Google is finally admitting that chat is the future of search. Bye-bye blue links, hello AI Mode.
The New Search Paradigm
Google's AI Mode represents a fundamental shift in how users interact with search engines. Rather than presenting a list of links for users to explore, AI Mode provides direct, synthesized answers to queries. This conversational approach mimics the experience of asking a knowledgeable assistant rather than sifting through search results.
While this shift is nothing new for anyone who has been using ChatGPT or Perplexity, it's the most meaningful change to the Google UI since they first launched in 1998.
For brands, this means the traditional SEO metrics of ranking positions and click-through rates are becoming less relevant. When users get their answers directly from Google, they have less reason to visit your website, even if your content informed the AI's response.
Measuring AI Visibility
In this new landscape, brands need to track different metrics:
- Brand Mentions: How often is your brand cited in AI-generated responses?
- Sentiment Analysis: When mentioned, how is your brand portrayed?
- Content Citations: Is your content being used to inform AI responses, and is it properly cited?
- Competitive Positioning: How does your visibility compare to competitors in AI-generated answers?
Profound's approach to tracking across AI platforms like ChatGPT, Perplexity, and AI Overviews is now essential for Google's AI Mode as well. By monitoring these new metrics, brands can understand their true digital presence in an AI-first search environment.
Adapting Your Strategy
To succeed in this new paradigm, brands need to:
- Optimize for AI Understanding: Structure content to be easily understood and synthesized by AI systems
- Focus on Entity Recognition: Ensure your brand is recognized as an authoritative entity in your space
- Track AI-Specific Metrics: Implement tools that measure your visibility within AI-generated responses
- Create AI-Friendly Content: Develop content that answers questions comprehensively and authoritatively
The shift to conversational search isn't just a feature addition—it's a fundamental transformation of how users find information online. Brands that adapt quickly will maintain their visibility in this new AI-driven search landscape.
Our Solutions
Profound is already building the tools brands will need to effectively track and optimize for AI Mode and other conversational search interfaces. Our platform provides:
- Comprehensive monitoring of brand mentions across AI search platforms
- Sentiment analysis of how your brand is portrayed
- Competitive intelligence on how your visibility compares to others in your industry
- Actionable insights to improve your AI search presence
As Google's AI Mode becomes the standard way users interact with search, having these insights will be critical for maintaining your digital presence and driving business results.
We'll be announcing a new integration between Profound and Google AI Mode soon.
Get Started
Ready to see how your brand appears in Google AI Mode and other AI search engines? Profound provides the comprehensive visibility you need to succeed in this new search paradigm.
Get a Demo
Video Overview
Watch our video overview of Google AI Mode and what it means for the future of search:
Nick Lafferty (00:00)
Hello, everyone. Welcome to Unpacking Google AI Mode. I'm the head of growth marketing at Profound. And here with the team, Garrett, Nigel, and Josh to go through and answer all your burning questions.
Josh is our leading AI strategist and overall AEO expert at Profound. Nigel is the founder of Organic Growth Marketing, and then Garrett is director of marketing at iPoll, Rink. And today, really, all we're trying to do is go in and just unpack the big announcement from Google, talk about AI mode, just kind of set the stage of, first of all, what is it, and then going into some more tactical questions of where do we go from here?
What do we do? Are we panicking about clicks falling off a cliff or not? That kind of stuff. so first I'll kick it over to Josh to kind of just give us an overview of what is Google AI mode? How does it work? What are your early thoughts on it? And we can go from there.
Josh Blyskal (01:01)
Absolutely. Yeah. think Google AI mode, mean, all right, let's start at the very beginning. You Google's got the Google AI overviews that appears on, you know, around 13 % of queries if we're doing the math. Um, but AI mode is the extension of that functionality, which is an actual full fledged AI chat bot experience. If you click the little AI mode button in the left-hand side of your browser, you're entered into a chat and that chat is going to converse with you and you're going to find everything from, um,
an integration with Google maps to Google shopping. And you're to be able to go and get resources, discuss things. mean, it really is. I think in my perspective right now, the interesting thing about AI mode for me, my like immediate thought when I saw it was like, boom, okay, this is the BC to AD moment that Google has finally admitted that like, okay, search is conversational. We're doing this. It's going to conversational search is the future. Cause I think people, I think people don't have a healthy appreciation that like this is
Yes, it is kind of like a pseudo experiment, but it's also not one of like the quirky little Google labs experiments. This is a feature of Google. Now this is part of the Google experience is a conversational mode of search. And I don't think we're going back anytime soon.
Nick Lafferty (02:15)
Great. Yeah, Nigel, I would love your kind of take on the state of things. And then Garrett, I'll send it to you. And then we'll go into some kind of rapid fire questions.
Nigel (02:25)
Cool, thanks Nick guys. I told you in the lead up to this, I'm gonna say it depends in a whole bunch of different ways here. so I think the first thing like when we really think about it, when a lot of questions are here are gonna be about what can we expect, what's gonna happen. So I think it's important to say what are the variables here? Like besides what we assume, what do we actually know and not know? Like so right now AI mode is an option. When is it going to become the default? How is this going to evolve over time?
One little hypothesis I have is actually that search is moving in a direction where if it used to be just Google traditional search, you now have the surface area expanded. It's multiple LLMs, it's social, it's video. And now even within Google, is there a world where there's a generational gap where older, less tech savvy people just don't want to use this and Google sort of has to find a way to appease them? I'm not saying that's going to happen. I'm just saying, I don't know. And then how is this product going to be adopted? How are people going to like it?
How is that going to influence their behavior? Like the overarching theme here that everyone can see is search is moving more conversational. So if we want to say, where is it going to be in a few years? I can sort of predict that with higher confidence than what's happening next week, next month, over the next year. If everyone remembers when SGE first was trialed at Google IO two years ago, there was this huge panic of, my God, look at these screenshots. What do they mean? And then that experience changed a lot and nothing actually changed for a little while.
I'm not saying nothing's going to change. We're already seeing the change from search. But again, the point being, if the question is, is the sky falling from Google AI mode? I think the answer to that for me is no. But what is the overall trend here? We're shifting paradigms. The old paradigm of Google is we organize all the world's information. You ask the question, we find the page that best answers it. We're now moving to, we've read all the information and how do we summarize it? And that is such a fundamentally different way to think about search, that it changes the way we measure.
the strategy, what we know, what we don't know, and all of that. So, and I think Garrett will touch on this too, but we're kind of moving into a stage where SEO becomes less of this, okay, we know the levers to pull, you make the content, you build the links, you get the traffic, you get the conversions, watch, repeat, and a much more of like uncertainty and bets where we can say, okay, here's what we see now, here's what we think is gonna happen, we think this will make sense, we'll try to measure it this way.
So for me, it is both uncertain and very, very fun. And I'm not really taking anything for granted and all bets are off. But yeah, that's my blurb.
Nick Lafferty (04:59)
Garrett, how do you feel about that?
Garrett Sussman (05:00)
Yeah, it's funny. There's a lot of overlap with Nigel. We're going to have our like strong, you know, polarizing opinions. And yet I totally align in the sense of, you know, going back historically timeline of chat GPT comes out late 2022. That summer we get SGE Google feels they have to be on the reactive. They've been controlling the narrative all around. You have to always go back to the money. Like how's advertising? What role is that going to play in all this? And then there is this
shift in paradigms, the fact of in Google I.O. along with the AI mode, we have this release of this idea of personal context. So nobody gets the same answers anymore. More and more AI mode is able to take into consideration everything that Google knows about you in the Google ecosystem. So what you see as a result of a search is going to be different than what I see as the result of a search. And that's going to make our jobs as marketers, as SEO significantly more different.
and more difficult, Nigel mentioned like kind of the paradigm shift of how we approach this. Like as SEOs in the past, we've been so focused on targeting a specific query, but we've also seen this idea of this fan out where now with the LLM, it can actually predict all the different relevant and reformulated queries that are related to what you're actually searching for. So it's pulling from content that we have to parse out and reverse engineer.
Which makes our job more difficult the other aspect of this is now this point is like nothing is gonna change immediately the difference between AI mode and AI Overviews is AI overviews was put front and center. It's very disruptive. You can't opt out of it It was right in front of you as you see, you know first five eight ten percent of queries would have this experience where all of a sudden you're not seeing the ten blue links you're seeing this generative
response right now out of the gate. were just talking about before this where even those technically been rolled out. We're not seeing it on just a normal search. see it on our labs where we've opted into like a special experiment mode, but we're not seeing it public. so right now it's just a tab where there's still that friction where you have to switch from the traditional search results to the AI mode. And until it becomes the default, which I think it will, like Nigel said, you know, we'll talk about when that runway is.
it's not going to have that disruptive experience quite yet.
Nick Lafferty (07:19)
Yeah, I think that's what's been interesting to me is that Google said it was going to roll out to everyone. And then all four of us tried to find it right before we went live. And it's not live for any of us right now, which is interesting to me. I think some of the chatter I saw online immediately is there's always the both sides. But I saw the panic about a drop in click volume to publishers, to any content creators on the web. And I'm kind of curious what y'all's take is.
on this of until the full rollout happens, until it's kind of like in your face like AI overviews is. I wouldn't personally expect a massive decline in clicks. A lot of publishers are seeing declines in clicks already and impressions increasing. But would love y'all's take on when and if we'll see a decline in clicks and how dramatic that drop is going to be. Josh, I'll start with you.
Josh Blyskal (08:10)
Yeah. mean, here's, here's what I think. I think it's a total lever of the adoption. like if, if 20 % of people, like, you know, AI overviews is on 13 % of queries. People are talking about the 30 % decrease in CTR from AI overviews. a well-known kind of well-cited stat. I think if you see even 10 % of searches going through AI mode, which would be drastically a lot, like to me, that would be like pretty solid adoption from Google. 10 % of people actually.
Opting into jumping into a second tab to have a more personalized experience. think, um, you know, I think the CTR is gonna, it's gonna undergo the same effect. I think we're going to see like that 20, 30 % decline in CTRs. Um, if not like totally punished within the, like the conversational experience. I think the interesting thing to see within AI mode is going to be the actual conversion rate. So within AI overviews.
You've got this kind of like one shot answer, really a lot of discussion. There's really no discussion to be had. There's really no way to kind of drill down and have that full conversation. But what we know from ChatGVT already is that the leads coming from ChatGVT and coming from these chat interfaces are much warmer. I've got experiential insights from brands that saying, we're seeing 25%, 30 % conversion rates on referrals from ChatGVT. That's great. mean, that's, to anyone who knows anything about conversion rates from referral traffic, that's a nutty.
And to have it from one channel in such a unified way, I mean, that's sending a strong signal. Now with AI mode, can we guarantee the same kind of high quality traffic pipeline? You're going to see the same thing. think that's a huge question. I'm pretty bullish on that. think AI mode is going to lead to some really high converting, high intent traffic coming into websites. And I think websites really need to start thinking about,
Are they going to be able to route those people the correct way? Are they going to be able to have experiences that just get them ready to take action, jump in there, make the purchase if they're already ready, if they've been through all the informational phases?
Nick Lafferty (10:09)
Yeah, but before I joined Profound, I was consulting for SaaS companies. And I've got access to a lot of HubSpot accounts. And every single one of my clients had deals that were sourced from ChatGPT. The biggest one was a chunky six-figure ARR deal, or referral source was very cleanly equal to ChatGPT, which is crazy. And I think that'll only continue as like this, the rollout happens.
Nigel and Gar, I'm kind of curious on that too. I saw that for AI mode that there is no refer information kind of being passed through right now. And so are y'all optimistic that that'll change and or that we'll see some of that data in Search Console anytime soon?
Garrett Sussman (10:43)
No, no shot. not seeing a Google Search. No, I mean, like there was talk about whether or not the referral stripping was apparently a bug, which has potentially since been rectified, but not really. I don't anticipate any sort of insights in Google Search Console, even though it makes sense. We've heard talks about like ads getting insights into what queries were used that will ultimately send traffic via AI mode.
What I'll say to your click question though is I think we need to reframe the way we're thinking about clicks and content in general. It's not, not that it's become a vanity metric, but the type of traffic that's going to come to your website and Josh kind of touched on this, isn't going to be from that like informational content anymore. Like we have a responsibility to provide information about our businesses and insights that will get them there of like why they choose us. There'll be entertainment factors through other channels.
But just doing the old like AI slop top of the funnel content, it's like, why do they need a business to do that? And so thinking about the, where the clicks come from, the other aspect of it is like how the UX plays out. Like we're already seeing different iterations, which I don't envy Josh's job. And you guys have profound of like, how do you track all of this? Because you're seeing different types of links. Like you're seeing little cards, you're seeing carousels, you're seeing links that go to other Google searches.
The whole reporting and tracking of all this is a completely different type of challenge in the first place. So I think ultimately the way we attribute all this is going to require a lot of untangling of messiness. Everything's going to continue to be really messy.
Nigel (12:26)
Yeah. Plus one on all that. And even when we talk about concepts, like I think like click through rate is even that's like a concept from a previous paradigm that like borderline doesn't really apply. But another note, we kind of have one foot in the old world and one foot in the new world right now. So we've been talking to our customers about, like we still get some traffic and stuff from Google. Certain things work for that. And then we have this new LLM stuff and AI mode. And how do we handle like
both of these together. So I think that's what we're all figuring out. then as far as, mean, I would not count on Google telling us anything. Will they maybe tell us something? Yeah, but just when you think about it from like a big picture and common sense perspective, if Google already strips a lot of queries out of Search Console, like Search Console is not close to as helpful as it used to be back in the day when it was Webmaster Tools. So like, and if that's partially because queries are getting more personalized and even now a bunch of them get stripped away like,
Say goodbye to all that. Like our best chance is going to be, I don't know, browser plugins, all the weird things. And even then that's only going to be a slice of it. We're probably going to have to go towards more qualitative research to start to understand, okay, who are our customers? And this sounds like a very cliche thing, but I do believe it. Like who are our customers and what are their problems and how do they think versus keyword research tools kind of approach.
Josh Blyskal (13:45)
I mean, yeah, I mean, like in this, in this paradigm, like what is a keyword, right? Like I think Google and, and chat GPT open AI, if they want to, if open AI wants to introduce an ads model, it's like, okay, I submitted my 10 paragraph query with like three pictures and like a CSV file. What's the keyword? All right. Like, go ahead. Like, what's the keyword? Like, what, like if I'm a brand, how do I optimize for it? Right. Where am I going to put my money?
So I think there's something that like here at Profound, we're also thinking about that question too, right? So we're always trying to model keyword volume and things like that. If you know anything about the platform, you know, we've got this awesome keyword volume tool that's coming to the works. But I think there's this place where we're going to kind of meet halfway on that. And I think that the actual model providers and the folks that are doing the generative experiences are going to have to start thinking critically about how do we decide what the unit is? Like, what is the unit? Like, are we talking about?
Is it like an answer thing? we, do we, does it, you know, is it about query regularization? Do I read in your query and turn it into one of 16.7 million standard queries that exists somewhere else that we can like think are like almost just like a fixed basket. Is it about like fixed answers? Does every answer have like a unique code with it? I don't know, but I think if, if we were going to still think critically and run tests on this stuff, it's, it's going to have to be solved.
So there are plenty of things about that area that are still yet to be decided even by the people who make the models and make the systems.
Nick Lafferty (15:09)
What role do you think citations have? Are citations kind of the new rankings now, a big brand like Figma or something might really care a lot that their perspective on design systems and design operations is kind of what surfaced in these responses from AI tools? so what role do you think citations play, having this kind of new mode of AI search? And if you're free, any of y'all to answer.
Josh Blyskal (15:32)
I'm to throw one out there, not to monopolize it, but I think, I think there's like three things about citations. think citations themselves are like, they're a means to a different end, which is like ultimate brand visibility in the risk, like being in the paragraph. Like citations are the way that you claw your way into being the first bullet point in the list of running shoes that, you know, AI mode or chat to BT recommends. Citations are also a way that you can influence the content of that answer itself.
Right? So you can, you can try to shape, you know, what are the factors? What are the features that are being described? Hopefully to try to highlight things that are most important to the customer, most detailed, you know, things like that. I think the third thing is just like this idea of like, maybe something wider than citations, like almost like informed, like informed queries. Like I want to inform the queries. I might care about being in the paragraph. I might care about being the resource.
You know, there are some brands out there, at least some that we work with here Profound who are like, I don't really care about being in the paragraph. I just really want to show my, you know, I run an ads model and a lot of my revenue comes from other companies paying us. I want to show that I'm still relevant in, you know, in search, in generative search. want, you know, brands when they come to us to say, to be able to say, yeah, we are one of the top sources in generative AI. You should still consider spending some money with us, placing some ads with us, putting some content out on our site.
because we're still relevant and we're still being cited and our content still sets a benchmark for some of these queries. So I think there is this kind of share of query of like, how do you, how do you decide what's more valuable? I think it also contextualizes into the brand's goals, but I think basically it comes down to this hopefully eventually quantitative, but right now very qualitative idea of like influence, just like this big kind of question mark of like, what is influence? How do you measure it? How do you care about it?
And how does it go down to the bottom line, which is revenue?
Garrett Sussman (17:24)
I completely agree with Josh. And one thing I want to call out too is we really need to distinguish the difference between AI mode and chat GPT. Like open AI is not a search engine first. Like that's not how they're built. Google 100 % is built as a search engine. So the way they think about using citations as trust signals is not going to be the same as open AI. And maybe over time chat GPT becomes that. So in the context of this call,
Citations still matter, but we do need to also expand two ideas. One is rankings are different. Like it's really not just about being number one. I think it's a general visibility thing because when we try going back to this probabilistic sort of answers and personal context, if I ask for the best bank for me, I'm going to get different answers from than Joshua and Nigel will. And I'm definitely not going to get it in the same order.
So like how useful is it to know who was the first list of the second list of the third list? And I just want to make sure that I am visible in most of the answers for the greatest group of people. That's my target audience. The other aspect of it is because like we're talking chat GBT citations, we don't know how much they ultimately matter outside of, know, Josh's point of influencing the content is the mentions. Like it's not just necessarily the links, but is your brand just specifically mentioned?
in other content, know, stepping away from the trust signals because usually the link was the vote. Now it's just like making sure that your brand is associated with content relevant to whatever the topic someone's searching for.
Josh Blyskal (18:47)
you
Nick Lafferty (19:00)
Is there a world where you optimize for chat GPT and that's different than how you optimize for AI overviews? Or AI mode?
Josh Blyskal (19:07)
Yeah. mean, chat
GBT, the rag model of chat GBT is made by academic. That's like an academic thing. Like engineers treat it like it's like optimizing cash. Like how's the performance of it? Google knows what like SEOs do. Like I think, I think chat GBT is a little bit naive. not in like a bad way or anything. It's like, I wouldn't expect them to like be, and I wouldn't want them to be considerate of like all of us, like over-optimizers. all said in the best way, but like, I, I do expect.
which actually be, and it was totally natural that they're the signals for their ranking are just much more utility based. It's like, does the content like, don't care. It could be from this random spammy domain, but does the content answer the question? So yeah, I think, I think things are going to be, I think things are going to continue to bifurcate. But one other thing I'll throw, like I'll throw kind of a coin into the blender here is the, the deprecation of the thing API.
That's going to be huge, right? That's going to change the whole game. Like where is Chastity BT going to go? Are we going to see this like flattening? Are we going to go on team Google? Is Google going to be open to a partnership or are we going to do our own thing with that OAI search bot, that index bot that they've been building? So diligently this whole time, maybe we're doing something else. I don't know.
Nick Lafferty (20:13)
Since we're almost halfway, I do want to kick it to a question that Alex asked in the beginning. Alex asks, what geotactic strategies have been most effective for your clients? Any unexpected insights? Garrett or Nigel, I'll kind of let y'all go first. And I know Josh has a perspective on this too.
One of y'all answer, please.
Garrett Sussman (20:31)
to monetize, you monopolize. wanted to let Nigel have the shot, but I mean, I think it goes back to the original, GEO study, which is what we experiment a lot with, which is a combination of citations, facts, authoritative language, and then, mirroring the sort of tone of the potential queries. The other big thing for us is reverse engineering, especially like this is our focus for AI overviews and for,
AI mode is what's been described as the fan out, the query fan out, which is trying to identify, like originally when SGE came out and AI overviews came out, people were confused of like, why is an article that's not ranking in the top 100 for this query showing up in the AI overview? And then we realized they were doing a reformulated related query fan out exactly sort of format. And so now it's a process of trying to identify, and we have some proprietary techniques of identifying like.
what that fan out entails. Because you can see it work with AI mode. You can see six different searches, 100 different websites. And so we're reverse engineering that and identifying what those queries are and then creating content that's hyper relevant. One last thing I'll say is focusing on Google and Chachipiti use Fraggles, very specific passages, and look for extreme relevance in terms of
what's called vector embeddings, where you turn words into mathematical values, and then you identify what's called the cosine similarity, which is how mathematically relevant is this paragraph to what someone is searching for? And so that's kind of our scientific mathematical approach for very specific queries.
Nigel (22:11)
I'd add like on a very simple level when you're doing like SEO, GEO, whatever for brands, historically we have to look at the incentives. At Google it was like, well, this is where the search volume is, therefore we care about doing the things that will rank here. Now we can kind of go to first principles and say, okay, if people can ask very specific questions, what are the specific questions that we feel we are best suited to, like that are gonna drive the most value for the business? So if it's a product, then it used to be,
What are all the like clear scope keywords that'll get us ranking now could be, okay, what are the things where we think we win deals and how do we make sure that we are specifically mentioning those and also thinking about what are the ways people self identify? So just like a hypothesis, but if the way people are using LLMs to search is more, it's not best CRM. It's, hey, I work at a 10 person multi-franchise pizza shop in Cincinnati. We have this and this integration. We have this much revenue. Here's how many orders we get.
We have to send texts and like we have a newsletter. Here's what we're trying to do. So you can think about, okay, what are all the verticals that we want to be across times the integrations and all those sorts of things. And how do we put out a signal? And I think the advantage you can have here is like how many pages are there on the internet that say we can solve this vertical and we have these products and we solve these use cases. It doesn't mean you have to include every single possible variation, but
Like I've seen like it's still early enough that I don't feel very confident saying, yeah, we've solved GEO. It's do this and this and this thing. But if very simply, when you have a specific page, it can be relevant for a specific question. And now we're thinking about how do we do this at scale in a way that still makes sense and isn't sloppy.
Josh Blyskal (23:55)
That's awesome. I love that. I think the man, there's so many little interesting things here. one of my favorite little insights right now, and I want to go back to what you said, Garrett, because that, mean, that's the name of the game, right? Like right now I'm obsessed about, like, I don't care what you put into the chat box window. I care about what the chat box says you put in the chat box window. I don't care about the seven paragraph query. I care. Like have I go
Yeah, I want to go to like a hot yoga studio and I want to go to one that's open at seven in the morning and I've never done it before. So I'm a little nervous. And is there one that's like 15 minutes from my neighborhood? That'd be really great. And it'd be like, yeah, okay. Hot yoga studio, Brooklyn, like early morning. Boom. That's the fucking, that's the query. that's the query though, that you're optimizing towards. So, I mean, what, what you end up trying to have to go back down to is, you know, how does that system work? Can you model that system?
How differentiated is that across users? think one of the coolest insights right now is that for ChatGPT's query regularizer, technically the word is regularize, it regularizes the queries, it'll add 2025 to the end of many, many queries, leading to, stunningly, the situation where pages that have 2025 in the URL slug and pages that have 2025 in the title tag end up ranking better.
Is that the way it's going to be forever? No, but that's what happens when, you know, PhDs build like search systems because yeah, that performs great at getting great, fresh insights and information. But now I'm absolutely going to put 2025 in my URL selects. Why not? I'll take the extra 2 % citation rate.
Garrett Sussman (25:33)
And to that point, that's another way you got to realize the differentiator between chat GPT and open AI very much chat GPT. They're both like wild, wild West. And Nigel has been joking, you know, like everything is, it depends at this point, but like there's so much of the old school search engine manipulation you can do for chat GPT that Google's already ahead of it. they're so obsessed with, you know, spam, like, you know, we joke about the whole idea of like white text on a white background, like
that would probably work on chat GPT right now. I think AI mode is gonna catch that sort of stuff. The other consideration, man, there's so many different directions that we can go in. Josh, we're talking about agentic models and the focus of that is a whole other consideration of at what point are we no longer trying to show up for people, but for their own personalized LLMs or their personalized agents.
But one other thing I'll call out from a practical standpoint is being hyper aware of your JavaScript implementation because you are also dependent on crawlers going to your website and being able to surface your content. so, you know, a lot of these chat bot crawlers don't necessarily render JavaScript and they can't get it. And so then it's invisible. So that's like one of the simplest things I would say, regardless right now.
you need to make sure that your content is accessible by these bots, regardless of what type they are.
Josh Blyskal (26:59)
It's the smartest people in the world. It stops at JavaScript. We're not crawling the JavaScript.
Nick Lafferty (27:03)
Do you all have perspectives? know Josh does, but I'm mostly like Garrett and Nigel of LLMs.txt. To my understanding, most crawlers aren't really using this yet. But from my perspective, it kind of just makes intuitive sense to serve up your content in a way that's easiest for these unsophisticated crawlers that prioritize the current year anywhere in the content or in the slug right now. What do you all think of LLMs.txt overall right now? And how should people be thinking about it?
Nigel (27:30)
I mean, I'll give a quick answer, which is that it's not that hard to do. I'm not very optimistic that anytime super soon, it's going to make that much of a difference, but it is also pretty early with all this stuff. I don't know. mean, yeah. To your point, Nick, the core concept of how do we give LLMs content in a way that they can easily consume? That makes a ton of sense. What way actually works like right now and will work is still a bit TBD in my experience.
Garrett Sussman (27:56)
Yeah, I'm with Nigel. It's really hard to get standardization across the entire internet for like a certain industry. If it happens, it happens. Like we can't even decide on what we're calling GEO, AEO, LLMO, let alone like using an LMS.txt type of protocol.
That said, do think, know, we in the SEO where we used to joke about structured data, like not having any actual valuable applications or results other than like, you little star ratings. will say structured data and making sure that your content is in the various knowledge and shopping graphs will continue to grow in importance. So just making sure that you have your information is structured and easily consumable goes back to the JavaScript thing.
for these bots is going to be critical for businesses.
Nick Lafferty (28:41)
Sweet. I'm going go back to a question. This one's kind of long. So I'll read out the most important part, and y'all probably will just see the big shout out to Josh in the beginning. Do you guys think that Google will segment AI search out from organic for improved tracking? My uneducated take on this is I think everything will just move to AI mode. think Google Search as we know it is not going to be around for much longer, but would love your guys' take.
Josh Blyskal (29:06)
I don't know, it'll be interesting to see. One day it'll be like, you'll have to opt into serp. I want to opt into the serp. I don't know. I think eventually, I think these things are gonna flatten at a midpoint, basically. I think there is gonna be an element of still serpiness. I think we're very far on the pendulum swing of conversation. I think it'll be conversation plus serp with a little bit of explainability. Maybe, I don't know, I like to think that there's always gonna be a synthesis between two of these points.
Garrett Sussman (29:35)
Yeah, to that point, think like multimodality, like the goal. think the future state of all this stuff is reducing as much friction as possible. You know, in the past, we've always had to adjust our natural behavior for the technology because the technology was limited. You think perfect examples keywords, right? Like we used to put in a very simple keyword because the search engines didn't understand the nuance of natural language. And, you know, with the natural language processing, bird, mom, and the advancements there, we feel more comfortable asking the full
a question with context, it's going to continue to get more advanced. And we think about using audio, video, text, like whatever is going to be the best way to search in that specific, for that specific question is going to be the way we do it. You look at what they announced beyond with live search and Project Astra at Google IO using videos to, you know, find out information. And ultimately, if we're thinking real future down the line, like search won't be its own thing.
it'll just won't be the core function of what our assistant is. It'll just be something that's naturally a function of however we interact with our day to day, like the movie Her, like will be, it's basically a thing.
Nigel (30:45)
Yeah, to me, the only uncertainty is like, again, like how long does it take us to get from here to there? Like when I think about my parents or extended family that's outside of like our tech bubble, it's like, are they, yeah, like when this gets shoved down their throat, are they like, how are they going to respond? And I don't know. That's what makes me a little less certain on the time horizon of all this. Like, it surprise me if the UX got cleaned up very beautiful, everyone uses it, they're forced to, and it's more useful? Yes.
But it also wouldn't surprise me if this took a while. Think about people in your life who really learned how to properly use Google in the last five years and how long has it been around. So that's the only thing that gives me a bit of hesitation around trying to predict the timeline of the future. Whereas I totally agree with the conclusion of where we're going.
Josh Blyskal (31:31)
Yeah, I this gives me hope that like my parents Alexa is finally going to work for them. I mean, I hope it, I hope if it, if if the intelligence increases, they'll finally be able to do things, know, in depth, agentically. I know I'm thinking about, I think voice is actually kind of underrated for those kinds of use cases, like non-technical use cases where like, I don't know if you've tried chat GPT's voice dictation. It's phenomenal. I use chat GPT's voice dictation more than I use like my Mac. Like I go to chat GPT to get the voice dictation from chat GPT.
to then go put into Gemini 2.5 Pro. That's how much I like their voice dictation. I think voice is underrated. I there's gonna be a lot of ways to interface with that. I wanna use an LLM in my car. I wanna have it like order me like a birthday cake and send invites to my friends in my car in five years. That's my hope.
Nick Lafferty (32:14)
Yeah, I was walking down the street in my neighborhood a week ago and there was a poster on the side with like a thing was happening at a certain date and I didn't want to stop and like open my calendar app and just like put it in manually. I wanted to just like take a picture of it and have it just automatically like add my calendar invite. It knows my wife is with me, add me to it. That's kind of like the LLM world like I kind of want to live in as we kind of cross over to that point. There's a question from Jesse that I have for y'all around.
per Google, AIO clicks and AI mode clicks should be aggregated into Google Search Console's overall click performance. Does anybody know how positions are factored in? My take is the resident ads person here is that Google ads used to tell you what your average position was, like 1, 1.5, 2, whatever. They got rid of that in favor of this metric called search impression share. And you have different metrics like top impression share, which means you're in the top part of the SERP.
absolute top impression share, which means you're effectively number one. So my take is that we're probably going to move to some kind of metric like that, like visibility in AI mode. But what do you guys think?
Nigel (33:20)
jump in. I actually think that like average position in Google Search Console is totally misunderstood and like very abused. Like if you actually think about, for example, you search something, you were in a country, you were using a device, that's a specific query. When you look at average position in Search Console, it's really an average of an average of an average of an average of an average. And we've gone deep down this rabbit hole because we built internal tools and we use the API.
And we noticed a discrepancy between different grains. Like, are you looking at page and query? Are you pulling in country? Are you pulling in device? But the other thing is that the, lot of the time when people look at where do we rank in Google, they're counting the organic ones. Search console actually just counts everything on the page. Like I want to say even like maybe below ads, but even like, so there's a people also ask that's a thing. So you're, you're averaging across all these different dimensions.
And then you're factoring all the things on the page. Like if you look at average position out of the box in Google and you compare it to like a rank tracking tool, they're totally different because U S desktop is different than averaging all the countries and all the devices. This is like a long winded tangent, but I think the point being that I wouldn't interpret that data too literally. And I think I just saw something the other day that everything within an AI overview is considered position one. But again, like who knows how that'll change and how much do we really trust Google anyway?
Garrett Sussman (34:43)
Yeah. mean, to that point, I know we, a lot of us are accountable to higher ups that have expectations when it comes to performance metrics and ROI and showing that we're not just like doing arts and crafts around here. But like the reality is we do need to start to shift the way we think about reporting on organic search, you know, and, kind of going back to what Josh was saying about like conversions, ultimately, like if we move with AI mode, if it's a probabilistic outcome.
and it's different every time, how valuable is that type of metric anyway? Like how valuable is it to know that, you know, out of a thousand searches, you get a certain percentage to what Nick was saying of like, you know, the top visibility or the bottom visibility or, know, like, what does that matter? I think it ultimately comes down to visibility in general. We do want to kind of track it, but we want to focus more on tying it to business revenue.
Um, and then the other thing we haven't really even gotten into this conversation is just how the organic search as a channel, um, and the founder at Apple rank, Mike King talks about this a lot has moving from a performance channel to a brand channel. So that changes the thought process as well as like, you know, our exercises, what we're doing is SEO, organic search, visibility in these channels supports all of your other marketing efforts.
So that as long as you're doing all of your other marketing and people are going to AI mode to find you, that's going to be a better outcome for what you're doing as a marketer. But we really do need to start thinking about how we're reporting on all this differently. And that's what Profound's working on. mean, Josh, you could probably speak to
Nick Lafferty (36:26)
Yeah, there was a great study by Siege Media, too, before Josh, real quick, that just talked about if you show up in AI overviews, that depending if you're B2B or B2C, you see between a 10 to 17 % lift in branded homepage traffic to your site. I'll drop this link in the chat. But that is an interesting shift of branded searches, direct visits kind of increasing as a result of people having much higher up the funnel first touch visits with.
Josh Blyskal (36:26)
Yeah, bye.
Nick Lafferty (36:55)
Google AI mode, for example.
Josh Blyskal (36:58)
Yeah, I feel like predicting this kind of stuff is like, there's this cross section of like, what is possible eventually to model and predict and then what is like reasonable and like productive. Like it reminds me of the, it's, I think I told you about this one, Nick, where it's like the meme of like two ends of the bell curve, where it's like you've got the kind of the idiot on one end, you got the genius on the other.
And they're both like, yeah, just track like the main query. Like it's going to be fine. And then you got the guy in the middle who's just like stressing. He's like, but you've got to get every possible variation of persona and every possible like user background and every personality and every like geo. It's like, I think at a certain level, they're, just, there's going to be this kind like particle physics, kind of like, just like fluid dynamics thing of like.
At a certain point within AI mode and these AI tools, you're going to be optimizing for like, if you really want to go to like the most granular level, you're going to be optimizing for like four ladies named Cindy in like Southern Tallahassee, which like you can like hats off, like, like go ahead. But I mean, it's going to be, you know, I think every search is going to be different. And I think there's just going to be a level where it's productive enough to just get the aggregate of whatever that unit is. and, and being in the discussion at some point is going to seed enough visibility.
is going to get you in enough of these prompts, enough of these discussions to move those metrics in the right direction. like, are we going to be able to say your position too, for the keyword like beach? I don't know. I don't think that's going to happen anymore.
Nick Lafferty (38:23)
is we kind of go down this AI search path where right now user experience and human experience is very much a thing. And what we've seen on our own profound site is that OpenAI is hitting our site 12 times as much as Googlebot. Some of that is a function of how big your site is. But I have to imagine even big sites, OpenAI is probably just slamming until they start caching some of the results that they hit.
Is there a world like, you know, maybe a few years down the road, maybe sooner where it's more about an agent experience instead of like a human experience, even like website design, like visual design might become less important. And, you optimize towards like truly how well is your site optimized for agents instead.
Garrett Sussman (39:10)
It'll be both. It'll be both. mean, you already see like a lot of the talk about the MCP, context protocols where there's more complex sites, marketplaces are building the ability to interact with their sites too. And even that was another thing that Google rolled out at IOWAS, the idea to buy products on the SERPs. I think that we're going to see this differentiation where you have to support both
types of experiences. the one hand, you might not be creating content specifically for organic visibility. It might be for your other channels, but still lives on your site as a hub, as a homepage. Whereas you also want to have this like going back to the structured data and all the data in merchant center or all these other feeds that makes it as easy as possible to purchase. I mean, we see the partnership with Shopify and OpenAI in terms of like that ability.
We also like, have to think about the different industries and business models too. Like I don't want to say media publishing is dead, but like it over the years, like it's been hit so many times to the business model that we're going to see so much moving behind a paywall because of, you know, protecting intellectual property. so in that sense, organic search won't be kind of a channel they'll even want to focus on. So maybe it becomes only an e-commerce focus.
Everything's going to shift. just need to try your best to prepare for how user be, how are people going to interact with their various kind of devices to find your business or your content or your service, whatever.
Nigel (40:40)
Yeah, I don't have much to add to that. The only thing I would add is if you think about like everyone's been pumping out content in similar templates, because that's what Google incentivized. And now we're moving to a point where like, think people sort of have pattern recognition to that. They're burned out on it. And when you actually do get a visitor, you don't want to F it up and you want to put a strong foot forward. So I think if we lean too much into the like, it's all agents, man. I mean, are we going to get to that point? Maybe I don't know. We're very far from it. So
I agree totally with Garret. Like you want to make the content as accessible to agents and bots and whatever. But the end of the day, if you want to get in front of someone and make an impression on them, then you have to think about like, there's actually design is a great level up right now, because if people are used to so much of the same looking stuff, if you can actually differentiate, then like the bar is high to get someone to stick around. Like think about how many searches y'all do every day on chat, jbt, Google, whatever. How often do you stick around on a website? It either has to be
Performing like a very strong utility for you or it has to really like suck you in with the overall experience So you have to crush one or both of those. That's my
Josh Blyskal (41:45)
Well said. Yeah, I like that. I buy into this idea of like, there's two tiers. There's like, I buy into this like two tier of the internet idea of like, I want to see more beautiful web pages. I want to see artistic websites, cool websites, websites that are terribly optimized for bot pickup because underneath those sites are longer LLMs.txt files. I don't know, call it you want. It's longer context, know, bigger amounts of information that like the information chunking, you know,
you know, vicious AI machines of the world, they're just going to inhale like vacuum cleaners and just be able to populate to people and then direct them to the cool front end, the videos, the interesting content, the multimodal stuff that actually lives for the people. I actually think this is going to be a really great change for user experience, letting people do what they want to do on their websites, letting people be creative, letting people do all this. I work with, like architecture firms come to profound and they're like, yeah, we like, we call our blog like.
Like the, like the hub space zone. I'm like, and whatever, it's not actually that, but it's like some weird abstract name, not a blog. And then I'm like, like, that's rough. Like, because we really need to like call it the blog so that like the LLMs and answer engines kind of know, you know, when they're scraping that page, it's going to be a good, like normal page to scrape. they're like, yeah, we do like, kind of this like, like kind of cipher code thing with our title tags. I'm like, that's interesting too. but I want them to do that. I just want them to be able to do that. Like there, there's no reason why.
With the advent of answer engines, we have to continue on that same paradigm. So I'm really excited for what the future holds.
Garrett Sussman (43:18)
Yeah, two quick call outs to Josh's point is one, that's such a good point that just the user experience, like, even if it's frustrating for us as businesses, I think users win with all the development of this technology. We're at a nascent stage, so there are going to be a lot of problems. There going to be a lot of bad results. There are going to be a lot of bad UX experiences, but I think we'll get there. The other thing to Nigel's point is that the tools, I know there's some people who are anti generative AI for content creation or whatever, but like,
The tools are also going to get better too. And we're going to be able to create some amazing things. If that isn't like your natural skill set. Like I just think human creativity is at like a starting point where there's a certain democratization, which I know there are reasons ethically that it's problematic, but for the sheer like fun creation, I get really excited.
Nick Lafferty (44:08)
Yeah, Rose has a good question in the chat. And we briefly touched on this earlier. But she asks, for example, would regularly interacting with AI tools or chat bots help to encourage Google's AI mode or chat GPT to mention your brand more? My take is that no, I don't think that would be a driver. Otherwise,
companies like Profound could have an outside impact on driving visibility for our customers. I will say that you probably have something hidden on your site that you're not using, and it's if you have any gated content, if you're in the B2B realm and you have this beautiful PDF document hidden behind a form that the chatbot can't access, I would look into how you would ungate that and try and serve that up to LLMs, as that's a wealth of good information that probably isn't being picked up right now.
Do y'all have takes on?
Garrett Sussman (44:57)
watching, I'm watching Josh.
Josh Blyskal (44:58)
Anyone else?
Garrett Sussman (44:59)
So my thing
is two things. One is don't ever get caught in your individual, going back to the, all get the different results, like inside thing. One time I asked, uh, chat, GBT, what's the best, you know, B2B SEO agency ever. And it told me I pull rank and that's the agency I worked for. And I was like, oh, well, why'd you say that? It's like, well, I know that you're the director of marketing at IPR. So that makes you feel great. And I was like, yes, it did. But like that's.
Not helpful for me. so whenever your, your interactions, like we are, you'll see like the option that whatever you're doing interacting with it is training data for the LMS. Now it's miniscule in the grand scheme of things, but there are also like, it constantly asks for feedback. Google does this as well. The thumbs up, thumbs down. And like that matters. Like every time you're giving a data point as to whether or not it's a good or bad answer, you're in essence.
training and basically the fact of like how many people are training on your brand Maybe it doesn't work, you know down the line but right now every thumbs up is probably going to have a little like, know But I don't think I don't know you should spend your time like, you know Hire someone for 10 hours today to sit and do thumbs up for all that which probably would have an impact But I do think it matters
Josh Blyskal (46:19)
I'm just imagining like, like a, like a warehouse full of servers, just clicking thumbs up on all the profound mentions in AI search. I mean, I think there's a few things there's like, there's the element of like, you'd be impacting the rag model. You probably don't even really care about the LLM. If you, if you're trying to train an LLM, you're on like a decades long journey. it's a separate one from ours, but it's, it's, it's a nice long one. I think.
It's really hard to say. think the scale would have to be incredible. I think you'd have to have like an incredible scaled operation across, you know, continents and time zones and coordinated across many different factors to just like, like everything that works well and dislike everything that doesn't. Maybe that's a factor. Yeah. Anyone else?
Nigel (47:06)
Yeah, I don't have a strong opinion on this at all. I mean, I've asked like an ML engineer this question a while ago and he said it's unlikely to work. I've heard people on the internet say it works, which makes me think it probably doesn't work. Maybe it does. I don't know. But to me, it's like, I don't mean it's in a bad way, but it's not the right question. It's like, I don't think this is the lever that's going to like give you the most bang for your buck. If you can figure out something, some weird thing that works here, I guess, go nuts. I wouldn't tell anybody though, like
Josh Blyskal (47:34)
Yeah.
Nigel (47:34)
Keep that to yourself, take it to the bank.
Nick Lafferty (47:37)
I think there is some element of that
Garrett Sussman (47:37)
And to that point.
Nick Lafferty (47:38)
right
now. There's probably things that are working and people aren't talking about them.
Garrett Sussman (47:43)
was just going say one thing to keep in mind, think is going back to the old school days of SEO right now is like, it's a while, while less experiment, run experiments for yourself, for your brand. you know, Nigel going back to Nigel's, it depends. Like it depends is when we talk about generalities, but when you get very specific into your niche, into your topic, into your segment, there are opportunities to experiment with different things. And if you find something that works, keep it to yourself, but figure out and double down.
Josh Blyskal (48:10)
Yeah, true that. And if it does work, if liking something on ggbt does work and it becomes public, I promise you within six months it won't. Or it probably won't. These things have a way of like, the exploit is known, the exploit proliferates, the exploit is patched. That's the cycle of SEO.
Nick Lafferty (48:26)
Right on. Well, hey, guys. Thank you so much for hopping on. think this will be probably the time that we start to wrap up. If there's any last minute questions in the chat, drop them now, and I'll flash what I think is the most interesting one on the screen. If you want to talk to anyone who you have seen here, Garrett at iPollRank would love to talk with you. You can find him on LinkedIn. Go search Google AI mode for iPollRank. The same for Nigel.
Nigel runs organic growth marketing, would love to talk to anyone. And then if you're interested in talking with Profound and you submit a demo request, we will try. We will try very hard to get back to you in a timely manner. And yeah, please stay tuned for more news from us in the coming weeks. So if there aren't any more questions, guys, I'll let you go. And thanks so much for hopping on. This was great.