Latest / Found in AI: AI Search Visibility, SEO, & GEO / What AI Engines Actually Want (And Why Your Blog Posts Aren't It) ft. Bryan McAnulty

Transcript

What AI Engines Actually Want (And Why Your Blog Posts Aren't It) ft. Bryan McAnulty · Found in AI: AI Search Visibility, SEO, & GEO

• 0:41
  SPEAKER_01: Hey, welcome back to Feminine AI.
• 0:43
  SPEAKER_01: I am Cassie Clark, a fractional content strategist and AI Search Visibility Consultant.
• 0:49
  SPEAKER_01: This is a show where we talk about what's actually happening in AI Search and, you know, what to do about it.
• 0:55
  SPEAKER_01: My guest today is Brian Matthanaldi, founder of Heights Platform in Latchloop.
• 0:59
  SPEAKER_01: Heights Platform has helped over 10,000 creators and entrepreneurs build online knowledge businesses, things like courses, memberships, and communities.
• 1:09
  SPEAKER_01: And Latchloop started as an internal AI coding agent and is now available to the public.
• 1:14
  SPEAKER_01: Brian spends a lot of time thinking about what AI engines are actually looking for right now, and his take is pretty specific.
• 1:20
  SPEAKER_01: I actually really agree with this take.
• 1:23
  SPEAKER_01: He says that the training data already has everything that existed before.
• 1:27
  SPEAKER_01: So the only content worth creating is what's new, what's happening now, and what real people are actually saying.
• 1:33
  SPEAKER_01: That kind of shifts a conversation from just blog posts intowards something that most brands aren't thinking about just yet.
• 1:39
  SPEAKER_01: That's where we go today.
• 1:41
  SPEAKER_01: Let's get into it.
• 1:56
  SPEAKER_00: Heights platform, we have now helped over 10,000 creators and entrepreneurs build online knowledge businesses.
• 2:02
  SPEAKER_00: It's an all-in-one platform for building an online course, membership community, uh, your website, uh, everything you need to sell some kind of digital product or info product online.
• 2:11
  SPEAKER_00: And then uh Latchloop is our kind of AI coding agent, uh general uh AI agent solution that we built uh mainly internally as a tool to help us build our software faster.
• 2:24
  SPEAKER_00: Um, but we we found it's been very useful.
• 2:26
  SPEAKER_00: And so now we're kind of uh making that available to uh the public as well.
• 2:31
  SPEAKER_00: Um but yeah, that's a quick background to me.
• 2:33
  SPEAKER_00: I started back in 2009 on a graphic design, web design agency that eventually turned into how can we build our own products, tried out a bunch of different SaaS products, e-commerce products, uh services, and then uh eventually, yeah, landed in uh where Heights platform is where we spend most of our time now.
• 2:51
  SPEAKER_00: Uh and of course, yeah, with building with AI.
• 2:54
  SPEAKER_01: So there's a couple of things that I want to talk about here, but one of the things we're gonna talk about is the kind of content that these LLMs really like.
• 3:03
  SPEAKER_01: I want to ask though, because you talked about the community aspect.
• 3:06
  SPEAKER_01: Are you finding that creators are kind of the key to holding some of this together with the GEO strategy?
• 3:14
  SPEAKER_00: Yeah, um I am uh really excited about the what like a community can do for either like to be your business and you're selling access to some kind of community membership, or to be something that can help grow your business.
• 3:28
  SPEAKER_00: If like you're you're running a business and you want to offer some kind of free community to your audience.
• 3:33
  SPEAKER_00: And it's something that I've always believed in.
• 3:35
  SPEAKER_00: And from the the Startup Heights platform, we we thought like, okay, how can we how can we help improve learning and help improve what these creators are trying to teach?
• 3:42
  SPEAKER_00: And we thought, okay, well, integrating some kind of community element in that was always something that made sense.
• 3:48
  SPEAKER_00: Um, but now with this age of AI, I think it makes even more sense because everybody wants to connect with the actual real human, you know?
• 3:56
  SPEAKER_00: And you I don't know what your experience is, but for me, if I go on these social media platforms, it's just so frustrating to see replies.
• 4:04
  SPEAKER_00: And it's like, I'm pretty sure this is AI, but even if it's not like it's so close to it and it's just annoying to see all this same stuff.
• 4:12
  SPEAKER_00: Um, and actually, like uh recently, uh I noticed on X they had this new feature that was like only people could reply who like your followers followed them or something like this.
• 4:24
  SPEAKER_00: And I read like one of the threads there, and it was so refreshing.
• 4:27
  SPEAKER_00: It's like, wow, this is I can actually tell these are real humans responding.
• 4:32
  SPEAKER_00: And um so yeah, I think everyone wants to be a part of a community of like-minded people who are trying to either share the same interests or accomplish the same kind of things.
• 4:42
  SPEAKER_00: And uh I think that uh I mean, I'll I'll wait to get into it a little bit more, but there's ways that I think this can really help your uh your AI search and like GEO and and all that as well.
• 4:53
  SPEAKER_01: So let's just dive in right there.
• 4:55
  SPEAKER_01: So, how how can this help your GEO strategy?
• 4:58
  SPEAKER_00: So if you think about it, like in our team here, we've been considering like, well, what does AI actually want compared to what like the Google algorithms and search algorithms before wanted?
• 5:10
  SPEAKER_00: And I think some of the fundamental differences are that it's really important to have the content of like what is happening now and what what is what is relating to your skills or whatever you're talking about that is specifically a new thing.
• 5:24
  SPEAKER_00: If it was in the training data before, then like just don't even consider to write about it or make content about it because it's already there.
• 5:30
  SPEAKER_00: And I think if you went back like 10 years ago, we were all thinking, okay, we have to produce all these blog posts or all these pieces of content.
• 5:38
  SPEAKER_00: And then if we can have something that other people either haven't written about before or haven't written up about as well, then we can rank for it and we'll get traffic.
• 5:47
  SPEAKER_00: Nowadays, the LLM already knows all that.
• 5:50
  SPEAKER_00: And the thing that it doesn't know is what are the ways that somebody's approaching something now in a particular skill, or what are the the things that are different now?
• 5:58
  SPEAKER_00: And so the LLM is really uh like focused on trying to find those things, and that's where it's gonna go search the web or or browse the web with deep research and things like this.
• 6:08
  SPEAKER_00: And so I would think about that in terms of your content.
• 6:11
  SPEAKER_00: Like if you're writing about like that's one of the reasons that all the AI content is doing so well, not AI written content, but content the content about how can I use AI to accomplish this, because the the training data doesn't know that already.
• 6:24
  SPEAKER_00: And uh that's one part.
• 6:26
  SPEAKER_00: The second part is where are like actual people having discussions and what do what are people actually sharing their opinions now is something that I think is really important to LLMs.
• 6:36
  SPEAKER_00: Because again, if you think about it, uh if you uh Google something, I'm sure you've seen like results where like it has Reddit threads and and people are talking about discussing something.
• 6:46
  SPEAKER_00: And often like you'll care about that as a consumer because you want to see, well, what do actual people think about a product or a brand or or whatever the thing is?
• 6:55
  SPEAKER_00: And that's also the thing that's constantly changing, but it's not in the AI training data.
• 7:00
  SPEAKER_00: And so I think businesses have a unique opportunity right now to become part of having that data and that information from their audience in a way that they can then capture the attention of these LLMs and then in turn find the people who are looking for them.
• 7:15
  SPEAKER_00: And so the core example there is inside Heights platform, we have this community feature.
• 7:20
  SPEAKER_00: And we introduced uh last year a feature called public community pages, where if part of your community is not like behind the paywall or not blocked off, then we automatically generate these landing pages for the individual posts.
• 7:33
  SPEAKER_00: And it's kind of similar to Reddit that like you can see the whole post, but then if you want to be able to see more of that, uh we don't show all of it.
• 7:41
  SPEAKER_00: You have to sign up and join that person's free community.
• 7:44
  SPEAKER_00: And so what this does is now as the owner of that community, you could look at it as before.
• 7:50
  SPEAKER_00: Well, you were helping people in your community.
• 7:52
  SPEAKER_00: It's it's better than just doing customer support because now anybody in your community gets to benefit from what you had to share.
• 7:58
  SPEAKER_00: But now it goes a step beyond that because now it might also show up in Google or ChatGPT that somebody finds that.
• 8:05
  SPEAKER_00: And then the other thing is well, before you used to take all this time to write these blogs and it had to be this polished information.
• 8:11
  SPEAKER_00: Hopefully it will catch on and somebody will look at it.
• 8:15
  SPEAKER_00: But nowadays it can actually be that your customer is the one writing the question and they're making the post, and they're probably writing it in the same way that your other customers or somebody else would try to search for.
• 8:26
  SPEAKER_00: And so then you just get to answer it, and you're not only helping that customer or that lead and everybody else in your community, but potentially attracting new leads as well.
• 8:36
  SPEAKER_01: So I can see how that would be really helpful for like the live search with the information that the AI can't come up with themselves.
• 8:42
  SPEAKER_01: So when you're thinking about training data versus live search, how do you separate the content for those two?
• 8:50
  SPEAKER_00: Yeah, I think um there's a couple different ways that that AI could come across your brand or your business in some way.
• 8:59
  SPEAKER_00: And I think that people have to remember that when you ask AI a question, sometimes it's just going through like it's pulling whatever is in the model itself, the training data.
• 9:08
  SPEAKER_00: Sometimes it's going to search the web.
• 9:10
  SPEAKER_00: And if it does that, it may get like a limited number of results and it kind of gives up there.
• 9:14
  SPEAKER_00: It's like, here's what I happen to find.
• 9:15
  SPEAKER_00: And then sometimes it does deep research and then it'll really go through a bunch of different sources and find things.
• 9:21
  SPEAKER_00: And so I think it's worth it for a lot of brands to consider how do you show up uh in each of these things as almost separate targets for your brand.
• 9:30
  SPEAKER_00: And the the training data is going to be one where it's gonna be more of like a longer term effort that your your brand just has to be maybe better known in some ways and and mentioned in more places.
• 9:40
  SPEAKER_00: Um, and then the the web search is going to depend on also like, well, where are uh where are the things that you just happen to come up earlier or things like that?
• 9:51
  SPEAKER_00: And then deep research, there's there's more chances, but uh the way that I've been approaching it at least is like to kind of work backwards and think of like what are the situations in which uh like a customer would want to come across this and making sure AI has content on our site to be able to answer that question for the customer.
• 10:11
  SPEAKER_00: And I guess more concretely, what I mean by that is what's really cool now is AI will kind of do selling for you in terms of like the customer says, this is the thing that I really care about.
• 10:22
  SPEAKER_00: Is there any product that that does this or aligns with my needs?
• 10:26
  SPEAKER_00: And the AI is going to try to find that, but it can't answer that question if your website doesn't communicate that in the right way.
• 10:32
  SPEAKER_00: And so it's now the brand's job to be able to say, okay, well, if you wanted to do XX and X, or if you were this kind of person, then showing like how your product can fit in with that kind of persona.
• 10:46
  SPEAKER_00: And then the AI can understand that and communicate that back to the reader.
• 10:51
  SPEAKER_01: So it sounds like to get that kind of content, you're really pushing like those community forums.
• 10:57
  SPEAKER_01: Am I understanding that right?
• 10:59
  SPEAKER_00: Yeah.
• 10:59
  SPEAKER_00: So I I think for yeah, communities are uh in some ways make more sense than blog posts.
• 11:06
  SPEAKER_00: Uh, I think for it depends on your brand and what you're doing.
• 11:09
  SPEAKER_00: Um, but then also I think uh it makes more sense to spend time on like uh articles that are even about like comparing your product to other products or comparing your product to like certain ways of being able to accomplish things.
• 11:23
  SPEAKER_00: And where like those pages, maybe you used to have like some of them on your website, but then it would kind of feel like boring because well, if nobody's gonna read all of them.
• 11:32
  SPEAKER_01: Um now the AI might take the time to read all of the ones that it thinks are relevant for that specific user, which is interesting because a lot of companies will put that down in the footnotes, like nobody's gonna look at it, so we'll just shove it down there.
• 11:46
  SPEAKER_01: So it's it's interesting that that's the thing that people are having to write now, um, which is fine.
• 11:53
  SPEAKER_01: But um, so we have we've covered like the community content and those pages like that.
• 11:59
  SPEAKER_01: Is there anything else that you've seen that helps influence the A AI or LLMs?
• 12:06
  SPEAKER_00: Um I think it's really important to just keep experimenting because like I wouldn't say I I know everything about this, I think nobody knows quite everything about this yet.
• 12:15
  SPEAKER_00: It's changing so fast.
• 12:16
  SPEAKER_00: And and even somebody who is maybe more experienced than me, who knows what happens next week at the at the rate things are going.
• 12:23
  SPEAKER_00: Um so yeah, I think it's uh it's important to to keep uh creating things.
• 12:29
  SPEAKER_00: Um, but uh yeah, I think maybe the the bigger piece of leverage is just like learning how you can utilize these models in order to aid with that process.
• 12:39
  SPEAKER_00: Um, and not in like generating a bunch of slap for you, but instead uh being able to have these agents that can help you do research and help you discover the things that you should be writing about or the things that you should be uh communicating better.
• 12:53
  SPEAKER_01: What what's your process to find the things that you should be writing about?
• 12:57
  SPEAKER_00: Um, well, I'm not the writer on my team specifically.
• 13:00
  SPEAKER_00: I'm focused mostly on building the software.
• 13:02
  SPEAKER_00: Um but still uh like I I find personally like a lot of uh I know this for myself.
• 13:09
  SPEAKER_00: If I'm like looking for a product now uh or I want to buy something, I don't go on Google.
• 13:13
  SPEAKER_00: I ask Deep Research and then it does like all the research for me, and then I read based off of that.
• 13:18
  SPEAKER_00: Um, and so uh yeah, I'm I'm trying to do a lot more research.
• 13:23
  SPEAKER_00: Um, and I can relate it at least to like the software development and how I'm building these products.
• 13:28
  SPEAKER_00: Uh, the models are so good now in a lot of ways that it's very weird how the interaction is currently.
• 13:35
  SPEAKER_00: Uh, I would say even less than a year ago, it was a lot about like, okay, I gotta try this multiple times.
• 13:41
  SPEAKER_00: Maybe I have to adjust the prompt, and then I need to review this because maybe the model like completely messed it up or whatever.
• 13:49
  SPEAKER_00: And uh the reviewing part uh is still a problem in a lot of ways, but it's very it's a different problem now than it was before.
• 13:57
  SPEAKER_00: And so before it might have just messed everything up.
• 14:00
  SPEAKER_00: But now the models are so good that you can kind of review it faster in a way.
• 14:05
  SPEAKER_00: And a lot of what I'm doing is I'm just saying to the model, like, I don't have time to read this.
• 14:10
  SPEAKER_00: Like, is it good?
• 14:12
  SPEAKER_00: And like, can it can I go through with this?
• 14:14
  SPEAKER_00: If not, tell me why not.
• 14:15
  SPEAKER_00: And like, I'm probably gonna send it back to AI and have it uh go and make those adjustments.
• 14:21
  SPEAKER_00: But uh it's a lot more about like you can better now like direct your intent into what you're trying to have the AI accomplish for you, um, rather than uh having to spend so much time in I guess like changing how you're communicating that.
• 14:39
  SPEAKER_01: So I'm curious because you just mentioned you use deep research more than you use traditional Google.
• 14:45
  SPEAKER_01: Like, are you still using Google at all, or is it just straight to deep research?
• 14:50
  SPEAKER_00: Um I I do use it sometimes.
• 14:52
  SPEAKER_00: Um, but yeah, it's definitely changed a lot.
• 14:56
  SPEAKER_00: Um, specifically in like day-to-day the things that that I will do.
• 14:59
  SPEAKER_00: Um, because even now, like I'll use uh my own like uh setup with I have uh latch loop, which started as a coding agent, but now like why not?
• 15:09
  SPEAKER_00: If it can connect to web search tools and other MCP tools and things that we're using, then I just asked that, like go and do the research and and figure things out.
• 15:17
  SPEAKER_00: And so um I've built like a couple of our own like research agents in that for specialized tasks.
• 15:23
  SPEAKER_00: And um yeah, I think uh I can I can see a future where everybody's got like their kind of personal agent that that knows some things about them and helps them with general things.
• 15:33
  SPEAKER_00: But then for businesses, you maybe have like this kind of research agent that really focuses on this thing and this one that focuses on some other thing.
• 15:41
  SPEAKER_01: I asked that because you know the conversation lately has been users are using AI to find something and then they go to Google once they find the thing and then they dive deeper on Google, but I haven't been doing that.
• 15:54
  SPEAKER_01: I just skip Google entirely and I just stay in whatever agent that I'm using.
• 15:59
  SPEAKER_01: So I always like to think about how other people are using it just to see if maybe I'm normal or not so much normal.
• 16:05
  SPEAKER_01: So that's that's helpful.
• 16:07
  SPEAKER_00: I I I think uh I think for me it's it it it would be kind of similar in uh in most cases, yeah.
• 16:13
  SPEAKER_00: Um whereas before I would have like really researched in Google and not even trusted the first couple things.
• 16:19
  SPEAKER_00: But if I'm really trying to learn, like really dig in there.
• 16:22
  SPEAKER_00: And yeah, now I feel that I can generally trust that I've got something good from the base of Chat GPT or or whatever, and then kind of go from there.
• 16:33
  SPEAKER_00: But usually it's not more digging through Google.
• 16:35
  SPEAKER_00: Maybe it's then validating whatever it found by like checking that company's website.
• 16:40
  SPEAKER_00: Um, but yeah, it's definitely different.
• 16:43
  SPEAKER_01: Which is interesting too, because just last week, the local news they do a poll every every so often, and it's just random questions that they have on Facebook.
• 16:53
  SPEAKER_01: But the question was, how much do you trust AI, like the AI overviews?
• 16:57
  SPEAKER_01: And here in the local community, it was do not trust 90%, which marketers and people that are working in this all the time.
• 17:06
  SPEAKER_01: I feel like we are kind of in a bubble over here using it so much.
• 17:10
  SPEAKER_01: And I think that's kind of important to really call out too that not everybody is on this just yet.
• 17:16
  SPEAKER_01: Like discovery is still happening in Google for some people.
• 17:20
  SPEAKER_00: Yeah.
• 17:20
  SPEAKER_00: Well, what I think is really interesting is like I do think there's a skill involved in knowing how and when you can trust it.
• 17:28
  SPEAKER_00: Because for the average person who's just used it for the first time and is does not yet understand how AI works, they probably shouldn't trust it.
• 17:35
  SPEAKER_00: And it's probably going to be more likely to do things where it's hallucinating to them.
• 17:40
  SPEAKER_00: But um, you can kind of build build the skill of realizing when you say something that it might be more likely to either fabricate something or or not.
• 17:49
  SPEAKER_00: Um and and yeah, so so being able to kind of build up the the internal model in your head of like the chance that it's gonna make something up for a given prompt is uh is something you have to get familiar with.
• 18:03
  SPEAKER_01: So do you think that brands need to also consider that too?
• 18:06
  SPEAKER_01: Like the user on the other end of this, when they find us, might not trust what AI is saying about us.
• 18:12
  SPEAKER_00: I do think there's there's this weird place that we're at right now of like where should the brand um, like how should the brand integrate their products with these AI tools?
• 18:22
  SPEAKER_00: And like it's something that we consider because we have uh Heights AI and Heights platform, and we had that since early 2023 as like one of the first kind of like co-pilot chat products that you could ask questions about our software, you could ask it even to like edit settings and things inside our software.
• 18:39
  SPEAKER_00: And uh it it works well and we're happy with it.
• 18:42
  SPEAKER_00: But the answers people are gonna get from that is gonna be very different than if people ask ChatGPT.
• 18:47
  SPEAKER_00: And so we've had cases where we'll get a support question and it's because somebody says, Well, ChatGPT said I could do this.
• 18:56
  SPEAKER_00: And it was like, well, ChatGPT doesn't exactly know the answer.
• 18:59
  SPEAKER_00: You could have asked RAI inside your account, it would have.
• 19:03
  SPEAKER_00: But um, but yeah, it it also becomes a thing.
• 19:06
  SPEAKER_00: Well, do we want to make a tool, a connector or an app or something that AI tools could access, um, or do we not?
• 19:15
  SPEAKER_00: Because on one hand, then we could, as a brand, maybe guarantee somewhat better support for users who want to use ChatGPT, which I would think is a good thing.
• 19:25
  SPEAKER_00: But then on the other hand, there are restrictions and limitations of what like Cloud or ChatGPT actually has access to that it can perform uh otherwise.
• 19:34
  SPEAKER_00: And so like we're trying to figure out like how do we not confuse our users or like segment our service weird by like, oh well, if you're trying to access your account through ChatGPT, then it can do this, this, but not this, and vice versa.
• 19:49
  SPEAKER_01: Yeah, which which brings me back to the whole customer education is the huge part of a geo strategy.
• 19:57
  SPEAKER_01: Like you're not just teaching, well, I mean, you are you're teaching the person audience about your brand.
• 20:03
  SPEAKER_01: You're also teaching AI about your brand.
• 20:05
  SPEAKER_01: Like there's so much teaching involved right now, it's not funny.
• 20:09
  SPEAKER_00: Yeah, and and it's definitely it's changing fast.
• 20:12
  SPEAKER_00: I think the probably your audience here, like watching and listening to this, is is well ahead of the the general users.
• 20:20
  SPEAKER_00: But yeah, we saw like that that we felt almost too early with some of our AI shot features in 2023 where people were using it.
• 20:27
  SPEAKER_00: We told them this is this is AI, it's not like an old chat bot or something, but people were treating it like a Google search still back then.
• 20:34
  SPEAKER_00: And now uh we're noticing that people are messaging it a lot more and and learning they've learned now how to interact with AI and what it can do uh at a better better level.
• 20:46
  SPEAKER_00: But it's still such a weird thing because there are like it's confusing when you all you have is the chat interface to know like what are the real capabilities behind it.
• 20:56
  SPEAKER_00: And that's where you see like the stories of uh I think my brother was telling me the other day, somebody made the this like joke content about Chat GPT.
• 21:02
  SPEAKER_00: They say, like, time me uh running a mile, and then they say back to it, okay, I'm done.
• 21:07
  SPEAKER_00: What was my time?
• 21:08
  SPEAKER_00: And it just makes up a time.
• 21:09
  SPEAKER_00: And it's like, I didn't even run.
• 21:12
  SPEAKER_00: Um and so yeah, so it can be confusing.
• 21:16
  SPEAKER_00: And I found that like looking at some of these uh other agent products as a consumer of them, you feel like, wow, this has a lot of potential, but at the same time, you feel like, well, what do I do from here?
• 21:29
  SPEAKER_00: Like because it's so open-ended, it's almost like you don't know what you can actually do to kind of implement it.
• 21:36
  SPEAKER_01: Okay, we covered quite a bit in that conversation, and I'm gonna stop it right there.
• 21:39
  SPEAKER_01: And I'm gonna leave you with one final reminder.
• 21:42
  SPEAKER_01: We are still very early to AI search optimization and AI use in general.
• 21:47
  SPEAKER_01: The general population is still learning how to use AI tools as a research tool.
• 21:52
  SPEAKER_01: And not every brain is optimizing for visibility just yet.
• 21:56
  SPEAKER_01: That's actually an opportunity if you move deliberately.
• 21:59
  SPEAKER_01: And because I like to leave you with a tip that you can actually do this week, here's one that I need to think about doing this week.
• 22:07
  SPEAKER_01: And honestly, the first week of July, since it's usually quiet for nearly everyone, since everyone's away on summer break, is a great time to do this.
• 22:14
  SPEAKER_01: Look at the last three pieces of content that you published and then go count how many places that each one lived.
• 22:20
  SPEAKER_01: Is it on your website only?
• 22:22
  SPEAKER_01: Did you put it on your social channels?
• 22:24
  SPEAKER_01: Is it in your newsletter?
• 22:25
  SPEAKER_01: Did you put it on YouTube?
• 22:27
  SPEAKER_01: Make a note of all those places.
• 22:29
  SPEAKER_01: If the answer is just your website, that's the thing that we need to fix.
• 22:33
  SPEAKER_01: AI engines build entity authority from seeing your brand and your expertise show up across multiple services, not just from that single blog post on your website.
• 22:43
  SPEAKER_01: Use these quiet weeks to redistribute what you already have before you go create anything new.
• 22:48
  SPEAKER_01: If you want a framework for thinking through this more systematically, head over to CassieClark Marketing.com and if you need an audit, you also know where to find me.
• 22:56
  SPEAKER_01: Thanks for listening, and I'll see you in the next one until then today listening.