ranking behavior.

[00:10:33] We got years of infrastructure leading to Google, officially going mobile first. The people who read the infrastructure signals early, they had already rebuilt by the time the indexing caught up and they were just ready for it.

[00:10:50] I think we are sitting inside that gap right now. But it's not mobile. It's for agents, it's for the machines. The infrastructure for this is being shipped, the indexing, the conversion behavior, the traffic behavior has started to flip, but the mainstream web publishers included is still reading this like it's the 2010 internet with some weird AI noise added on top of it.

[00:11:21] This is where it's better to be honest than trying to be helpful. The consensus read on the two webs thing is this is a rebalancing and the publishers will adapt and the agents will become another monetization channel and you know, things will even out eventually. I don't believe that. I think this split is permanent.

[00:11:40] And let me show you why retailers can monetize this second web, the agentic web, because the agency is doing exactly the thing that the retailers want. They're picking products, they're reading the specs, they're checking out, this is happening. If you look at the Adobe report where the 42% better conversion rate for AI referred traffic, that's on agent shopping.

[00:12:05] That's a human sent by an agent doing the shopping, but the AI brought the human to that website, so it's still completing a lot of that journey. That's real. That's real money. And every agent visit, or human sent by an agent is a potential buyer that has a higher chance of converting

[00:12:24] most of the publishers, web publishers, they monetize by page views. They need display ads, affiliate links, video inventory. When the agent's task is to summarize the page in a chat interface or wherever none of that happens. The 393% traffic growth year over year number. Doesn't show up in their analytics because that traffic is really not a human landing on the page.

[00:12:50] There's no ad being displayed, there's no affiliate links being clicked. There's no purchases after those affiliate links are being clicked. So the agent just reads the content, synthesizes the content, and the user never gets to interact with you or your brand or see your website. And the obvious pushback is, yeah, publishers can adapt.

[00:13:12] Just going to be AI native ad formats that exists to some extent, subscription pushes. Sure. Licensing deals with AI labs. It happens and some publishers will do all of that, but the ad supported content below a certain traffic threshold has always been very fragile, and the business model had thin margins before the agents showed up.

[00:13:40] And when the traffic drops 30% globally and 50% locally in one year, ah, the fragility of that threshold moves. And because below the new threshold, the publishers don't really adapt. They stop existing. Think, adapt that's a word where you use when there's a version of the business on the other side.

[00:14:04] And for some sectors. We need to be real. There simply isn't a new reality on the other side if you keep doing the things that you'll be doing so far. So I don't call it rebalancing, I, it's sorting, retailing, transaction driven businesses. That, that, that's the second web. They, they adapt and they work in the agentic web.

[00:14:27] What are the economics of that web? Reward them, add, monetize content. That's a huge chunk of the web. It used to be it's getting smaller, but it's getting smaller faster now. All that content kind of loses a home in the second agent web and the growth of agents share visiting your website is bad for you here.

[00:14:52] So for some sectors the new floor is not. 20 or 30% lower than what it was, it's, it's zero. And that is not a rebalancing, that is a permanent restructuring of who the web works for.

[00:15:08] 

[00:15:10] And since I keep saying this is not a CloudFlare episode, let me be specific about what their role is in all of this, the pace of their announcements in this agent's week, and not just this week. It makes the second web function faster, but it doesn't do anything to soften the collapse of the businesses that are better suited for the first web.

[00:15:37] So a good scanner for agent readability doesn't really restore publisher traffic. No matter how high you score a browser run with web MCP support doesn't monetize a news article.

[00:15:51] And CloudFlare is building the infrastructure that lets the second web exist, and that's a real thing that is worth building, but nobody is building anything to save the first web, the one, the only one we've had for a very, very long time because. There's no infrastructure to build there. It's not an infrastructure problem.

[00:16:12] This is an economic problem,

[00:16:14] and there's one smart pushback. I would like to name here the licensing deal argument. The idea that the publishers can license their content to AI labs, and that becomes the new revenue model. There are two problems with that. As a general solution, number one the aggregate payment pool is small compared to the ad revenue we replace.

[00:16:35] So the math doesn't really math at scale. Number two, the licensing deals reward the publishers that are big enough to negotiate the New York Times. The Guardian, i, I, I don't even know, not the middle and the long tail of it, and the middle and the long tail of it is where most of the web has always lived.

[00:16:57] So what does any of this mean for your website?

[00:17:02] The practical move depends on the sector of the web you're in, and I'll give you three quick things. If your website's revenue comes from transactions, e-commerce, SaaS, signups, lead generations, any conversion flow. Agent readiness just became tied to revenue. The retailer pattern is not live. You can run the scanner on your website.

[00:17:25] You can read all the checks, you can learn about them. Don't look at the total number that you're getting. Everyone will share the number. I shared the number, but the number is not the story. Look at all the checks and fix the ones that matter to you against Real Agent Runtime Today, The Readiness Tool is it agent ready.com.

[00:17:45] Go there and check your website. It, it takes less than a minute. If your website revenue depends on the on page, view driven display ads, affiliate links, video impressions, diversification of what you're doing is not a we should do this anymore. It, it, it, it's your survival ticket. So the question to answer today is this, what percentage of your revenue depends on page views and what does your p and l report? Look, if you had 30% less traffic, just model it. If the answer breaks the business is already breaking. Start that conversation about whatever other revenue stream. It could be licensing.

[00:18:33] It could be subscription. Anything adjacent, I mean, anything. Start that conversation today and don't leave it for when it's too late.

[00:18:46] If your product might be used by an agent acting on behalf of a user booking, buying, querying, doing any kind of transactions, the web MCP plus oath for excess layer, that CloudFlare ship that that's not your, we should look at this. In 12 months, no, this is something that you should look at right now because it became something that is real with the CloudFlare agent runs.

[00:19:16] It's extremely easy. It's super easy to learn. Any developer will figure it out right away, and this seems to be getting more and more real.

[00:19:34] So back to back to this website, back to no hacks.co. Just scan on that. Is it Agent ready.com gave me 33 out of a hundred. I felt insulted. I felt called out because I, I spent a lot of time building this website and making it AI ready and readable. But then I flipped one toggle that I didn't see when I ran it for the first time, and I got a 67.

[00:20:00] The point is not the number. I'll get a hundred by the way but that's not the point. The point is that there's a scoreboard for this thing right now, and two webs means two different scoreboards. The one CloudFlare is shipping with, is it agent ready.com is for the second web for the agent web. They wanted measures whether the first web, the only one we've had.

[00:20:24] It pays for itself whether the publishers survive, whether the ad monetized content survives. Nobody's shipping that because it's not a measurement problem. It's a structural problem. I think the split is permanent. I think the retailer side wins the next three years.

[00:20:43] I think the content monetization side never goes back to where, where it was. I would love to be wrong. I'm watching for all the signals I can to show me that I am wrong. But in the meantime, the practical question is which side of this split your website is on. That's the question we're answering right now.

[00:21:05] So go to that. Is it agent ready.com? Scanner? Test your website and look at all the individual checks and what they mean and. If you're on the newsletter, just reply to my latest email and tell me what was the score. Were you surprised? Were you shocked? Were you insulted like I was? And I read every response.

[00:21:28] So go to no hacks.co/subscribe and just, send me a note. This is no hacks, episode 223. I'll talk to you next week. 

[00:21:40]