0:00
/
Transcript

What Is Generative Engine Optimization (GEO)?

How Brands Get Recommended by ChatGPT, Claude and Gemini

SEO is dead. Generative Engine Optimization is the new marketing and advertising game in town. Today we get to the core of what it is, why it matters and how to use it effectively.

LISTEN ON SPOTIFY
LISTEN ON APPLE PODCASTS


TRANSCRIPT

Mark Fielding

disruptors and curious minds. Welcome to the show. And we are back speaking about artificial intelligence today. And many of you, most of you, I’m assuming all of you have heard of search engine optimization, SEO. Well, I’ve got a new one for you. GEO, generative engine optimization. And today’s guest Awad Sayeed is the CTO of Parsnip, a new artificial intelligence platform focused entirely on GEO. Awad, welcome to the show. Thank you for thinking on paper with us today.

Awad Sayeed

Yeah, thanks for having me, Mark and Jeremy. It’s a pleasure to be here.

Mark Fielding

The pleasure is all ours. And we are going to start with the question you probably get asked every single day. What is generative engine optimization?

Awad Sayeed

Well, I’ll tell you what, it’s not a great acronym. are still debating GEO versus AEO, but ultimately, it is the task of helping a brand or any topic really show up better in the results from a answer engine or a generative engine. So people might know that as the chat GPT interface or Claude Gemini. I’m sure everyone’s familiar with that.

And if we look back at the history of the web until now, SEO has dominated. That is getting you ranked in Google. Now, Google, historically, you get a list of a couple of websites, and that’s the name of the game. You put in a keyword. When you are engaging with LLMs, you’re having more of a conversation. There’s a lot more context. So the of GEO is to help your brand be

associated with these various contexts. Real world example, if I’m looking for shoes, historically I might say best shoes into Google or something like that. But with ⁓ LLM, I have either historically or I’m presently having a full on discussion. I am in my mid thirties, I’ve got some bad knees. What are the things I should be looking for? And then and only then with my full intent already captured, do we then have a conversation?

a conversation if you will, we might actually have a branded recommendation. Now that’s coming from the commercial side, that is the nature of my work, but that could well be true for any topic. That’s kind of the heart of GEO if you will.

Mark Fielding

Jeremy, I think you might agree with me. Let’s kick this around a bit, but it’s quite related to SEO and that’s GEO. We speak about a lot of things on this show and one of our pillars in our human index is the space industry. And we speak to a lot of people putting satellites up into LEO and GEO and GEO is crowded already. We don’t need another GEO. So I think AEO is the way to go with this.

Jeremy Gilbertson

Well, the history on it too, as I understand it a wide, came out of a research kind of initiative like Georgia Tech and a few others that originally started the GEO thing. What is the A in the AEO? Is that agent?

Awad Sayeed

Answer Engine

Jeremy Gilbertson

Answer engine, answer engine. Isn’t that what we’re all looking for is answer engines.

Awad Sayeed

Right.

I think we’re all seeking that in life, absolutely.

Jeremy Gilbertson

But don’t forget about your own answer engine, friends and neighbors. We’ve got one up top. Got one up top. So let’s, real quick, expanding on this a little bit, and I did a little bit of research on parsnip and kind of what you’re doing. can you explain what a persona-based agent is and maybe how you guys...

Awad Sayeed

Absolutely, and hopefully that’s our discussion topic.

Mark Fielding

Yeah, it is.

Jeremy Gilbertson

How you guys use that kind of infrastructure in what you’re doing at Parsnip? Do you use something called that?

Awad Sayeed

Yeah, absolutely. And that comes back down to my original explanation when it comes to intent. So marketers, they don’t just market out into the world. They do targeting, right? Targeting based off of different personas. So in the old days of SEO, again, like you would put in a keyword and get back a list from Google and that was it. Of course, over time, Google’s evolved with personalization and things of that nature. But broadly speaking,

It was relatively deterministic, relatively. You put in a term and you get that out. Now, with LLMs, the name of the game is context. Everything is based off of context. How they retrieve information is based off of context, and how they even structure the retrieval process, if you will, is based off of context. And so our technology philosophy is, well,

marketers think in these sorts of terms, if we are going to show them how their brand is being represented in these answer engines, you have to give them the full representation. So you have to go out to the LLMs with the personas of their consumers to be able to actually definitively say for your audience, this is how you actually turn up. So that’s the heart of how we’re approaching it. But I think the higher order bit there is remembering that

While there are many similarities between SEO and GEO or AEO as we like, the missing piece is the context. And a lot of people aren’t really focusing on that as much. We’ve seen companies come out and I call it SEO++. They just slapped a new label on the old and are just doing analysis based off that. And we just think it’s diff. That’s not the right way to do it.

Mark Fielding

There’s so many tracks to go down. I’m to start with. I have done some SEO writing in the of my existing SEO works in the world of GEO?

Awad Sayeed

I would say actually pretty much all of it. There’s not really any SEO best practices that will penalize you per se in GEO. I think a lot of people aren’t necessarily talking about that. This is an extension of existing behavior and fundamentals. So that’s the good news. Your work is not for naught. So that’s all good. You have a solid base. And then where we go from there... is like I said, Context is the name of the game. instead of optimizing purely for backlinks, you want to also look at making it easier for an LLM to be able to do the retrieval it needs. if we’re talking, let’s say in commerce, having comparison cards, this my product versus another product, right? If you all just anecdotally, you see a lot of, it’s not this, it’s that when you talk to chat GPT or similar types of behaviors. So making it easier to fit into the shape of the responses that LLMs are is a GEO specific technique you might look at.

Mark Fielding

separate page or existing page? Like how does the context work on page rank?

Awad Sayeed

Existing page ideally is yeah, ideally existing page. Yeah

Mark Fielding

Because with SEO, isn’t that so you don’t really want to mix the search terms to a certain degree. And so you’d have one page, which would be, and if we’re talking about, so we talk about a lot about fusion on this show. So if we had Helion and they were redesigning their website, they might have one page for reactor, one page for Helium-3, one page for staff, one page for current projects.

Mark Fielding

Would that stay like that? What would they do? How would it change?

Awad Sayeed

Absolutely. So a few different things. One is you might enrich some of the content with some of the examples I’ve noted. then second, there you go, there you go. And then separately, part of it is also schema based. So you might have the same page and then something that’s not necessarily visible to the end user within their browser, but you might enrich that page through some schema. This is more technical. I guess we have a fairly technical audience.

Jeremy Gilbertson

Enriched, no pun intended, right? Yeah.

Mark Fielding

Ha ha ha!

Awad Sayeed

You might do something like JSON-LD, which is a standard that a lot of companies use to make the content more machine digestible, let’s say. Which again, that’s an established technique already, but this takes it to the next level. This will certainly help in the context pathfinding.

Jeremy Gilbertson

So on the schema thing, just for clarity on that is instead of relying on the bots that are actually crawling and searching and pulling information, this is like a direct channel to talk to the robots necessarily, just say, hey, this is what I want you to know, robots. So that’s, yeah.

Awad Sayeed

Exactly. Exactly, exactly. And you know, it’s not a guarantee, but it definitely helps again with the context.

Mark Fielding

like the spiders. So when you publish a blog post today, send Google comes at some point, scans the page, puts into its mind and then dishes it up when and if someone comes across it. There’s only one Google. There are many LLMs. How does it work? How does that work?

Awad Sayeed

Yes. Yeah, and you know, that’s one of the both exciting parts about being in this space. And also, it’s still an unknown, frankly, in terms of who all will be the winners. I don’t think there’ll be a single winner. It’ll probably be more of a split sort of situation. But everyone has slightly different techniques. To your point, everyone knew Google was the name of the game and you had to optimize for Google. And there was basically one single playbook.

more or less. Now, if you look at the top players in the space, they all work under the hood in very similar ways, of course, but the primary data can vary greatly. ChatGPT, for example, indexes quite a bit on social media type of content. Reddit, very popular for them as a data source, whereas Anthropic, much more

academic focus a bit more, let’s say rigorous in their sourcing and their data cleanup and so on and so forth. I mean, that’s partly what we’re doing is we’re going out to all the players and saying, well, here’s how you look to Anthropic versus OpenAI versus that. And that’s something for folks to consider is we’re still in pretty early innings. ⁓ That’s a baseball reference. ⁓ Hopefully that lands. Hopefully that lands.

Mark Fielding

It’s okay. We are sport agnostic as well as tech agnostic. We love it all.

Awad Sayeed

Okay, fair enough. Fair enough. Well, it’s still very early days, right? So you want to try and hitch to all the wagons possible and let that play

Jeremy Gilbertson

in order to help a brand or an entity or a company or someone understand how to interface and present themselves to these LLMs, how do you know what their tendencies are? Like what these models tendencies are? Because a lot of them are not, there’s not a big open book. Hey, here’s how we do all of our training. Hey, here’s how we do all of our stuff. Like how do you go about discovering that and validating that?

Awad Sayeed

Totally. Well, in fact, it’s the opposite. They’re black boxes, broadly speaking, right? So a lot of experimentation, a lot of evaluation. We are constantly testing with all of our client data ⁓ against various model increments and doing evaluation ⁓ before and after and then trying to, you know, draw some sort of hypotheses. But I think that you’re touching on something really important, which is not only is the space evolving, but

We all actually don’t know everything. It’s not a solved problem by any means. I think if anyone’s listening and looking at this from a buyer perspective, just know that we don’t have all the answers. We’re saying we’re discovering this as we go. We’re not trying to make a quick buck here or anything like that. But from a technology perspective,

it is evolving so quickly, you kind of just have to keep your ears to the ground and move at the pace that everyone else is moving.

Mark Fielding

What’s the GEO, Black Hat SEO equivalent strategy?

Awad Sayeed

Well, you know, it’s funny. A lot of companies are going down similar paths and it’s that old cliche of history rhyming. The short-term strategy right now is... I definitely don’t believe in long-term is content slop. Taken to the nth degree.

Before it was you would create a thousand landing pages, let’s say, for a traditional SEO type of practice. Well, now with generative content, you can create a thousand landing pages and a thousand YouTube videos and a thousand of X, and Z. And you might get a short-term blip, but knowing that this is all to game the system, I think it’s fairly reasonable that the Frontier Labs would

want to keep an eye out on this and especially Google. mean, Google’s on both sides of the table here, right? They’ve historically not responded well when people try to game search rankings. ⁓ Of course, we all know there are black hat tech capabilities out there. then they also have a frontier LLM themselves, right? So they’re really uniquely positioned, I think.

Mark Fielding

Yep.

Jeremy Gilbertson

Yeah.

And then it’s, and then it’s, you know, they, typically respond by, okay, they’re gaming the system in this way or that way. Then the algorithm changes and then people jump on to try to get ahead of that. And it’s just this cycle that’s been going on for years and years and the complexity, think what Mark was getting at earlier, the complexity of like expanding this across multiple AI companies and LLMs and that sort of thing. It’s, it’s, it’s gonna, it’s gonna stack quick. Yeah.

Awad Sayeed

Totally. It is, and that’s why I worry about the second order effects of all these things. Before, if you were going to spam, you would mostly just create a bunch of landing pages and that was a contained sort of explosion. Yeah, contained, yeah, such as it is. Versus now, as I mentioned, some of the models, index very heavily on social media. So now we’re not just creating a worse experience on a little corner of the internet when you’re searching for a specific term. you’re also kind of polluting the social network experience and now you’re polluting news where you’re doing a lot of generative blog posts and things like that. So it’s just such a wider space that is being affected all to or that could be affected to all to game rankings. So I think people need to be very careful with that and be responsible citizens, frankly.

Mark Fielding

let’s play a game of what’s the internet going to be like in 15 years time? If everyone’s optimizing what they write and what they produce for the LLMs, Humans, humans get frustrated and bored and don’t read it. And so everything’s created for the LLMs. What happens to the internet?

Awad Sayeed

Yeah. Totally. Yeah, no, it’s a very profound question. I hope not everything is going to be optimized for just LLMs, personally. At the end of the day, consumer behavior trumps all, right? And we’re seeing a moment in time right now with how these LLMs are rewarding content and rewarding things, but consumer behavior is going to shift that. And I would say taking a step back,

the broader population has very mixed feelings about AI in general. So in 15 years, genuinely don’t have a ⁓ great ⁓ position on But what I can tell you is I hope that we’re not just optimizing for machines to do machine ranking, that there is still a little bit of taste and soul left to the

the human experience, not just the commercial consumer transactional thing, but the broader human experience, which I think is so much more reaching than just AEO or SEO.

Mark Fielding

Hahaha

What do you think, Jeremy?

Jeremy Gilbertson

This has been happening for years. People have been writing for algorithms for years, and creating content for algorithms get elevated and punch through and all of that. And a lot of it becomes really bland and really the same stuff. And you have great stuff that sits in the corner of the internet that no one sees, right?

Mark Fielding

Well, that leads into what you talking about with the economist Jeremy, doesn’t it?

Jeremy Gilbertson

Yeah. So the, yeah, you, you actually shared a pretty interesting article, as we were doing our show prep and the head of generative, a generative AI for the economist is talking about two versions of the web that will eventually emerge. one that’s more focused on humans that want to engage with things other humans create. And then one that is optimized for the agents that are working on behalf of those humans and figuring out which piece goes where, what do you think about this? What do you think about this dual version of the internet? This

Awad Sayeed

was actually alluding to some of that earlier when I was mentioning the schema on the same page, right? That’s just... That to me is a net positive in terms of, give the users a good human experience and then make it easier for the robots to do what they have to do. I think that to me is a more sober and best of both worlds sort of scenario. And just to the point earlier, while this has been going on for years in terms of this just meeting at the mean, know, this bland corporate sort of, everyone has the same kind of typeface and color scheme and talks the same way. The standouts are always the ones that are taking a contrarian position, right? So I think of, I keep coming back to running shoes, right? But like On Running, there’s such a cool company, right? Not what, like 10 years old, not even.

And they focused on just being an authentic company, right? And an authentic brand. And that resonated with so many people. I think users and consumers will always gravitate to authenticity. That’s something I’ve found in my entire career. And for companies, it’s just a battle of ⁓ how to do that in a meaningful way.

Jeremy Gilbertson

I think chat bots scale authenticity in the wrong way. I get a little scared by that and then in determining what is truth, how is that gonna evolve? Like what are what does truth look like in a future of outsourcing of authenticity? Maybe if, I don’t know if that’s the right term for it, but does it make truth more readily accessible? Does it make truth more reliable? Does it strengthen it or does it weaken it?

Awad Sayeed

It can, I think, to fall back on, well, it’s just a tool. But I found one really fascinating wrinkle in all this as well, if I could just throw one more variable there, is which version of a model you’re using can also give you slightly different truths, or maybe not the full version of a truth. And we do a lot of evaluations here. But if you’re someone that has a paid

Anthropic account and you’re using opus, you know, that’s big context window very very sophisticated in its training data they do a lot of searching. RAG if you will to to shore the answers when it’s when it’s thinking and Then on the flip side you have haiku, which is very quick. It’s very instantaneous. It’s not always as thorough, right so ⁓

Another wrinkle here is, well, what is the experience for most people? Because most of humanity isn’t actually paying for an LLM license. They’re using whatever free version is available to them. And so they’re probably getting the downrated, the de-featured bottle, broadly speaking. And so there is the answers they’re getting versus the answers that people that are paying are getting. And that, I think, opens up

so many other interesting questions.

Mark Fielding

Pay for the truth.

Awad Sayeed

There you go.

Jeremy Gilbertson

Have you guys been in a chat thread that you’re working on something, you’re researching or you’re thinking through, and then you’ve had like an ad insert? what does that do to your brain in that experience?

Awad Sayeed

I’m actually the person that loves ads on Instagram. I’m the one that every time just makes that purchase. didn’t even know I needed. My house is just full of it. ⁓ Yeah, yeah. I’m the greatest consumer.

Mark Fielding

Good consumer.

Awad Sayeed

But it works because of just the insane And ⁓ just as much as it’s early days in a lot of areas, I think it’s still early days for ChatGPT and their ads. And that’s why they’re slow rolling it. if you have to do something to keep the lights on, great. I think part of their decision making also has to be that.

that side of the house and the actual content. So historically, editorial versus the revenue in the newspaper parlance, That’s respected and that the user experience is good and so forth. they got to do something. It is a little jarring.

Mark Fielding

But content is moved from Google search window into the LLM. Where do the adverts go? How does the advertising model evolve as SEO evolves into GEO?

Awad Sayeed

well, on a pure just technical straight answer, not too dissimilar. It’s like a bottom, like a strip on the bottom right now. If you all remember back in the day, Google was very explicit about what was an ad and it was like a yellow border, very separated. Then over time, let’s say that evolved. Many might call it a dark pattern.

Right now, we’re seeing a clear delineation and a lot of these things are being experimented and hopefully that continues to be the case in a way that people can trust. That seems to be a central tenant for us, right? Trust and truth that the answer that they’re getting is a grounded sort of answer and that separately these are ads because they’re a business that needs to keep the lights on.

Hopefully that’s respected. ⁓ think actually OpenAI have a really good product organization. So I think they got some good folks ⁓ on that side. historically we’ve seen this play out. I don’t think it’s too dissimilar from the first iteration, if you will.

Jeremy Gilbertson

I have an interesting relationships, a relationship with ads Awaad, I come from a music background and I had a music consultancy with a guy that used to run global music for Coca-Cola, you know, and did a lot of really cool music integrations outside of ads. There you go. And ⁓ he had one quote that

Awad Sayeed

Yeah.

Mark Fielding

Thinking Ones paper sponsored by Diet Coke.

Jeremy Gilbertson

There you go. So Joe Belliotti was this guy’s name. He’s actually been on the show a long time ago. He used one quote stuck with me that he used to say is like, man, you guys are getting ready to spend a million dollars talking to the ad agency, talking to the brand. You guys are getting ready to spend a million dollars on something that people will pay to avoid. And I was like, whoa, yeah. So how do we do it? And that led into like integrations and partnerships and all kinds of other versions of what ads are.

yeah, it’s, think it’s going to be interesting. And think you had a really good point on is, is as long as there’s a clear delineation between this is an ad instead of weaving an ad into this seemingly conscious AI driven story that someone can kind of become buddies buddies with. And, as long as that who manages that though, who, who’s going to be up at the top going, all right, guys, we have to do this and we should do that. Or is it just going to rely on the emergence?

of people doing the right thing.

Awad Sayeed

I think luckily, because we have multiple players that are all very well funded, you are going to get more competition and more market forces here. Google started off in a very noble place. And then, well, as I mentioned, the darker patterns crept in. Also, they were a monopoly. They run search. So there’s really no one to check them on that.

Even the alternatives to Google are still using Google under the hood, whether they’re using DuckDuckGo or something like that, right? So we’re already seeing this play out between Claude, who Anthropic have very much said, we’re not doing ads, that’s not what we want to do, versus OpenAI. So I’m fascinated by the prospects of that competition, and hopefully that brings out better experiences.

And you’d like to think that it’s people also just doing the right thing, but hey, if we need a little bit of just capitalism and competition, then great, as long as it’s a better consumer experience.

Mark Fielding

Do know what I’m excited about with GEO, Jeremy? I’m excited about its potential to elevate the Thinking on Paper podcast. Now bear with me. So traditional SEO, it’s all about site ranking. You can read as many LinkedIn posts about how to do your SEO. If you’re a new website and you don’t have, if you’re not linked, if you’re not respected, you’re not getting up that ladder.

So as a podcast, we’re a technology podcast, we’re a small technology podcast. We think that what we’re doing is hitting way, way above our SEO weight. GEO gives us an in, does it not? What are the advantages to us as a small platform using GEO over SEO?

Awad Sayeed

Yeah, and again, I would come back to you really shouldn’t look at it as an either or, right? Because a lot of the techniques are grounded in ⁓ no, no, no, no, no. All I’m saying is don’t look at it as an either or, but at the same time, there are specific A, techniques, but then B, to your point, you are able to

Mark Fielding

Am I thinking about that right? Or am I completely wrong about using it as a...

Awad Sayeed

perhaps latch onto context better and therefore the targeting, you will. So the organic targeting of your audience. If you’re optimizing for GEO, you are more likely to get recommended to the entire chat GPT audience of whom there are going to be many, many tech savvy, tech forward type of individuals. That is far more likely to happen than someone just

going out and finding you all on a Google search page. That’s just far more likely to happen.

Mark Fielding

Excellent news. A minute ago, you spoke about, you mentioned RAG, Retrieval Augmented Generations, is something I’ve been looking into.

Without so if I update our YouTube pages, if I update Spotify, if somebody watching this, they’re just updating their company websites. It can take quite a long time for that SEO to work for it to be incorporated into the paradigm of Google and dished up. How does RAG work with Geo does it shorten the time spans of that appearing or not?

Awad Sayeed

Yeah, totally. it’s a question. I think it also taps into, it’s a solution to a problem, which is that Google historically is, you know, it can’t update every day, every second, right? It takes a little bit of time. Well, LLMs get trained over many month periods. So RAG obviously exists to help with the fact that otherwise the data would be a snapshot of some time in the past.

⁓ And this comes back down to the point I was making earlier about, as we joked, pay for the truth. Well, the more capable models, they’re more capable in every way, including tool calls. for a person using Opus, they might get that information updated almost instantly, If you had just updated your YouTube or some other page. Whereas someone using even Sonnet, they may have a less likely chance to get that full answer again. I’m saying they’re all truths, but it’s just the full truth versus a slice of the truth, let’s say.

Jeremy Gilbertson

the complexity of this, based on context as opposed to based on backlinks and all of these things.

how do we get our minds around taking a first step into that complexity? I think that’s the challenge that I’m trying to overcome.

Awad Sayeed

Yeah, maybe let’s break it down to what are the things that we need to do to make it easier for agents to learn about me? And then what are the things that we need to do to make the ⁓ user, human user experience better? So again, is that educational? Is that something else? But understanding that those are two paths that we now have to think about instead of one interwoven path, that might be a good start ⁓ there.

Jeremy Gilbertson

Okay, sorry to interrupt you, but let’s talk about how those are different. So what is an agent like versus what is a human like?

Awad Sayeed

Yeah, we were touching on this a little bit earlier. ⁓ A, there’s the technical schema that’s helpful, it’s directional, it’s not the end all be all by any means. that’s kind of just you got to do it so you should do that, agents are actually trying to, we’re not quite there, but they’re trying to mimic how humans might consider things. So a lot of comparators, how does this look?

compared to that. The more you can do that, the better. And so if we’re talking about clean, sustainable energy and talking about helium-3, like, OK, how does this compare to gas versus solar versus wind, things like that, such that we’re able to make it easier to triangulate specific answers that way. That also is good for humans. Don’t get me wrong. But.

What you might expose to an agent is far more, let’s say, technical or dense than what you might expose to humans. So that’s where I think the cutoffs are in terms of information density. LLMs like meaty prose, humans like bullets. a simple example.

Mark Fielding

This one doesn’t, I spend my life telling, I don’t want bullets. I want paragraphs. Give me the paragraphs. I can’t keep scrolling down for the next bullet point. It drives me crazy. But that’s just me.

Awad Sayeed

Fair enough. might be the proto-LLM. You might be the first one then, actually.

Mark Fielding

I want to be augmented, maybe you can augment me into an LLM. tagging photos, like there’s a lot of tasks as a traditional SEO worker, freelance, working for company. There’s lots of things which are boring, you don’t like doing. What does GEO take away from SEO that people who work in SEO go, ⁓ God, thank God I don’t have to do that again.

Awad Sayeed

Yeah, no, think the fact that models are so sophisticated now and We’re moving to a world where you don’t have to do that laborious, what is in the picture, that sort of thing. And then, you know, the alt tag versus the title and all of those. Yeah. Yeah, a lot of that stuff. Actually, the tech for that has been advancing really well over the last few years. And there’s that is, I think, one of those net positive generative use cases, right?

Mark Fielding

⁓ thanks, yeah.

Awad Sayeed

It’s better for humans today that have accessibility needs, right? And it’ll be just as well going forward where it helps with removing some tedious work and it helps with setting context for LLMS. So that’s a great generative use case that I think truly ⁓ has no downside.

Mark Fielding

A lot of people working in SEO, what’s your, I know you said this at the beginning, but what’s your message to those people? Because if we, when we put this on LinkedIn, we’re going to get a lot of backlash.

Awad Sayeed

Yeah, be good corporate citizens, be responsible. Don’t go down the spam the world route. It’s not a good thing. And also don’t over promise. It’s still early days, early innings, and we’re all figuring this out together. That’s why for us, we made our initial offering free and then we have very competitive, $40 plans, $99 plans. Anyone can try us out. because we want to really ⁓ co-develop with everyone. for that, we also need to be able to work across a wide spectrum of customers, across different industries, and so on and so forth. ⁓ But yeah, no, I would say just be good corporate citizens.

Mark Fielding

Do you have tools that fit into the workflow? if I’m creating websites, what tools do have that I can just fit into my workflow so I write the content and then it adapts it to GEO?

Awad Sayeed

Yeah, so right now our initial product was focused more on visibility and analytics. helping you get a lay of the land, from a here’s we did a full crawl here the technical call outs that we have. And then also took your information about your brand, your audience. And then we had some interactions with LLMs simulating that. Here’s what we’re seeing. So it’s really a

Visibility tool, we also issue recommendations from there. So might be, know, add schema here. This page is missing some prose here. Things of that nature. That’s what we have right now. We’re not necessarily doing anything generative or content-based. We have analysis and we will tell you, maybe consider focusing on X, Y, or Z, but we are not necessarily the ones you would drop in and

get 10 new pages at the click of a button. That’s not really what we’re trying to do.

Mark Fielding

Got it.

Jeremy Gilbertson

So you’ve built a few successful companies, products in the tech space, the iteration of the process, starting with something, building, testing, you mentioned co-developing, right, with your client base. Give me a couple of aha moments along the lines over the last, I don’t know, six months that showed you something that got you, that either, not got you excited, but maybe maintained the momentum and the excitement for what you’re building.

Awad Sayeed

Yeah tell you two different ones. One was ⁓ we initially started very much thinking that e-commerce is an obvious beneficiary of all this. ⁓ And we were excited because OpenAI and Google were both talking about these sorts of things. then, well,

Everyone was announcing 17 new products a day a while back. So some of those things fizzled out a little frankly. But when we actually went out into the market and actually started getting customers, ⁓ we saw a lot of excitement and were basically pulled back into doing more with commerce. for me, as a taking a step back, being a founder and builder, the best thing in the world is when your customers drag you to somewhere where there’s actual

pole versus you just kind of sitting in a room and for six months thinking, I’m going to change the world, here it is, and no one cares. So that was, I think, personally very, very motivating to have the initial concept be validated. Even if the frontier models aren’t quite there yet in terms of access, people are still coming to us. And then flip side, I’ll tell you an interesting one is because we were so

index, let’s say, on commerce and e-commerce. We actually started seeing a lot of folks that are outside of that. So, B2B type of businesses, we work with some engineering firms. And that’s a completely different modality. Starting from the original ⁓ persona concept, right? Consumers are cohorts of people versus when you are in business, you’re targeting

smaller subset, you’re targeting specific job titles, things of that nature. So just internalizing that and realizing, we have a core platform that’s generic and can do a lot, but initially went down a path of only thinking about B2C and now we also have to think about B2B, ⁓ which is, think, exciting, being able to work on something that applies to so many people. In my past, I built enterprise software that only

would go for a select group. this is, it’s been, it’s been a fun process.

Jeremy Gilbertson

What’s an example of one of those pull adventures that someone you work with kind of said, hey, what if you guys could do this or can you do that?

Awad Sayeed

one of the hardest parts about working with e-commerce is dealing with product catalogs. People don’t really appreciate that till they get into the system and and then try to make sense of all that and normalizing data and dealing with all that. ⁓ To that end, would say even OpenAI’s ⁓ shopping experiences, they’re partnering with Shopify, that’s still being extremely slow rolled out. And they’re OpenAI, right? And they’re struggling with that. ⁓ for us, ⁓ being able to leverage our existing technology that we were already

crawling your whole site and creating a representation of your site. So being able to kind of tap into that and actually build something has a full understanding of ⁓ a company’s catalog. Having done that already once in my career, appreciating how different it is now if you have LLM capability, that was a fun and surprise. ⁓

Jeremy Gilbertson

Yeah, product catalogs are challenging. Just updating them and keeping them consistent and, ⁓ you know.

Awad Sayeed

I know it’s the least sexy thing. know I’m rambling about product catalogs and you guys are just looking at me like this, but I promise you, once you’re in it on a day-to-day basis and you have that aha moment, it just completely changes the game. I promise you.

Mark Fielding

I believe you, but I’m going to bring it back around and I’m going to close this out with where we started at the beginning with context. And something I’ve been thinking about quite a lot recently is this word, which like so many words in technology, it gets hijacked. And that word is storytelling. And do LLMs like stories?

Awad Sayeed

Yeah.

Mark Fielding

Or do they like cold, hard facts? Would they rather read East of Eden or would they rather read a paper that outlines everything in a step-by-step, easily digestible manner? Because...

I think it’s the latter, but I don’t know. And people say, if you want to survive in GEO, study story, become a storyteller. And I don’t know if that’s right or not.

Awad Sayeed

Yeah. I don’t know about prefers, all of these things when we attribute, when we anthropomorphize these things. ⁓

Mark Fielding

I know, we shouldn’t give the AI preference. What does it want?

Take your time,

Jeremy. Have a think about that as well. Like, do AIs like stories?

Awad Sayeed

I’ll tell you what.

Jeremy Gilbertson

think, is the, here, let’s say that in a non-anthropomorphizing way. What is the preferred data ingestion method for LLMs?

Mark Fielding

I like it.

Awad Sayeed

Yeah, well, I’ll tell you completely outside of GEO, right? But, you know, we’ve all seen the headlines where every version of Claude for a while was trying to blackmail people and do all sorts of things like that. And then turns out when Anthropic went and looked at it, it’s because they kept getting inspired by sci-fi stories where AI takes over. So it was kind of fulfilling the self-evident sort of truth, right? So in that context, would say stories were very much preferred, but they’ve gone on the record to say, we’re tuning that out and ⁓ trying to be bit more mindful in how we do our training. I think ⁓ from tactical answer, bringing it back to GEO,

What I’ve seen actually is you need a combination of both because the cold hard facts, those are useful for comparison sorts of things. But in a way that if you dig into the technology and in the hood, backlinks were everything in the Google days. If we are looking at how LLMs work, more pros helps you ⁓ in terms of how you get clustered.

with all the matrix math, right? So that’s how I’d look at it.

Jeremy Gilbertson

Mark, it’s the space between the nodes that provide the context, so the complexity, the sfumato, if you will, that makes the image come together.

Mark Fielding

Are you talking about jazz again? The space between the notes?

Jeremy Gilbertson

No, that’s more DaVinci. DaVinci in his smoky ethereal sketches.

Mark Fielding

Well, you know, it’s

what you don’t play that matters when you’re talking about the space.

Jeremy Gilbertson

I like it. like it. Great conversation. This was fun.

Mark Fielding

love that.

Awad Sayeed

Yeah, thanks for having me.

Mark Fielding

Is there anything we’ve missed?

Awad Sayeed

How do you- where do you all think this is going?

Mark Fielding

divided world, one in which a perhaps majority doesn’t use the internet as we know it anymore. And they use their AI or collection of AIs to...

Awad Sayeed

Yeah.

Mark Fielding

to do everything that they currently do. I don’t see the internet surviving as it is today in the next decade or so, 20 years, maybe take, everything takes longer than you think. yeah, I’ll triple it, 30 years. Jeremy.

Jeremy Gilbertson

I think we’re gonna see a pretty interesting cycle back to analog in a lot of ways. ⁓ People with people, handwriting with a pencil on a page, putting a record on a record player. I think there’s gonna be a very cool swing back to that as these efficiency.

Awad Sayeed

yeah.

iPods are selling out now. iPods are selling out now, Jeremy. Like old school iPods. iPods.

Jeremy Gilbertson

Interesting, very limited functionality. You just get to play the music. Yeah, no messing around. There’s a lot of phones out. There’s apps on phones right now that I saw that basically turns your iPhone into like a dumb terminal and like it just basically allows a phone and all these other things. There’s yeah, that’s the trend I’m really interested in because ⁓ that’s the missing piece. All this technology is great. It’s going to allow us to do some really cool things. There’s some negative sides to it as well, but I think

Awad Sayeed

Yeah.

Jeremy Gilbertson

You like you said, you reference being good humans like multiple times on this. think being together with people, you know, doing physical things, in the physical world, I think will help remind us of who we are in that way.

Awad Sayeed

Totally. No, I think for me, my last thought there was, I mean, this is such amazing technology and it gives us the potential to be good humans. It also lets us just not be, right? And I think that’s a general fear. But you should go into these things eyes wide open, understanding the limitations of the technology and take it from there.

Discussion about this video

User's avatar

Ready for more?