Measuring AI Search Visibility: Harley Helmer on Prompt Tracking, Backlinks and Why You Shouldn’t Bet Against Google

ICYMI: Harley Helmer catches up with 10 Minute Martech on the four-quadrant approach to tracking AI search and why marketers shouldn’t abandon their SEO foundations.

“If you don’t measure something, you can’t improve on it.”

Your brand appears in an AI-generated answer one day and disappears the next. A competitor keeps getting recommended, but you’re not sure why.

In this episode of 10 Minute Martech, Harley Helmer, SEO Team Lead at Americaneagle.com, joins host Sara Faatz to discuss how marketers can make sense of those changes.

Harley shares his four-quadrant approach to prompt tracking, explains why brand visibility extends beyond your own website and considers what AI agents need to navigate digital experiences. He also makes the case for keeping backlinks and established SEO practices firmly in the conversation.

Harley’s Big Idea: Build the Baseline Before You Chase the Answer

AI search measurement comes with uncertainty. The same question can be phrased in countless ways, and repeating an identical prompt doesn’t guarantee an identical response.

“If you start tracking something now, four months, six months, eight months down the road, you will have trend data on at least some sort of data set that you can begin to look at.”

For Harley, the useful measure is competitive share of voice within the prompts being monitored. Is your brand appearing more frequently? Which competitors show up consistently? Are there particular types of questions where your company remains absent?

That information gives marketers a starting point for investigation. Existing SEO work may contribute to stronger AI visibility, or the results may suggest a need for more targeted content and offsite activity.

Starting now gives teams evidence to work with later, even while the measurement methods continue to develop.

Harley Helmer’s 10 Memorable Moments on 10 Minute Martech

1. Start Tracking Before You Have Everything Figured Out

Harley’s first recommendation is to establish some form of tracking when a company expresses interest in AI visibility. Waiting until the tools and methods are fully settled means losing months of potential observations.

The initial prompt set doesn’t need to answer every question about discovery. It needs to provide a consistent starting point that the team can monitor, learn from and use to identify areas for further investigation.

2. Look for Share of Voice, Not a Single Winning Answer

Seeing your company recommended in one response can be encouraging, but it says little about how consistently the brand appears.

Harley recommends comparing visibility across competitors over time. His example is a brand that initially doesn’t appear at all, then begins showing up in half of the tracked responses while still appearing below other websites. That reveals progress and a remaining competitive gap. It is more useful than treating one favorable answer as proof that the strategy is working.

3. Use Four Quadrants to Broaden Your Prompt Set

Sara raises an important concern: how do marketers avoid simply tracking the questions they imagine customers are asking?

Harley’s approach considers two dimensions: whether a prompt is branded or non-branded, and the questioner’s level of product awareness. Combining those dimensions creates four groups for developing and testing prompts.

The purpose is to move beyond questions that already assume familiarity with the company. A brand needs to understand whether it appears during broader discovery as well as when someone explicitly asks about it.

4. Follow the Question from Discovery to Comparison

Harley uses Progress Sitefinity CMS to illustrate how questions change as someone learns more about a product.

A person unfamiliar with Sitefinity might ask a general question about choosing a CMS or a capability they need. Another might ask, “Is Sitefinity a good CMS for X?” Someone comparing options could ask whether it is an alternative to another platform.

Taking a handful of questions and developing variations across the groups gives marketers a more deliberate tracking set, rather than a collection of prompts centered only on their own brand.

5. Let Patterns Guide Your Content Decisions

Once recurring patterns emerge, marketers can investigate where to act. If a competitor consistently appears for a question about implementation partners, what information might be supporting that recommendation?

Harley suggests several possible responses: adding relevant content to your website, contributing information to a third-party platform or publishing a press release.

These are potential tests, not guaranteed fixes. The value of tracking is that it helps teams choose a specific intervention and observe what happens afterward, instead of producing content without a clear reason.

6. Your Website Is Only Part of Your Brand’s Story

Harley observes that AI systems appear responsive to broader sentiment and repeated messages across multiple sources. He points to social media, data aggregators, Wikipedia and company websites as places that can contribute to that picture.

For marketers, this expands the scope of visibility work. An accurate, detailed website matters, but so does what other sources say about the business. Teams need to consider whether their expertise and positioning are consistently represented wherever people encounter the brand.

7. Adapt a Message Across Platforms

Harley distinguishes between repeatedly publishing similar content on your own website and reinforcing a message across different channels.

His suggestion is to take a message from your website, adapt it for social media, then adjust it again for another relevant platform. Each version should suit its setting while retaining the same underlying information.

This gives marketers a practical reason to coordinate website content, social activity and third-party contributions. The aim is to communicate consistently across several sources without creating competing pages on the same topic.

8. Accessibility Is a Useful Starting Point for Agentic Browsing

When Sara asks how websites can remain human while becoming accessible to agents, Harley points toward familiar territory: accessibility.

He sees similarities between what agents need to interact with websites and the principles covered by the Web Content Accessibility Guidelines. For teams concerned about agentic navigation, he recommends examining those standards as a starting point.

That is not a claim that accessibility conformance guarantees agent compatibility. It is a practical direction for designers and developers investigating how to make digital experiences easier to interpret and navigate.

9. Don’t Write Off Backlinks

“Backlinks are still as relevant as ever.”

Harley’s reasoning is that Google remains central to product discovery, while its AI experiences draw on parts of the search infrastructure it has already developed. In his view, marketers shouldn’t assume that established signals lose their value simply because the interface is changing.

He also acknowledges that Google doesn’t disclose exactly which ranking systems its AI products use. His argument is for continuity in the underlying work, not certainty about a specific formula for appearing in AI answers.

10. Pay Attention to the Context Around a Link

Harley contrasts repeated exact-match anchor text with the way someone would naturally recommend a company.

A genuine recommendation might link the brand name and describe the work it performed, rather than repeatedly linking a phrase such as “SEO services.” The surrounding explanation helps establish the relationship between the company and its expertise.

He connects this with how LLMs interpret broader context. The useful information isn’t confined to the linked words; it also includes what the passage says about the brand, the service and the particular project.

Harley’s Martech Hot Take

“Don’t bet against Google.”

During the conversation, Harley predicts that both Gemini and AI Mode will surpass ChatGPT in usage by the end of the year. It is a forecast rather than a settled outcome.

Behind the prediction is Harley’s confidence in Google’s existing technology and investment. The company has spent heavily developing search systems, and he expects that work to remain valuable as its AI products evolve.

For marketers, his position is a reminder to keep Google in view while investigating newer discovery channels. Studying AI-generated answers doesn’t require abandoning the content, links and website improvements that support search visibility. It requires measuring how those efforts translate into an expanding set of experiences.

Harley’s Reading List

Harley looks to people who share practical observations about search while challenging marketers to think beyond optimization tactics.

• Wil Reynolds, Seer Interactive – Harley appreciates Reynolds’ forward-looking approach to AI and LLMs, along with his straightforward delivery. He recalls a conference message about doing great marketing first rather than doing poor marketing for the sake of SEO. That principle aligns closely with his own outlook.

• Lily Ray – Harley also names Ray among the people he follows for updates and perspectives on where search is heading. He values voices that help him understand both what is happening now and how to think about the changes ahead.

His recommendations reinforce a theme running through the episode: stay curious about new technology, but keep the quality and purpose of the marketing itself in focus.

Listen to the 10 Minute Martech Episode

The Four-Quadrant Way to Track AI Search: Harley Helmer

→ Watch on YouTube

Next Up in the 10 Minute Martech ICYMI Series

Ana Mourão, brand-side marketer and author of Strategic Marketing Skills, joins Sara Faatz to discuss how marketers can lead technology decisions, design better experiments and know what to automate—and what to keep human.

Want to Keep Reading?

Here’s the full transcript so you can explore every insight from Harley’s conversation with Sara.

Sara: I’m Sara Faatz, and I lead community and awareness at Progress. This is 10 Minute Martech.

Harley: Backlinks are still as relevant as ever, and the reason for that is because Google is still, by and large, where product discovery happens. All that we know is Google has spent billions and billions and billions of dollars to develop this system. It makes sense that they’re gonna recycle some great technology. The message there is don’t bet against Google.

Sara: That’s Harley Helmer, SEO team lead at Americaneagle.com. Let’s get started.

All right, Harley, I am so excited to talk to you today. I’ve been really looking forward to this episode because the last time you and I met, you were doing a session about AEO, SEO, GEO, any kind of acronym you want, but in the SEO world, and I know that this is a topic that’s really gonna resonate with our audience, and you have so many interesting things to say.

I don’t even know where to start. I guess my question to you is, if somebody is looking today, they’ve been doing a lot of work from an SEO perspective, but they’re thinking, “I probably should think about SEO, EEO, AEO, GEO,” what would you tell them just to start with?

Harley: Tracking is a great place to start. If you don’t measure something, you can’t improve on it. The other area that I would say—and I would also caveat this, there’s a lot of “is it beneficial, is it not beneficial” debate—but some form of prompt tracking. There are a litany of tools out there now that do some form of prompt tracking.

All of it boils down to: if it’s not measured, it can’t be improved in any way, shape, or form. What we’ve found so far is the best way to get a sense for how you’re doing with the prompt tracking in particular is to try to understand a share of voice between competitors.

So, as we know, a prompt could be said in a million different ways, and even if you do happen to get the same prompt twice, the answer is different every time, and there’s all kinds [Sara: Right.] of studies on how AI systems are even designed to do that. But the fact of the matter is, if you start tracking something now, four months, six months, eight months down the road, you will have trend data on at least some sort of data set that you can begin to look at.

And so that’s where I think from an agency perspective, that’s where we’re leaning: anyone that even says that they may have an interest or are concerned with AI and LLM tracking, they say, “OK, let’s figure out something to track so we can track it now. And maybe the SEO work that we’re doing will have a positive impact on that, or maybe we will need to define specific tactics that are for addressing the things that we’re tracking within the tracking software.”

Sara: I think everybody should be doing just as you’ve said, but how would a company make sure that they’re not just looking in a mirror saying, “I think this is the prompt my customer uses”?

Harley: The way that we kind of approach prompt tracking is trying to understand branded versus non-branded, and then kind of like an awareness versus non-awareness of the product.

So non-branded, not aware would be a prompt or question about a particular topic. Maybe we’re talking about Sitefinity—it’s people who are not aware of what Sitefinity is, and so they’re doing a general prompt on a CMS, or one of the great things that Sitefinity does, and it’s unbranded, so there’s no brand. But then there might be a branded version of that question: Is Sitefinity a good CMS for X? And then you can get into some competitor—would Sitefinity be considered an alternative to another brand? And so we try to kind of break it up into those four quadrants—branded, unaware, non-branded, aware—and mix and match those, and then you can take a handful of questions and iterate that question through each of those four groups.

And then again, there’s that trend data of: OK, at the beginning of the year, you weren’t showing up at all. And now midyear, you’re showing up 50% of the time, but you’re consistently showing up below these other websites.

Sara: Right.

Harley: And so then you can start to understand, OK, now we’re seeing some trend data—we have an understanding of when we type this general type of question of “who is the best implementation partner,” we can start to see consistency, and then you can test for: OK, maybe now we wanna put content on our website about it, or maybe there’s one website that’s a third-party data aggregator that we could potentially influence by getting content on their website about it. We could put a press release about it. So you have the ability to sort of isolate, but again, this is much more—I mean, we always say SEO is a long-term play, but I think even this is even longer term, because you gotta have that time to understand trends.

Sara: One of the things that you kind of touched on were some of the—placing some of that content offsite, off of your own domain. Can you talk a little bit about what those signals are for offsite brand versus onsite, and how that kind of varies between traditional SEO and how we think about SEO today with AI involved?

Harley: What’s interesting is it seems as though AI systems are much more aggressive about general population sentiment, and overall repetition almost. And so while your website is one place where you can get this particular message out, it’s—in contrast, it’s bad for SEO to repeat the same message over and over and over again on your website from a cannibalization perspective. Whereas LLMs, they’re looking at social media, they’re looking at data aggregators, they’re looking at Wikipedia, they’re looking at your website, they’re looking at all of these different places. So if you have the opportunity, take one message on your website and tweak it for social, and tweak it again for maybe your data aggregator, and tweak it again for a different platform. And then you’re getting the same message, the same repetition, but across multiple surfaces, which is attractive to LLMs.

Sara: How do you think about agentic searches—meaning, if in one area we are optimizing content for humans and human ingestion, and then in the next we have to think about what the agent’s going to want to read or what it will pick up on, and that would be more on your own site. What are your thoughts on that? How do you create a presence that’s very innately human, but also accessible to agents?

Harley: So I would say from that perspective, there is a massive technical gap that I think web designers and web developers are going to have to understand—how agents, what agents need in order to interact with websites. I would say where the misconception comes in is it’s not anything necessarily new. If you actually look at the WCAG standards for accessibility, they’re very similar. So I would say if you are concerned about agentic navigation—how agents interact with your websites—go look at ADA compliance and the WCAG standards and pay attention there, because it’s very similar to what I’ve started to understand about how some of these agents are working, and how it compares to that particular aspect of website design and development.

Sara: When you think about, back in the early days of SEO, everybody needed backlinks, and you didn’t necessarily know how many you needed or where you needed them to be, but you knew you needed a lot of them—maybe some people had it more dialed in than that—but with LLM discoverability, or being surfaced in your favorite AI-powered answer engine, does that strategy change?

Harley: I would say backlinks are still as relevant as ever. And the reason for that is because Google is still, by and large, where product discovery happens. That’s a lot of how Google uses, or how Google determines, quality—or at the very least, popularity. And in conjunction with that, AI Mode and Gemini both use some of Google’s ranking [00:09:00] systems. Now, obviously, they don’t say which ones, otherwise we would figure out how to exploit them. All that we know is Google has spent billions and billions and billions of dollars to develop this system. It makes sense that they’re gonna recycle some great technology.

And then there’s one thing that I will add to that. Interestingly enough, Google has officially updated their—I don’t remember offhand if it’s their spam documentation or exactly where they put it—but when they are analyzing for link spam, so backlink spam, now they are looking for websites that have a high ratio of specific anchor text links. So if every time I have a backlink it’s “SEO services” or “best SEO services” on another website [Sara: Right.] and it’s linked to Americaneagle.com, they’re saying, “OK, why do you have 27,000 backlinks that only say SEO services?” A more natural backlink would be, “Get your SEO services at Americaneagle.com,” website link.

Sara: Right. Right, right, right. Or the brand name is linked.

Harley: And so we’re actually even seeing Google slowly start to cinch down on all the SEOs who are like, “Oh, we need exact match anchor text for all of these different keyword phrases linking to our key pages.” Well, a normal person that’s not thinking about SEO would say, “Americaneagle.com link did great SEO work.”

When we’re talking about content and context with LLMs, they read the whole picture of: OK, this is an email about this thing, and we’re talking about this brand, and it’s in this context with this particular implementation of Sitefinity. So it kind of goes a step beyond—we’re seeing a shift in both spaces.

Sara: That is great. That’s great insight. Would you have a hot take right now in this space?

Harley: Maybe not a hot take, but at the beginning [Sara: OK.] of the year, I was approached by—oh my goodness—WarpStream, the major PPC company, and they were doing a kind of expert roundup of predictions.

Sara: Yeah.

Harley: And I predicted that Gemini would match or overtake ChatGPT by the end of this year. And it just came out in the last month that Gemini and AI Mode have both matched ChatGPT’s usage. And so my guess would be by the end of the year, both AI Mode and Gemini usage will surpass ChatGPT. So the message there is: don’t bet against Google.

Sara: There you go. Well, that is a hot take. That’s a good one, and I will actually follow up with you in another four months and see if your prognosis is correct. One last question: Who do you follow for inspiration or information?

Harley: I am a huge fan of Will Reynolds with Seer Interactive. They’re an SEO agency, but Will is very forward-thinking on AI and LLMs. I love the fact that Will is also straightforward. I was at a conference that he spoke at last year, and kind of his opening line was, “Don’t do crappy marketing for the sake of SEO. Do great marketing first.” I’m very much aligned with that—that’s the one that sticks out to me. There are so many great people that I follow constantly—Lily Ray and a handful of others—that are great for just sharing what’s going on and how to be thinking about this into the future.

Sara: Well, Harley, it is always a pleasure. Thank you so much for your time today.

Harley: Yeah, absolutely. Super, super fun.

Sara: Listeners, thanks for tuning in. Make sure you like and subscribe wherever you get your podcasts. Until next time, I’m Sara Faatz, and this is 10 Minute Martech.

Keep reading: 10 Minute Martech recaps.