Surfaced - AI Search Intelligence - November 2nd, 2026

Hope everyone's fall is in full swing. I'm really enjoying the pre-holiday work hustle.

Quick update: I've been really happy with how Surfaced is coming along and the work I'm getting to do. That includes helping people build programs from the ground up, helping agencies shape better offerings, and supporting very mature AI search programs as they tighten up their operations.

This newsletter has been rewarding in so many ways, and I appreciate everyone who's been along for the ride so far. It keeps me sharp and lets me give something back to everyone navigating this very disruptive time.

Quick offer: If you've got a burning question about AI search (the hard ones AI can't help with), send it my way. If there's a challenge you're working through or something you want advice on, just ask.

I don't know the first thing about sales, but I do know how to educate, enable teams, and build champions. So use me. Seriously.

Onward!

Google's October 15 update puts more links higher in AI answers and includes a test giving official pages their own section at the top when there's a specific task to complete. Google says this builds on how it already puts links front and center when someone wants to go to a particular site.

As a marketer, I love this. As someone who's typed a company name into Google just to get to its website, also yes. Please let me through.

This is where branded search gets interesting. Someone typing “Figma login” has probably made their choice. Someone typing “Figma reviews” or “Figma alternatives” is still figuring something out.

Same brand name, but a different audience, different intent, and a very different job for the results page.

So I wouldn't read this as Google giving every branded query special treatment. The distinction is whether you're trying to reach a destination or need help making a decision.

My take on SERPs in general: If I'm comparing options, a useful summary could save me time. If I'm trying to reach a website, making me scroll past an essay is just getting in the way.

The best search experience knows when to help and when to get out of the way. Google (eventually) makes things better for the searcher.

Clearer links are welcome. Whether they bring useful visits is what I'd measure next. Jury’s out.

Noah Greenberg at Stacker shared an idea I like. Before a piece of coverage comes out, pick the questions you'd hope to show up for and check how you're doing. Then check the same questions after it goes live. Save any answers that cite your new coverage.

Simple enough. And much easier than trying to recreate the “before” picture when someone asks if that big PR win did anything.

My addition: start with questions your buyers actually ask. Noah suggests using AI to generate prompts from the upcoming article. That's a useful starting point, but have someone who knows the customer check them.

And when the article shows up in an answer? Take the win. You can show that it was used as a source. Claiming it caused all the extra mentions, let alone new business, needs more work.

Run the questions more than once, save the answers, and note what else changed. You can also watch a comparable group of questions the article wasn't meant to help, though coverage can spill into other topics too.

I’ll leave it to you to justify the business value of a citation, but if you’re in PR, you’re used to this challenge. 🙂 SEOs have had to justify the value of links for DECADES.

Kevin Indig's Does AI drive more direct traffic? was excellent. He looked at 12 sites (a relatively small sample), mostly B2B, to see whether AI visitors were hiding in Direct.

Across the traffic he included from January through September, Direct sessions were up 140.9%.

But engaged Direct sessions were basically flat: down 1.0% in monthly counts, or up 2.3% when accounting for the different number of days in each month.

Lots more visits, buuut very little extra engagement. He concluded that bots were making the growth look a lot more exciting than it was.

He also ran 632 tests across four brands to see whether better AI visibility was followed by more branded searches or Direct traffic. Some results looked promising until he checked whether the timing actually held up. He called it inconclusive.

That doesn't mean AI has no effect. It means he couldn't reliably spot it in this data set.

But it's one of the first comprehensive Direct traffic data sets I've seen, and I might tweak how I talk about Direct as a proxy for AI performance.

Last issue, I covered how broken tracking can push AI visits into Direct. This is the other side of that conversation. We can't call every Direct spike an AI success story just because attribution is a mess!

Check who's visiting, where they land, and whether they do anything. Ask buyers what brought them over and validate your assumptions. Primary customer research is a huge untapped opportunity for most companies.

Direct is the junk drawer of your analytics. It’s been that way for a long time.

Jason Saltzman at AirOps shared this excellent early look at Muse versus ChatGPT, Gemini, and Google AI Mode. This gentleman is worth a follow.

Look at those differences!!

Muse averaged 25.1 citations per answer, compared with ChatGPT's 6.9. But the answers were much shorter: 777 characters versus 5,192.

It also pulled a bigger share of citations from reviews and user-generated content: 17.6% versus 3.4%.

Jason's take is that Muse seems more interested in getting to a decision than keeping a conversation going. Lots of reading, short answer, almost no questions back.

This is an early snapshot and some of the first Muse data I've seen. The post doesn't share the full methodology, though I trust Jason. We also don't know yet whether shorter answers lead to better decisions or more sales. Keep that in mind before anyone starts selling the definitive Muse playbook.

What interests me is how many sources can go into such a small answer. Your page could help inform the response without your product being the one it recommends.

My take: report these platforms separately. If you mash them into one number, you're hiding differences that could change what your team works on.

Pick the AI surfaces your audience actually uses and spend some time with the answers. And remember, we're all very early to the AI reporting party.

I dug into this in my newest piece, Your AI visibility score starts with the prompts you choose.

Here's the slightly ridiculous thing: you can make a score look much better just by changing the questions, and I’ve done it to prove a point before.

Same goes for margin of error. Evertune is very serious about being the best in the market at this, and that matters when you have a lot invested in search.

Say you show up in 80% of answers when the question names your brand, and 20% when it doesn't. In my made-up example, shifting the question list from 20% branded to 80% branded takes the overall mention rate from 32% to 68%. Every prompt gets the same weight and the same number of test runs.

Same performance in each group, and you have a happier dashboard. But nobody new necessarily found you.

That's why I care so much about the list underneath the chart. Once your manager starts watching that number, it will influence what your team spends time on.

The move this week:

Walk into your next QBR or performance readout with that confidence.

A client recently asked me this.

I get it. You want one number to follow. Your manager wants one north star metric to report on. Nobody is asking for another spreadsheet with 14 tabs and a glossary that requires a stats degree.

A visibility index isn't terrible. It gives you a quick read on how you're showing up across the questions it tracks. If you're taking the lazy route and only following one thing, fine, follow that.

I just wouldn't let it run the program.

For example, imagine you're trying to grow a particular service next quarter. Your big educational guide keeps getting cited, so the score goes up. Meanwhile, the pages explaining that service, answering buying questions, and helping someone choose you still need work.

The guide can be useful! But are we getting any closer to the thing we said we wanted to accomplish?

That’s the conversation I want to have before we get the balloons out. 🎈

The tool doesn't automatically know your plans for next year, which pages matter most to those plans, or what you've got the people and budget to fix. Some platforms let you customize the setup. Great, do that. Someone who understands the business still has to make the choices.

And then there's the recommended to-do list. Who decided item one matters more than item five? Based on what? There's a lot of blind faith in those priorities, and prioritization is subjective. I argue with platforms all the time.

Andrew, so like, what would you follow?

I'd start with what we're trying to sell and the questions people ask before they buy it. Are we showing up there? Is the information right? Are we being recommended for a reason that makes sense? Then I'd look at useful visits, inquiries, and what people tell sales about how they found us.

I'd keep the index alongside that. It can help flag a change worth looking at. It can't make the business decision for you. And it's so much easier to tie those questions to success.

My concern is that you can spend months chasing easy wins on that dashboard number. The score goes up quickly, the returns take forever, and everyone wonders what happened. Then you spend the next quarter digging into why.

If your dashboard is looking great but you're still struggling to explain what improved, I'd love to hear where you're getting stuck.

Appreciate you all. Until next time,

Andrew