From Surveillance Marketing to Truly Knowing the Customer: The "Near Me" to "Know Me" AI Evolution
For years, marketers have chased the promise of personalized experiences, but why did so many approaches fall short, and what makes this wave of AI different?
Today's AI systems aren't guessing. They remember customer preferences, understand context, and build a picture of the individual based on what users actually tell them.
The episode breaks down:
Why traditional personalization failed and why AI finally changes the landscape
How AI memory and context actually work
Why every user now see different results
Why location used to be the strongest signal, and how AI is shifting personalization toward deeper understanding of the individual.
What brands must do to stay visible
If you're a marketing leader navigating the shift from guesswork to AI-powered relevance, this episode will help you rethink how personalization works and how your brand shows up in experiences shaped by memory, context, and user intent.
Episode Links
Transcript
Rebecca Colwell (00:03) Hi, Christian, welcome back.
Christian J Ward (00:06) Hello, good to be back.
Rebecca Colwell (00:07) I'm really excited, really excited for our conversation today. ⁓ Marketers have been chasing personalization for 20 years and they've built these giant tech stacks, lots of third party data, identity across devices. And we still end up with this bad targeting and wasted spend. Why is personalization so hard to get right in digital marketing?
Christian J Ward (00:31) Well, first off, happy holidays. And secondly, I would tell you, look, I think this is a problem that we've of grown into over the years, which is there's a lot of data vendors out there. There's a lot of tools that people bought, CDPs, tracking, surveillance. All of that was built around this premise that your behavior online kind of dictated and told us who you were. And that you can feel this a lot. Like every time, like during the pandemic, I think people really felt that where they'd be discussing stuff, like you and I'd be talking via Zoom. about a topic and then suddenly we're seeing ads for the things you told me to look at. Very creepy. Now, the reality is, is most of that has been born on my behavior and tracking and surveillance. I think what you're seeing is, is that personalization is desirable both by the consumer and businesses, but the method by which we went after it was all wrong, which was taking all of these signals and deciding what to show you. What we're seeing now is AI is probably the real opportunity in personalization that we all always thought it should be, but that's gonna come with a very different set of rules and tools than what people had originally built for.
Rebecca Colwell (01:40) It's interesting. ⁓ Can you talk a little bit more about the surveillance state and what marketers have been trying to do through digital marketing and how that's not really happening in the LLMs?
Christian J Ward (01:53) Yes. So originally when we think about the browser world, the browser is what allowed us to track what you looked at and where you went. You could clear your cookies. I think only maybe 3 % of humans actually even know how to do that on the browser. But the reality is, when we look at those technologies, that was this idea. We had this premise that that was allowable. Now I will warn you, like, I think if I followed you around all day and where you went and who you spoke to and kept tabs on it, In the real world, that's called stocking. So we should have immediately known from the beginning, this is not okay. This is in fact terrible. I would say I understand why they got there, but yes, we've always been very anti this sort of tool. And it's mostly because we think that the answers, the knowledge should drive people towards the right business at the right time. And what's happening in the LLMs is I'm having a conversation with an AI assistant that has all the knowledge about a business, but it also has knowledge about me. And so our marketing team had written up this brilliant piece earlier around, instead of near me, it's know me now. And I think that's just a brilliant concept of this idea of how do you piece together, who am I, what matters to me? But that's a very far cry from what marketers were used to. Marketers were used to, we have all the data. Now the large language model has the data and it's using it to create different types of memories. tied to where you live, what you're interested in, what your job is, how many kids you have, what their allergies are is the one I always like to use because of my daughter. It's one of these things where that knowledge set then interacts in a way that marketers have never really experienced before.
Rebecca Colwell (03:33) You know, it's interesting because when in a surveillance state, there are a lot of inferences that happen, right? So I'm shopping for school supplies. It might label me as a parent, ⁓ but it might just be that I'm looking for some supplies for the office ⁓ or I'm a homeowner because my zip code, you know, isn't a high home ownership rate. So that's, you know, ⁓ why I think it gets it wrong so many times. And, you know, as an LLM, it actually knows the things that I've told it because I've said, you know, I have kids that are this age ⁓ or whatnot. So. it seems interesting. Do you think that, ⁓ this is part of why LLMs are so much better at personalization? Like we're actually telling it about us, about ourselves?
Christian J Ward (04:12) Yeah, it's a great question because really there's two forms of knowledge here that we're talking about. There's sort of the explicit and sort of the implicit. And explicit, there's two forms with the LLMs. The explicit is saying, hey, ⁓ I want you to remember that this is where I live, or I want you to remember that this is the type of car I drive to help me with cart maintenance. I want you to remember ⁓ X, Y, and Z. Many of the things you're talking about are implicit. So many brands other than purchase history or engagement or subscriptions, those are explicit memories that you're building with a brand that they could store. That would be in a CRM or something like that. Implicit is where a lot of this surveillance stuff was, where it sort of implied because you visited this site, you must have kids that need school supplies. So you're implying that. ⁓ A lot of times the implications get wrong. And that's one of the issues. With LLMs, what's happening is I'm explicitly stating things and they can actually also then build smarter implied memories. So if I say something to an LLM just in passing, like, hey, I'm looking for a lunch spot before I pick up ⁓ my son at La Crosse. I wasn't trying to tell it to remember that I have a son, right? I was trying to look for lunch. So that was the explicit thing that I stated, but it gives an implied a dataset as well that is far more accurate than trying to sort of pull along where I am based on the browser. So I think what you're seeing is, is LLMs have a huge advantage in the dialogue that we are having with it is such, there's so much rich data that is literally first person. So instead of third party tracking, which is what we classically call it, this is first party knowledge sharing to create much more interesting, explicit and implicit memories that the AI can use to personalize the experience.
Rebecca Colwell (06:09) That context is so important too, because we often are so much more descriptive when we have conversations with the LLM about why we're looking for something. ⁓ I'm curious, you mentioned remember this about me, remember this about me. How does memory in an LLM actually work? And do I have to ask it to remember things about me?
Christian J Ward (06:27) Yeah, so there's a couple things. It's a really important question that I can tell you directionally where people are today, but almost only chat GPT. So Gemini, we absolutely know this has memory. We know Claude has memory. We know perplexity has memory. They do not explicitly show it to you in an action point. Claude is starting to now inside of projects. You can see some of it, but what you're going to see, I think, is they're going to reveal more of this. So rather than guessing, there are ways to force it through an explicit statement to say, I want you to remember that my oldest has a tree nut allergy. Now it knows that. And you'll see a little flag in ChatGPT, which you don't see in the other models. I hope they adopt this that says, hey, I updated my memory. I'm going to remember that. So you know it took. But the other side is there's this level of memory around every dialogue you've ever had with it. So you could also... ⁓ prompt it to say, you remember when I talked to you about getting a new dog for the holiday, like for the family six months ago? Can we go back and revisit that? That's another type of memory. That's actually wrote like the literal recall of everything you've ever said to it. Look, I think memory is gonna evolve a lot. In fact, Sam Altman said on stage probably maybe six months ago, right after ChatChip D5 came out, he said, look, ChatChip D6 is all about memory. So this is where we're gonna see this leap. I think what a lot of people have seen in the foundational model is getting better and better and better. That's just normal improvement in the foundation of how the model works. But once you add memory and you really get it working to where people both understand what it's doing, are comfortable with what it's doing, realize the outcomes are better with what it's doing, you're gonna see this become just normal operating for everyone when they're searching. which is it's gonna be a little bit of mixture of how it uses memory, how I tell it what to remember and my comfort level. I'm gonna get much better results in my searches.
Rebecca Colwell (08:24) Interesting. So will you and I see different results for the same search?
Christian J Ward (08:29) Yes, today we could say probably we already do to some extent, ⁓ obviously location being the most important thing that drives a lot of those results. But yes, there was always sort of this rank brain idea. But it's real important to remember with search, search is really around giving you showing you a ton of results and then ranking. So the ranking was always determined. That's always a deterministic by Google. They could determine, they could change the ranks, all that sort of thing. So the idea of you and I seeing the same ranked things is highly unlikely in LLMs and AI. We might see ⁓ over a period of time, probabilistically, we might see similar things, but rank is going to be, to us, ⁓ a completely personal experience. So we all talk about search fragmentation, where people are searching in these places in different areas. We haven't really dove into ranking fragmentation, which is absolutely going to be driven by AI. And so that's why I think we call it visibility. I know a lot of other people call it visibility. I think we have to start looking at what is your relative visibility across both Rebecca's view and my view over time in certain geographies on certain topics. And that's going to be much more helpful to brands to focus on than running a query and saying, we were mentioned first. I'm like, that's not really how this works. That might work there. But remember, the AI already knows you work for that brand and they might want to just make you happy. We know the AI is a little sycophantic. So we know. there's elements here in its memory that maybe aren't giving you the right signals. You need a much broader understanding of how those rankings operate.
Rebecca Colwell (10:04) The implications for marketing analytics are huge. I mean, it's so disruptive to how we measure success and even report up on what's working. ⁓ It's really interesting. Earlier, you mentioned this concept of near me and personalization really came down to like the number one thing was location. It's simple, it's reliable. Near me was really important. You mentioned this shift towards know me. ⁓ What implication do you think that has for marketers?
Christian J Ward (10:08) Yeah. First off, I think it's going to be a little bit of an uncomfortable time because marketers, I know Mike Walrath talks about this a lot, marketers tend to want to control the message. And really what's happening now is you're providing the knowledge to the AI and it controls the message. It's going to rewrite your information. And one of the most important things to think about today is how can you control it? What level can you control that? And so when we say near me, I remember there used to be ⁓ an ER in Dallas that changed its name to ER near me. And then there was like Ventus near me. They literally changed the signage on their building because they had bought so far and says, I would advise against that with the know me current that's going to happen next. But I can tell you, I think it's a lot more about using the word knowledge. No, to know not only thyself, which is the AI does for us, but it's also going to be knowing the business. And so making sure as many facts
Rebecca Colwell (11:13) you
Christian J Ward (11:28) are woven into the most cited places with AI is critical. And so we had done that study a month ago on 9.6 million citations. What we found is the number one cited thing for both branded and unbranded queries by three different AI models globally was the first party website. So your pages should have as much knowledge on them as possible in sort of that localized format. ⁓ But also that same data should be in the second most cited platforms, which are the third party data sources we classically call listings. And so making sure the two data sets in those areas, because they capture, I believe we said roughly 83 to 84 % of all citations on close to 7 million citations across the world, those were all in those first two categories. So for marketers, you have to get comfortable with this idea that if the AI knows my insurance, it's going to comply. completely ignore every other dentist, doctor, ER, or child surgery center if that insurance is not listed on that page to let me know. It's gonna unrank people that don't have that. And that's the part you really wanna focus on because the AI is eventually going to know the most important thing in this decision hierarchy is what insurance you take. It's totally different for other topics. So depending on what business you're in, you've gotta really build out this strategy of providing more knowledge to the AI if you want to take advantage of this.
Rebecca Colwell (13:00) It's interesting because as marketers, we've had this concept of a funnel where you start really wide and then you narrow and narrow and narrow. And that concept is being totally collapsed, right? As if the LLM is already eliminating options that they know are not a FET, my top of funnel has suddenly become very small. And maybe it looks more like a straw or pipe or something, right? Where it's not getting more narrow. Any thoughts on what ⁓ conversion rates might look like in a model like this?
Christian J Ward (13:29) Yeah, it's a great question because we're going in this world, like you said, where the funnel was this and this was like my advertising, my marketing, and it was consideration and then all the way to transaction. It's going to feel like this is now the top of the funnel because this is all being taken over by AI. But the reality is the actual number of checkouts, the actual number of people through is going to be equal. So it's not necessarily a problem. The problem is how you measure the benefit of this. And that's very difficult for most marketers to understand. But I think this will feel uncomfortable, but the actual throughput will probably be even better quality leads once they start adopting this. And you've seen that today, I saw there was an article earlier on ⁓ how many billions of dollars went through with AI. But that's, we're gonna see, I don't think it's so much about the number of people buying through Chach-IPT, it's about the behavioral change. of going through chat GPT or Gemini. And so that's the thing marketers should be focused on is the behavior is changing at such an incredible rate. The dollars will follow, but the behavior is already changing. We can see that in early sort of returns from Black Friday.
Rebecca Colwell (14:43) Right, yes, I did a lot of my holiday shopping ⁓ and research using AI this weekend, and it was so helpful. And yeah, I did convert much, much faster. I was like, OK, I'm confident it's done the research for me. It feels great. Well, maybe next time, let's dive deeper into some of the analytics components and the other things that ⁓ are changing. That's a wrap this week for the Visibility Brief. Thank you so much, Christian, and I hope you have a wonderful holiday.
Christian J Ward (15:09) You as well, Rebecca. Thank you.
Rebecca Colwell (15:11) Thanks.
