How To Win the Mentions that Win the Market | Ali McCarty and Alex Halliday

Hi everyone. Welcome. Really excited to talk about offsite. Why don't you tell everybody who you are?

Okay, great to meet you all. My name is Alex Halliday. I am the CEO here at AirOps. We've been building the company for the last 3 and a half years to help marketers meet this moment where discovery is changing really rapidly. We really believe in helping folks kind of learn together. So we're all students at this moment. Things are changing really fast and our webinar motion and things like this have been really core to how we want to invest in helping people learn, experiment, get ideas for plays they could run, and obviously help them with our platform as well. So we're really excited to do this. We love doing it. It's a big part of our kind of activity in the market and something that we've had a lot of success with. So keen to continue that trend today.

And then hi everyone, I'm Allie McCarty. I am our VP of Strategy and New Verticals here, and I'm spending all of my time right now thinking about how third-party websites can help brands drive influence across AI search and SEO. And so really excited to go deeper and, and chat through this as a thing that we all need to start thinking about, um, as we're thinking about our strategies for 2026 and beyond.

Overall, we're seeing this large macro trend that's happening in some areas really quickly, in other areas. A little bit more steadily, but the way in which people are finding and buying their next favorite product is shifting. And between you and your next customer sits a new layer, which we think of as agents, but AI Search, AI Overviews, we're kind of thinking about that generative layer as being a major challenge. An opportunity for marketers. And our role as AirOps is to help folks navigate this shift. We want folks to be in the practice of experimentation and learning in this moment, and obviously building strategies that, that win and hopefully get people promoted along the way.

When did you realize, like, think that this was going to, like, the aha moment of like, this is the next big channel versus it just be like the latest fad?

I, so I think when you start seeing people around you divert their first query to a generative experience, like in your life, even my mother, right? She's 72 years old and she's, and she's sitting there and, and, and in ChatGPT and, and is using that to decide her holidays, her vacations, her which, whether she should upgrade her iPhone. And you start seeing behavior around you change. And the other big thing was just we speak to CMOs and marketing teams every single day. And from, I would say, mid-January 2025, it became the singular topic on, on every single call we were on. And, you know, that was, that was massive signal to us. Clearly, our customers were starting to see it come up as a channel in their weekly reporting. And that was starting to drive urgency and questions given some of the challenges that this channel poses, I think, was, was, was actually one of the first kind of factors of concern. And now ever since then, I've been building, hopefully, the playbook and the platforms go after that and figure out how to navigate it. But unlike many channels, you know, it's also one that shifts very quickly. Every single time OpenAI snapshots one of its models and pushes it to ChatGPT, there are subtle behavior changes in there as well. And so Much like a Google core update, these things can have pretty big shifts in things like the pages that are getting cited and retrieved. And so I think it calls for a pretty big rethink in how we build the systems to go after this channel. So we talk a lot about this in other webinars, but it is a, it's a big shift in how we organize ourselves as marketers, how we think about the role of content and how we think about driving sentiment and influence in agents and humans. And so a couple of things. One of the important shifts here is that there is a behavior change on the, on the, on the buyer side. In the old world, there would be a single short search term, 2 to 4 words, hotels Berlin. And in the new world, if we go to the next slide, folks are actually doing much more substantial search queries. And what I think is interesting is that not only is the search query getting longer, but it's also being enriched by some meaningful personalization context. So platforms like ChatGPT are building context payloads about you as a buyer to then really make sure that they're retrieving the right specific information relevant to you and your preferences. And Google, on the other hand, has also added in recently Personal Intelligence, which brings your YouTube preferences, Google Drive, Gmail, all of that into the generative experience to make it increasingly personalized. And this pushes a lot of the, the, the retrieval activities of these models into the long tail of search, whether it's consumers increasingly going to ChatGPT to ask questions to find their next favorite product or even confirm a buying decision they formed off, off, off in another channel, or it's Google increasingly becoming generative as a surface in its own right with first AI overviews and then the increasing affordance of AI mode. You know, the majority of queries will have at least one generative component in them, and that surface is going to become higher quality, more personalized, more efficient. For buyers, which has a really meaningful impact on, on the stakes, right? You really need to make sure that your product, your service is mentioned, that your product is represented accurately, and that ultimately you're winning that increasingly shorter buyer journey through your activities, of which offsite is a massive portion. Depending on your category, when we'll talk about this today, as much as 85% of top of funnel visibility can be driven by context that is retrieved from unowned content on the web. So concretely, this stat comes from research we did in end of last year, which is for B2B SaaS top-of-funnel terms. And when we analyzed all of the pages being retrieved and cited, 85% of those for those, for those queries that we looked at were from unowned domains. And so why are we here today? Why are we talking about offsite? It's because If you're going to place your chips as a marketer to increase awareness and top of funnel, then there is a huge opportunity in what people are saying about you off your own website. And of course, we should talk about backlinks for a second because, you know, we come from a world where backlinks were a big part of the traditional SEO playbook, and they're still important for, from the SERP authority sense. But the new dimension of this is the quality and specificity and depth of the mentions that are made of your brand on third-party websites is now a really important surface area and one that I think people don't fully understand. You know, if, if ChatGPT is forming an opinion about your business and it starts talking about your shipping speed or your customer service being, being maybe not as good as the competition, that sentiment and response is being driven by the consensus that it is observing across all the pages it retrieves. And its own training data. And so what's really important is that, um, you make sure that you understand the levers, um, and the contributing factors to, um, the, that sentiment that's being expressed, and you're able to drill down into that and, and start to, uh, address some of those things. Should I take this one? So yeah, for sure. So the AI answer engine shift, um, I think as I mentioned a little bit how LMs decide who gets named and who gets cited. It's, it's, it's, um, it's a probabilistic engine, right? So, um, LMs have their own training biases. They have, they have knowledge they, um, are, uh, they are trained on initially. So they understand a model of the world and, and, and, um, the ontology of the world. And then at runtime, when someone puts in a specific question, there's— they make a determination whether or not they have sufficient information in their training set, or they need to reach out of that to get new information. So they need to go retrieve, other sources on the web. And so the way it works today is for queries, particularly temporal queries where it's like best of X 2026, or where shall I go next week in New York for dinner? They will, they will opt to reach out of their known universe of knowledge and perform a set of fanout queries to then retrieve content on the web, which uses traditional search. To go and fetch the latest information. And it uses the authority graph and domain authority, topical authority, and all the things that have been built in traditional search to understand which sources should be at least looked at. And then a portion of those will be cited. Is the second— go ahead.

Even there. So how much do we have a sense on from an LLM perspective? How much is pulling from the model training data versus the net new like web scrape of live updated content, or is that still opaque?

It's, it's, it's opaque. It really depends on the nature of the question as well. We, I recently spoke to a large travel website and they were looking to do FAQ content on their, on some of their listing pages, but the FAQs that they were looking to create were things that would very much be in the training data of the LLM, right? So it's like, what year was the you know, the Sydney Opera House built. Now putting that on an FAQ of a page is not going to help you in AI search because that information is very much in the training data of the model. What, what, what times the ticket office is open in 2026, however, is much more likely to be something that the model is going to go and try and find an authoritative source. So answering a totally different question than the one you asked, kind of. But, but the point being that you want to make sure you're the content you're creating is aligned with intents that you observe consumers expressing, or real questions people are asking, but also things that are likely to be, um, outside of that core training data, so you increase your chances of citation. So because, given the probabilistic nature of these things, we should avoid talking about them as like rule-based engines, as if there is a rule. Like, you, if, if, if you can observe behaviors easier than you can back out rules, meaning like testing things, seeing what works, doing structured experiments in your vertical for your, for your business is the best way to go here. In our view, one thing that Ethan Smith actually talks about, we just ran an event with Ethan from Graphite and he pulled— this is actually from our data at AirOps. We collect about 3 million AI search answers a day and we analyze those. And one of the interesting stats was just how long the long tail is. So you'll see a lot of charts online that will be pie charts of like most cited domains, and Reddit will be like represented as like 30% or something. But that's only as a percentage of the top 5. It's not as a representative percentage of the entirety. And so it's really important to understand the distribution of current citation influence across the spectrum of domains that are being looked at. And often you will find a lot of opportunity in the long tail of domains that you can build relationships with or conversations you can participate in.

Which is so interesting because that's so different than, than for Google. Like long tail never mattered before, where today it does.

Yeah. And I think long tail is, is the— if there was one takeaway from the shift to AI search is the long tail. And it's, it's, it's because the, the, the specificity of the fan out queries and the uniqueness of the fan out queries me, um, along with this long tail of, um, uh, of citations and the fact that domain author— lower domain authority websites actually can perform really well in citations means that the long tail is kind of having its blow-up era. Um, and, uh, there's a lot of opportunity there, particularly because most people aren't looking at it or building content for it. So definitely, if you can organize yourself in systems, and really the AROPS is, is is really excited about this movement. But the idea of building systems that can get off the long tail systematically, I think, is one of the most interesting opportunity areas in organic right now. So lots of great stuff there.

And that's what we're seeing across the board with the customers we're helping support on these efforts, which we'll get into a little bit later, is actually driving mentions in these long tail sites is seeing a visible increase in AI visibility post. So like we'd never would expect, honestly. So there's a ton of opportunity now to invest in. I think the question is we've, we've branded this offsite. So kind of, we think about it as like, what does that really mean? It's not your team offsite. We're not headed to Vegas altogether, but we really think about it as all of the real estate across the internet that you don't own. So everything on your domain, blog, product content, help centers, that's all things that teams should first and foremost be investing in today and, and really driving, you know, refreshes of content, making sure that you're structuring content in ways that the LLMs can, can cite and, and answer the questions that folks are asking. But for offsite specifically, these are all of the websites driving this influence that don't belong to you. And so some of that could be competitor sites. We're obviously not going to have a huge say in what can happen there. A lot of it is also social. So having, you know, strategies around your social and community sites is really important right now. But as Alex was mentioning, there is a huge opportunity across, you know, kind of this middle and long tail of affiliates, blogs, review sites that potentially are opportunities for you to work with and see if you can get your brand as part of their story and mentioned in those articles.

I think that one of the takeaways is prioritization as well. I think it's just there's so many things you can go after and it really starts in auditing where the LLMs are going today. So in our product, in others, there's, we spend a lot of time going out and figuring out where the foot traffic of the agents is going currently across the web. Now that's like, you don't want to over-rotate on that because obviously that's the known universe of content that it's pulling from. But it is very interesting to start to get a sense of the types of content, types of domains, um, and you'll start to see patterns that will give you clues. But you don't want to just singularly like over-rotate on what's happening today because obviously there are probably publishers out there that have great topical authority. That you could build content into that would perform very well. Um, and so it's a, it's a, it's a balancing act as it always is. And, and that's part of the art of what folks on this call do to make those judgment calls. Um, but it's good, it's good kind of detective, uh, it's good detective, uh, fodder for, for starting your journey.

Definitely. And these are the types of sites that we consistently see, uh, being used across you know, LLMs as, as the cited sources to get the response in the, in the answer engine. Um, so everyone's talking about Reddit right now. It's a hot topic. We're seeing that user-generated content be, uh, really important to have an unbiased opinion and tell the real story about your brand. That's why I think the LLMs really love, love Reddit as a source, even though it can be quite unhinged at times. Um, but a huge uptick in the affiliate sites Comparison review sites are huge. It goes back to that question of how does your product compare to a competitor and so forth, and being able to answer that in a structured way. So oftentimes we see listicles get picked up across the board, and it doesn't matter if it's from a NerdWallet-level site all the way down to alexreviews.com. If it's very topical for your brand and subject matter, we'll you'll see it cited quite a bit, specifically if it answers a very specific question as well within like a product line or feature set. External blogs, so this is everything from your competitors' blogs, adjacent company blogs, as well as, you know, individual users that have built, you know, content around the subject matter. And then more so, and this is a trend that continues to pick up, is YouTube videos. We're starting to see YouTube continue to, to get cited. And I think as Google takes over more and more share, um, of LLM usage, we're going to see YouTube be a bigger and bigger channel here. So it's one area to start thinking about, um, as well as then obviously like industry publications that are really trusted. Trusted sources for your, for your specific area and so forth.

One thing I will say is that like this also shifts, this mix shift just changes a lot by industry, stage of funnel, and also just over time. So it's just important to keep an eye on it. Listicles is one where there was a, there was one model snapshot which suddenly kicked listicles out and then they came back again. Like I think that the most durable strategy here is to keep an eye on it as, as things change. YouTube is an interesting one. It's a good example of, especially in B2B, there's so few YouTube videos that answer basic questions that doing even simple YouTube videos that answer B2B questions can perform really, really well because there just isn't that much there. And YouTube is a trusted source, especially if you're a lower DA site. And we've definitely, we've seen success across a number of companies creating, you know, pretty basic videos, but with rich descriptions, great titles that align with the questions that can, that can really crush.

Um, so I'm going to pause. Ashlyn, are there any good questions in the chat that we should address?

Yeah, we had a juicy one. I don't know which of you guys might want to tackle it. Um, but it was a question around, does AI know if the mention of the link is sponsored placement versus earned? Um, and does it matter as long as, uh, the quality of the site is high and it's a trusted site? Uh, I don't know if either of you guys want to tackle that question.

There isn't like a formal standard around this. I mean, essentially what's happening is that the model is retrieving the page and reading the contents and then making its— going through its algorithm. Candidly, we haven't done enough research on this one yet. I think we've observed in a lot of cases that sponsored posts and other things are being cited perfectly fine, but I wouldn't say we've done definitive research on that one yet. And I don't think there's a hard rule yet around this or any guidance provided. So it's very— This is very early in the, in the sort of maturity of this ecosystem. So definitely a great research topic. I think we'll chat to our research team about that. Maybe we can get that prioritized.

As you think about offsite as a channel and working with, you know, third-party publishers and collaboration, I think it all comes back to ensuring that the content that you're developing answers the question for the site and the type of content that's being published. So for videos, it's how does it work? Social, it's really reviews from, from real users. These affiliate sites are, you know, X versus Y, which one is better for whatever use case that I'm going after. And, you know, blogs are all over the place, can be a bit more open-ended, but it comes back to similar similar to what you should be creating for your own domain is, is really getting back to this piece that Alex keeps hitting on of how do you answer the question that users could be having a chat with their, you know, chatbot interface, um, is really key to, to winning here versus just in this world, the previous world of link building, just making sure that you, it linked back to your website. All everything here comes back to what is the narrative about my brand and how do I make sure we structure content to answer the right question in the right format is really key as you start exploring this as a channel.

I'm going to quickly jump in. The chat is telling me that I missed an important question that would actually be a really good discussion one. The question is, they say, I don't know who they is. They say top of funnel is dead. So is it really better to work on driving AI visibility to middle of funnel and bottom of funnel terms? I'm sure you guys have thoughts on that.

I think it really depends. I certainly think that like basic informational top of funnel is not, hasn't— it would not be where I would place the same degree of chips as I would have 2 years ago. At the same time, there's a lot of top of funnel questions which are kind of more long tail, like more interesting, more specific, more temporal, where I think there's still a lot of opportunity or questions, top of funnel questions where you are in a unique position to provide the answer in the— I'll give you one example. So Zola, which is a wedding planning site that works with us, they leverage their pricing data to answer really interesting top of funnel questions and because they are authoritative, they have something unique to add, and that is a shifting metric. Those, those pricing, that pricing dimension shifts over time. It's actually a really great strategy for them to go after it. So as with most things, it sort of, it depends. But I would not build the same informational pages that we would have, you know, 18 months or 2 years ago in this new world. That's probably not the best use of your time.

Okay. So this comes up quite a bit, um, from folks I talk to is, is how is this new offsite effort different from traditional link building? And I think Alex has already touched on this a bit, but the biggest thing in this world is it's all about how your brand is described and talked about and what is being said. Um, and that's from everything from, are we talking about the value props correctly? We're talking about the product lines and feature sets correctly. And the sentiment behind it. Is it positive? Especially from the user-generated content as well. So everything is all around what is being said because that's being ingested by the models and then being determined as, as part of the response back, like a research report of taking it all in versus in the traditional link building world, we're just trying to build authority. You know, and credibility that there is enough traffic to our site that, that it deems, um, that it's viable and, and should be ranked higher. Um, so it's really all about the narrative, uh, versus just, you know, traffic from other sites coming in. Um, and that's the opportunity set and also what makes going after this area a bit harder. Um, because you have to think about it from a narrative perspective and not just a volume play, it's a huge opportunity for those who are willing, willing to invest right now.

Yeah, I think this is where the opinions, these kind of— it's not just about your brand showing up, it's about your products showing up. It's about the accuracy of the products, the features, the attributes, and often like people also changing their business, you know, going through pricing changes. Like new rollouts and these, this sort of information that's feeding the model's understanding of those product attributes and a new business is gonna lag reality. So there's a big change management process here as well to think about. So it's a fascinating new challenge.

We've actually helped a customer specifically with that where they've pivoted their business model and wanted to, you know, going through a transition period of, of wanting to be completely known for something new and having to essentially reach out and work with publishers to retell that story so that the LLMs were picking up their new version of their product and not citing all of this old, out-of-date, out-of-date copy and things like that. So that's a very real use case across the board. I see a lot of questions coming in. Ashley, is there anything we can answer?

Yeah, if I'm understanding the question correctly around YouTube, and I think the same question applies on Reddit, um, like, is the authority of YouTube as a site or Reddit as a site actually taken into account, or are large language models looking at the authority of the individual, like, subreddit or YouTube channel? It's a really hard question, but I'm curious if we have any thoughts.

Alex, I'd be curious to hear your thoughts.

I mean, it's about— the thing about models is they're a black box, right?

Yeah.

Someone didn't write— sit down one day and sort of write a bunch of if-else statements to kind of like make them do a thing. There's obviously, there's obviously a set of instructions they're given. There's this giant kind of neural net that's created to, to come up with, with conclusions. And then there's certain behaviors that are encouraged, um, by the, by the training process. And this is why, you know, it's like really treating it as a system that needs— you need to collect evidence on its behaviors relevant to your industry. And so really trying to figure out what works. And what we hope to do is like provide snapshots that we've gone and gathered and drawn conclusions from them to give you clues as to where to go next. And so, as has always been the case with organic, actually place your chips, see the results. Double down on what works. It's the same playbook. It's just the system is a— the behaviors in the system are a little bit more sophisticated and variable in this new world. I do think this is going to move quickly though, because silly things work today that really I don't think will work in a year from now. And so as with SEO, you know, there's a little bit of a cat and mouse in some of the short-term tactics. And so building authentic, real conversation, high-quality content, these things that we think of much more durable strategies. Than simply just optimizing for some shortcuts today.

Absolutely. I saw in the chat, Sam had said, one of the most cited YouTube channels in our space gets like 20 views per video and 5 posts a day. That's the perfect type of opportunity, Sam, though, to reach out to that creator and understand how can we work together as you know that, you know, they're not, they don't have a million subscribers today, they might be more amenable to that. And it creates, those are the types of opportunities today that we're seeing across the long tail that brands can capture if you're focused on the space and, and drive, drive real impact. And so on impact, does it actually work? Like, you know, we keep talking about this, we know that there are third-party sites are driving this influence, but if I actually set up time with my team, and we focus on this, are we going to actually see results? And so across the industry, so we're not going to just talk about what we see at Airops, but across the industry, we're seeing early signs of folks focusing on this. So, you know, one of a close partner of ours, Kevin Indig, did a recent analysis, you can see this on his LinkedIn and on his blog that is validating this, that AI trusts these comparison sites more than your brand page. And so going to third-party sites to get this data is where the LLMs are going. So how do you, again, focus and take the energy and time to invest in being there and start to develop strategies, strategies to do so? And then the team over at Graphite actually ran an experiment themselves where they reached out to publishers to be included in, in these articles that we're talking about, you know, space they wanted to win. So for example, becoming best SEO agency, they had absolutely no visibility across the LLMs before the test. And then after the test, they saw they were in the top 3 answers. And this is a short period of time, I believe his test was like 6 weeks long. So real effort in short sprints is already showing significant improvement and just shows that like this is an area to invest in now to take advantage of being an early mover in the space, taking advantage of things that work today that might not work down the road but can build a foundation and then thinking about how to continuously evolve and systematize efforts over time. To really capture more and more share. I saw a great question. Uh, Ashlyn, is there something we should answer? Yeah.

Um, I think one of the questions was around like, how do you control the narrative out with third parties that are mentioning you that are maybe talking about you negatively? Maybe they're talking about you, um, with like an outdated product positioning. Like maybe you've launched new products since then and third parties don't know about that. Or maybe they're talking about you being really good for like small businesses versus enterprise, how can brands approach solving that problem?

I have thoughts, Alex, but do you want to go first? Please, Alex Halliday, take it away. Awesome. So one thing that is really key right now is having fresh, up-to-date content. So not only can you reach out to these specific publishers and let them know and provide them with the right content to be included. Um, because nobody wants to tell the wrong story. So there is incentive there, but also positioning it as if you refresh your content right now, that is more likely to actually get cited and drive more traffic as well across AI search. So there is a win-win angle there. Um, but I think just first and foremost, being forthcoming that this is, this is where our product has shifted. You know, we help support customers do this all the time. And we've found that publishers are actually very amenable to working together on making sure that the content is correct. But also leading with this input of, you know, refreshing your article should drive more traffic is also something that has been received very well across the market. Alex, do you have any other thoughts?

No, I think that the appetite for people— I do think there's a real opportunity to do this in a kind of win-win way. And to Ali's point, like content quality, especially in pages that potentially have gone a bit stale, has never mattered more. And so framing this as like, hey, like your content is out of date, like keeping it up to date not only helps with SERP, it helps with citation, it helps with the hygiene of the content you've already created, I think is a way to align incentives really nicely. So yeah, there's definitely a lot to do here. And I think we're sort of in the early innings of this, but that we, from the, from what we've seen in the market, there's a pretty good, people are pretty responsive to this kind of message.

Definitely. And there's a fun question for you. When will it be possible to build AirOps workflows via MCP?

So you can now create grids through MCP. You can, I believe, run grids and run workflows through the MCP. Creating workflows to the MCP is, uh, I don't want to say it's the wrong question, but I think we're going to shift the paradigm soon away from workflows a little bit. So that will be announced next month, and that will be available by MCP. So good. I think people increasingly want to work through MCP or CLI, and we know that, so we're building a lot there. That's probably to other webinars that we should do.

Yeah, there's a little breadcrumb.

But like, definitely email me, Alex at AirOps, if you have specific questions on that. I would love to hear your use case as well because we're actively working on that.

Breadcrumb, huge news, love it. We, uh, at AirOps, we've been supporting customers through driving offsite visibility, um, here since the beginning of the year, um, and have already started to see, similar to Ethan and Kevin's research, started to see, you know, really big improvements when this is a focused and concerted effort. So for example, Synthesia is a B2B AI video creation platform, and they are consistently launching new feature sets that are really important. It's a very crowded space. And they want to make sure to talk to their core ICP. And so we've been very focused on specific features within, you know, all the things that they're tracking to drive AI visibility in and hopefully break away from the competition in terms of winning with these specific topics. And so right away, we've seen a 15% improvement in just brand mention growth compared to the topics we haven't focused on. And overall, we've seen almost a 3x improvement compared to the entire set of prompts that they're tracking is the ones where we landed brand mentions across the board are just outperforming from a visibility perspective. And this is the interesting thing is this is growing over time. Early on, we saw meaningful improvement from January to February, but then it compounded in March because is one thing and one trend is citations across the LLMs will stay month over month. And oftentimes we'll see certain ones grow and become more authoritative for a specific prompt or topic cluster. And so it's really cool to see that this is a compounding effort versus just a one-time small hit. So that's been, been really awesome. And then on the opposite side, I know we mentioned Zola earlier. Uh, we have been helping Zola, which is a B2C wedding website brand. Um, and they're seeing very similar, uh, opportunities off after, you know, not a huge concerted effort, like small efforts are still driving big brand impact. Um, again, focused on topics that we want to win, we're already seeing, you know, 35% improvement in brand mention lift, as well as up to a 60% improvement on single prompts that are very important for them. So specifically, they're thinking about how to build out their vendor business more so. And so we can be hyper, hyper focused on what are the citations that actually drive influence across prompts or like keywords that we really want to win. Um, but most importantly, and this is something Alex hit on earlier, is we've seen it actually start to change the sentiment. So back to how folks are talking about the brand and how we position across third-party channels, um, is actually changing that sentiment from negative neutral to more positive, um, which is a huge, huge note and early learning.

Yeah, there's a question on building pages specifically for AI. So just the first thing is I wouldn't build pages. That are solely for AI, that like things like LLMs.txt, unless you're seeing a lot of discovery driven by true agents, meaning like Cursor or Lovable, things that can do multi, um, can do multi-level navigation, it doesn't really make sense to do any of the kind of LLMs.txt type stuff because, um, those don't get favored by Google Search, which is the primary, um, a retrieval intermediary for fetching content. So don't do that unless you actually have true agents that are driving discovery. So if you're a developer tool, yes. And, and, and if you're not, I wouldn't, I wouldn't do that for now.

So Zola shouldn't do that.

Zola, I don't think people, I mean, maybe some, maybe some, uh, someone is using Cursor to plan their wedding. I don't know. You never know. Um, but, uh, I think in terms of like thinking about, um, building for agents specifically, I'm not a big fan of doing markdown. Versions of the page, you actually lose a lot of page fidelity. And the conversion of HTML to Markdown for the models is like one of the most optimized pieces of tech in the world. Like, that is very efficient. And I actually think you have fidelity loss and quality loss if you just automatically convert things to Markdown. And Cloudflare offers to do this, but I don't— I personally don't think there's value there. And we've heard that from multiple people at OpenAI and Perplexity. I think for the page itself though, there is a need for deeper content, and that content can be compared more for citation. So headings matter a lot. We have a lot of research on this, but headings matter a lot. Sentence complexity matters a lot. Directly answering the question, reducing fluff. FAQ content can help a lot. And then other page hygiene like schema, which we're not gonna get into schema on this call. It seems to split the audience, but my POV is do it. And then things like, you know, general technical SEO. So, you know, the story of this world is very similar to the story of SEO, but the emphasis on certain things just shifts just because of model preference and things like the model's ability to consume more content. But, but, but many of these things will be familiar to folks in these schools.

Schema databases, multi-phase.

Okay. So it divides the audience. But we spoke to Bing directly. They said do it because they use it for retrieval. So I'm like that. Do the schema. We can all save ourselves some time.

So as you start to think about actually operationalizing this and like where to get started, I think first and foremost, most, uh, start measuring the right, the right data. Um, a lot of this data, you know, you were not measuring before in a traditional SEO sense. Um, so what are things we recommend? Mention rate. How often is your brand, um, appearing in the AI answers for specific prompts and topics? Um, share of voice. How often is your brand mentioned when a brand is, is mentioned? Because not all answers are going to bring up specific brands and in the, in the actual text response. And so those two combined and monitoring how they're trending are obviously what we're driving here that will then lead to hopefully page traffic, organic conversion, and so forth. Beyond that, really understanding where are you first, where are you third, how do you compare against your competitors, because that is a really good way to measure in the space of, you know, with answer engines. Again, to this piece on like share of voice, not every time are we going to be talking about brands. And not every time do you get the link specifically back to your competitor. So it's really like, how do you compare in the space? And how do we beat out the folks that you deem directly competitive? And and also gain share in certain areas that you might be behind or keep your lead in areas where you're leading by quite a bit. A question just came in. How are you measuring non-deterministic answers for a categorical mention of your brand's position?

We collect, we have a sampling frequency, which is quite low by default, but we allow you to scale up the sampling frequency so you can get a a more, a more static sample size for that. So this is a, this is a big problem specifically in like lower, like longer tail or yeah, longer tail questions in particular. So yeah, you can, you can dial up the sampling frequency and we do that for priority topic clusters. Oh, there's also the other dimension of this is like persona and geo as well. So like, yeah, so you can imagine persona, geo, long tail, sampling frequency. There are a few dials here to play with and think about the and this is why prioritization is important because you like the volume, the volume gets big real quick. And so dialing in where you want to move the needle in any given month or quarter, making sure we have really good data sufficiency there and then moving on is the, is one way to do it. Love these questions. By the way, these are the best questions I've had on a webinar in a while. So you guys are crushing it.

Oh, that's so great.

Yeah, I love this.

We love this group.

Love this group.

Well, and it's a, it's a great part. It's like there's so much you can go after. So it's, it's, you know, our recommendation as you think about this as an effort is really stay focused to start. Um, pick a specific topic that you, you know, maybe two topics. One where you're already winning, um, but your competitors continue to, to gain share. They are starting to be mentioned more and more. Um, so you want to defend your position. And then pick a topic where you're not winning. Um, and start to invest in going after, you know, third-party publishers and sites that could be potential collaboration opportunities. Um, you know, once you land those mentions, uh, start to measure how you compare over time. And I think you'll be shocked at the impact right away that can build the business case to then come back to your team and say, hey, this is actually a place we should start investing, and we should get the entire team, whether that's leadership PR, whether it lives in your SEO side of the house, aligned to go after this. One thing that's been interesting that we've run into, and Alex would love to get your thoughts here, is like, because this is a different motion than backlinking and touches on some elements of like digital PR, and you know, we talk a lot about the narrative and the storytelling, we've seen that, you know, some teams don't know where this should live. Do you have any thoughts on like where this best fits in within an organization?

Typically, like these things get a lot of momentum, to Ali's point, when they start getting wins. And so I wouldn't take the first swing at this as being the ongoing operating model. So like just figure out how to get your first test launch, your first, and maybe 2 or 3 tests actually simultaneously, a few shots on goal, get a win. And then I think turning it into more of an ongoing motion, I think is the right way to go about it. But it is somebody that understands the brand and has good taste and understands the voice of the brand, but also understands systems and as well is a great owner for things like this. So really, you know, a lot of traditional SEO folks who've done strategies that are in this universe are great owners for this.

Yeah, it's interesting. We've seen teams say, no, this is definitely PR. No, this is definitely SEO. So I think it will shake out over the next year. To understand, we have to rethink how the team is structured, and everybody's crossing each other at this point to build the right story. So I think the biggest takeaway right now is now is the time to move. There is a really unique time in the market where we're seeing ways to quick wins. There is a playbook of working directly with publishers to collaborate and actually seeing visible increases. And the folks that adopt this now and fast will win in the, in the market and really build their foundation across AI search. And so this is our encouragement to all of you on this call to go do a test and experiment, run this with your teams, see if you can see an impact. And hopefully invest in this longer term as well. I see a good question come in. Can you connect citation rate to conversions beyond directional correlation?

Yeah, I think that the problem is that there's, there's, there's all of the attribution data is imperfect by itself. I think what we really try and do is first structure things in a way that gives us a good sense of incremental lift from a visibility and traffic standpoint. So you know, thinking about splitting pages or topics into two and performing activities on one and then looking at the inbound lift, the traffic lift. But it's not, there's no single data source, unfortunately, that gives you the whole picture. And so part of what we're working on and releasing soon is a connection through to conversions so we can combine both owned and inferred data to get to get the full picture. But for now, it's about structuring the experiments in the right way, measuring all of the changes that are being made in the offsite placements that are being done, and then being sure that you do full, like, through signup processes and conversion experiences, having the implicit attribution, but also asking people where they heard about you.

So another good question. How is mention rate calculated?

It's of responses that contain one or more brands. What percentage of those responses contain your brand?

And I would be remiss not to talk about what we can help you with. So let's say you did an experiment on your team and you've realized that this is a channel you all want to invest in. AirOps specifically now offers an offsite offering where we really help you, you know, first and foremost understand what does the universe impacting your brand look like when it comes to third-party websites. Um, you know, some brands that means it's 30% Reddit, some brands Reddit's less than 2%, um, some brands it's all these listicle sites and affiliate sites and so forth. So just really understanding who's, who makes up the universe and how can we go after them. And then we actually can help you run the process end to end. So understanding, you know, outreach to publishers, content development that should fit your tone and voice, your brand, understand your value props and so forth to ensure that that is what is being included in these articles, all the way to landing placements and so forth. And then the most important piece of this all is is actually showing the impact. So how is all this effort that we are putting in driving visibility and lift and hopefully overall organic traffic? So if that is something you're interested in, and please, would love to have a conversation. I'm gonna skip to the end. Would love to have a conversation if you're early in your journey and just figuring out how we set this up for your team to run an experiment, our team is always happy to, to set up time and chat through that as well. This is a really new, exciting time, and we're here to help you all win. Alex, any final thoughts?

No, and like, we're all, we're all kind of learning as well, and would love folks' thoughts on any of the topics covered, but also the other things that cropped up, like how you guys want to work MCP, AROPS platform, like my DMs are open. I love hearing from people. You guys are great. And we, we really do build what we, what we get asked to build in many cases, and you're the high-touch users that we are constantly thinking about. So talk to us about Offsite and other things. We are, we always love hearing from you.

Thank you all for such a great chat and all of the wonderful questions. Like Alex said, our emails and DMs are always open, so please reach out.

Appreciate you guys. Thanks, Ali.

Thanks all. Bye. Bye.