Navigate the probabilistic world of AI search

Garrett says: “Treat AI search as probabilistic, not predictable.

Build content strategies rooted in deep audience understanding, because what shows up in search results depends on who's searching and how they search.”

Does that mean that there’s only a probability of whether or not your content is likely to surface in front of someone?

“Bingo. We're getting used to this idea of LLMs and AI search impacting traditional search, so our whole mindset around rankings and organic links is changing – and it's going to accelerate over the next 6-12 months.

When we're looking at AI overviews, AI Mode, ChatGPT, and what they put out, it's not the same answer every time. There's a probability of giving yourself the best chance to show up – whether it's for the targeted query or within the ‘query fan-out’. Query fan-out is the idea that, when AI is doing a search, it's actually looking at 20, 30, or 50 related searches and pulling relevant info from there. It essentially predicts what people are likely to continue to talk about in the conversation.

What is the probability that the content you produce is pulled in, whether it's for that specific keyword or related keywords, and it helps your brand become more visible to your audience?

Ultimately, the goal of AI search is to increase the satisfaction rating and give you the answer that you’re looking for, based on your search intent. Query fan-out is trying to accomplish this by understanding what other people have searched in the past, understanding the relevance of a topic and the sub-related topics, and thinking about your entire conversation journey.

Then, it can identify whatever you could be looking for with that search and answer it in one go.”

Do you still do keyword research prior to deciding what content you're going to focus on?

“There is still going to be a transition to this idea of topicality, which is ultimately where we're going. That is more of a grey area, and it’s a little messier.

We're so used to thinking about keyword search volume: ‘I'm going to target this keyword because 100,000 people are searching for it.’ What we have to start thinking about is that the way people are searching is going to change.

As we get more conversational in our searches, there's less friction, and we can trust that the output is going to be specific to what we're looking for. We’re moving away from the ‘keywordese’ that we’ve been trained to use by the capabilities of a search engine (those one-word or two-word queries) to get something close to what we're looking for.

That keyword volume perspective doesn't have the same value because you're never going to see that one-word search anymore. People are going to use natural language. Over time, we're going to see the shift from keywords to general topics and natural language searches.”

Is creating a piece of content targeting a keyword phrase and writing extensively about a topic becoming a thing of the past?

“To some extent, yeah. When you think about information that anybody can supply, it's commodified.

We're so used to using copycat content to be the one chosen by Google or whatever search engine. The value of being the one to provide that information, if it's commodified, has decreased. The value you can now provide is through proprietary information, data, or insights – or when someone is specifically looking for your brand.

Content strategy is going to shift. There will still be times where it might be appropriate to try and target a specific keyword, but that’s going to become rarer as search behaviour changes.”

How does an SEO or Head of Marketing measure SEO success moving forward, if it's not possible to target something specific with known keyword volume?

“In a lot of ways, it is going to become trickier, similar to brand performance metrics. We're so used to being able to put data into it, but now you don't have the data, and everything's messy.

We've been having this conversation about attribution for a while. It's not as precise as we want it to be, but it's still somewhat accurate. You're going to want to look from the bottom up. Look at those key business-driving metrics and view SEO, AI search, GEO, or whatever you want to call it as a support for all of your marketing, holistically.

You will still see traffic driven to your website. It's not going to be the same volume, but what we're starting to see and hear about as people experiment is that, when someone does search and goes to your website, they're usually much further down the funnel. They have a stronger intent to find your services, and that traffic is more likely to convert, become revenue, and be more valuable.”

Is it possible for exactly the same question or keyword to result in an entirely different piece of content being pulled from your website?

“Definitely. Think about the way that personas work. ChatGPT has introduced a memory layer, where it takes into consideration all of your past interactions with the platform, based on your account, and it gives you answers based on that experience.

For instance, if I put in my custom instructions that I am the Director of Marketing at iPullRank, and then ask, ‘What is the best SEO agency?’, it's much more likely to tell me to give iPullRank as the answer. Now, don't get me wrong, it is the best SEO agency, but it knows my relationship there.

Back in May, at Google I/O, Google announced that they're taking more personal context into consideration. Whether it's your search history or your Gmail usage, any data that Google has about you can inform the output it gives, personalised to you.

We've had personalisation in SEO in the past. Just think about localisation results, where you get a different result based on where you live. Now multiply that by 100, where Google or ChatGPT is taking all the information about you to give you a much more personalised experience.

We'll see this mostly impacting things like e-commerce, but we're starting to see the impact already. It will get much more significant over the next 6-12 months.”

How do you research who your target audience is and map the intended content towards them?

“It's the old school ways, and there are a bunch of ways to do it. You don't want to find yourself in a sample bias where you're only looking at research from a specific segment that doesn't actually reflect who your existing customers, past customers, and potential future customers are, and what the entire market looks like.

That could be survey data, direct interviews with people, or looking at search behaviour through tools like SparkToro, which look at the places that your demographic tends to spend their time. Take all of these different types of market and audience research and calibrate it to create relevant content, infused with your brand messaging and positioning, so you have the best chance to show up across different personas.

As an industry, people joke about personas, but there's real value there. You're never going to get the individual level because that doesn't scale, but if you can generalise across a segment of shared attributes for different users and get to the heart of what's important to them, you have a much better chance of creating content that's not only going to resonate when they do go there, but it's going to show up when they search for it.”

How do you decide where your audience is likely to be interacting?

“It’s a classic SEO answer, but it depends.

We have a finite amount of resources in terms of where we spend our money and focus our efforts. The current and future state of marketing is an omnimedia strategy, creating different types of content on different types of channels.

Bigger enterprise brands are going to have a bigger budget that they can spend across all the different personas that they're trying to attract. If you are a small niche brand, you really have to know who your niche audience is and the content that appeals to them, with a very targeted sense of market research.

It depends, but a good marketer is going to do the research and understand who they can afford to target and how to prioritise their resources to create the most valuable content that's most likely to resonate with their audience in the places where they spend their time.

When it comes to finding out where they are, you need to use the tools or the people at your disposal to find out that information. You just have to do the work.”

Is there a best way of looking at the data and deciding which platform to test initially?

“You can get a sense by looking at trends, analysing, and trying to identify those places with things like survey questions.

If everyone says that they're spending time on LinkedIn, then you go to LinkedIn. If everyone's searching for your products in your industry using a certain type of language, you're going to use that language when you're describing your product.

You can use a tool like SparkToro to do audience research and find out whether they listen to a certain podcast or spend time on a certain social media channel.

There's a digital PR element there. Beyond just going to your own owned media, you may want to be present on Reddit or be in trade publications where people are looking at listicles. As much as we want to deny it, listicles still have value in 2026.

It's a moving target. At iPullRank, we recently conducted a UX AI Mode study, looking at how people are searching (despite the fact that AI Mode isn't the default version of search). There's a lot of friction to get there, but one thing that we found is that there's no search channel that is definitively the right search channel for every type of search.

For news searches, people want to go directly to the publications that they already trust. For e-commerce searches, most of the time, they go to Amazon because they want to see visuals. If AI Mode isn't giving them a high-quality carousel of products, they prefer not to go there.

Over time, we're going to continue to see this fracturing of potential search preferences, depending on demographics, search type, and modality – meaning video, text, YouTube, etc. All of this is a moving target.

Marketing's never been more fun, but it's never been more difficult.”

What is relevance engineering, and how can you use it to align with how AI surfaces content?

“It’s the framework or discipline of how we think of modern SEO at iPullRank. Our founder, Mike King, concocted this perspective.

We’re realising more and more that all marketing is holistic, especially as SEO evolves and AI search becomes the default way that we search for things. Relevance engineering is the integration of five different disciplines: information retrieval, content strategy, user experience, digital PR, and AI.

It’s about how those five different pillars come together to give you the best chance of being visible, based on the probability that you appear in all of these different channels, in different omnimedia modalities, that your audience is using to search for you.”

Is relevance engineering suitable for every type of industry?

“I definitely think so. Regardless of where you are predominantly getting your audience from – whether it's organic search or other channels – everything should support everything else.

Ultimately, people search. Search isn't going away. Making sure that you are visible on whatever search platform, in whatever means necessary for your brand and your audience, is the foundational layer of what relevance engineering aims to achieve.”

Garrett, what's the key takeaway from the tip you shared today?

“Search is becoming probabilistic. The way we thought about search in the past is changing.

It's not just about rankings anymore; it's about creating the content that gives your brand the best chance to show up.

It's like buying lots of raffle tickets for a lottery that's changing every single time someone searches. You need to understand your audience, do your research, and make sure that you're everywhere that your audience is to be found.”

Garrett Sussman is a Director of Marketing at iPullRank. Find out more over at iPullRank.com.