I built a system to measure whether AEO was working. The proof showed up first.


Issue No. 04 - June 30, 2026

Last week I started a project I should have started months ago: setting up a real way to measure whether our AEO work is actually doing anything. That meant finally sitting down with GA4, which stands for Google Analytics 4, and which I'll now explain the way I wish someone had explained it to me.

Think of it as the thing that tells you where your website visitors came from. Did they click a link from Google search, from LinkedIn, from an email, from typing your address directly into the browser. GA4 sorts all of that into buckets so you can see it. I have technically "had" GA4 running on our site for a while, the same way some "have" a gym membership. It was there. But (maybe) not really being used.

What I actually did last week was make sure GA4 was properly hooked up to the website (there's a piece called a "container" that has to be installed and configured correctly, which took some doing) and added something called UTMs to our newest content. A UTM is tag you stick on the end of a link so that when someone clicks it, the system knows exactly which post or campaign sent them. Without it, a lot of traffic just gets dumped into a generic "referral" bucket and you lose the detail. The goal was to be able to see, specifically, whether traffic was coming from an AI source (like Claude, ChatGPT or Perplexity) instead of just an unhelpful "other."

While I was in the middle of that setup, something happened that made the whole project feel different. It's not just one week anymore. Sales call after sales call, when I ask how someone found us, I keep hearing some version of "I asked ChatGPT" or "it came up when I searched with AI." This has been happening for weeks now, not a single lucky stretch.

I still don't have a dashboard. I don't have a single chart. And I already had proof that something is working, weeks before the tooling could show it to me.

It's been a couple months since I started the AEO push, rewriting landing pages, fixing schema markup, structuring content so AI tools could actually parse it. I didn't expect to hear anything back this soon, especially not directly from the people we're trying to reach.

What struck me is how out of order this felt. I'd been thinking about measurement as the thing that comes first, you build the tracking, then you wait for the data, then you find out if it worked. Instead the signal showed up before the system to capture it did. A handful of conversations told me more than any report could have, and I almost missed it because I wasn't asking the right question on those calls until I happened to.

That's the part I keep coming back to. "How did you find us" is a question that's been available to ask this whole time. It costs nothing. It doesn't require GA4, UTMs, or any of the setup I spent last week on. And it gave me the first real evidence that the AEO work is landing with actual buyers, not just with whatever Google's crawler thinks of our pages.

The GA4 work still matters. Anecdotes from sales calls aren't the same as seeing it at scale, and I want to know which specific pieces of content are doing the work so I can make more of what's working and less of what isn't. But last week reordered things for me. The measurement system isn't what tells you the work is real. It's what tells you what to do next once you already know it's real.

Since then I checked, and I have an actual number now, even if it's a small one. GA4 is tracking a separate line called "AI Assistant," traffic that came in through tools like Claude, ChatGPT or Perplexity instead of getting lumped into "referral" the way it used to. Over the past few weeks it's 58 sessions. Less than one percent of our total traffic. I'm not going to dress that up as a wave.

Here's the part that got my attention. Those 58 sessions have a higher engagement rate than almost every other channel on the site, second only to organic search. People who land on us through an AI tool are sticking around and doing something once they get there. It's a small number, but it's not a hollow one.

I wanted to write this down now, while the order of events was still fresh, because I think the order is the actual lesson. The sales calls told me the work was landing before the spreadsheet did. Now the spreadsheet is starting to agree.

Where this lands in the three buckets

Immediately actionable: Start asking everyone who contacts you "how did you find us," and actually write down the answer somewhere, even a simple spreadsheet. You don't need any tooling for this, and it might tell you something your analytics won't show for weeks.

Start thinking about: If your website has Google Analytics (most do, even if you've forgotten about it like I had), you can create a custom channel group that specifically captures AI referral traffic. Inside GA4, under Admin, you tell the system to look for visitors coming from domains like chat.openai.com or perplexity.ai and give them their own label. Without that step, the traffic gets buried under "referral" or "unassigned" and you can't see it clearly.

Next phase: Once you can see which specific pages or posts are driving AI referral traffic, use that to guide what you create next. The goal isn't just to know AEO is working in general, it's to know which parts of it are working so you can do more of those.

I'm glad you're here.

~meghan

Meghan Brenner is COO at JB Sales and founder of The Operator's Notebook. The Muddy Middle publishes every Tuesday.

The Operator's Notebook

The Muddy Middle is a weekly newsletter for non-technical operators figuring out AI in real time. No tidy conclusions. Just honest notes from the muddy middle.

Read more from The Operator's Notebook

Issue No. 13 - September 1, 2026 There's a course for everything now. AI for beginners. AI for operators. AI power users. AI mastery. Become an expert in a weekend. Get certified. Get ahead. I've skipped all of them. Deliberately. That's not because I think learning is a waste of time. It's because the framing is wrong. "Mastery" implies a destination. A body of knowledge you work through, complete, and then apply. That model works for a lot of things. It doesn't work for this. The...

Issue No. 12 - August 25, 2026 When I first started using AI seriously, I read everything I could find about how to do it right. Set clear instructions. Tell it who it is. Tell it who you are. Give it context about your tone, your audience, your goals. The more specific, the better. So I did. I built it out carefully. Then I started seeing a different take. Creators I follow, people who work with AI every day, started posting about pulling back on instructions. The argument was...

Issue No. 11 - August 18, 2026 I've been cheating on my AI tools. Not dramatically. Just shopping around. For the last few weeks I've been deliberately splitting my work across different platforms. Strategy questions go to one. Creative work goes to another. Not because I made a formal decision to do it but because I started noticing the outputs felt different depending on where I asked. And once I noticed, I couldn't stop noticing. This isn't a platform review. I'm not going to tell you...