The Hidden Algorithm That Decides Which Software AI Will Recommend | Tim Sanders, G2

Most companies investing in AI visibility are optimizing for the wrong thing. Being cited by an AI response and being recommended by an AI response are completely different outcomes, with click-through rates that differ by a factor of 70. Tim Sanders, Chief Innovation Officer of G2 and executive fellow at Harvard's AI Institute, joins Craig Smith to explain the hidden mechanics behind how ChatGPT and Gemini actually decide which software to recommend, and why the answer has almost nothing to do with what's on your website. When a user asks a commercial intent question, both models enter a "validation layer" process that specifically down-weights vendor content and seeks verified third-party signals: appearance on authoritative lists (41% of the recommendation weight), awards and accreditation (18%), and online reviews (16%). None of these can be manufactured. They must be earned. The conversation covers the structural transformation in how B2B software buyers behave, half now start their search with an AI prompt, up from 29% a year ago, with two-thirds projected within a year, and why most companies' organic search traffic is on a structural path toward zero. Sanders also delivers some of the most specific competitive data on AI model usage available anywhere: ChatGPT does live retrieval 100% of the time on research queries; Claude does it less than 40% of the time. ChatGPT and Gemini account for 81% of G2's AI research citations. And more than one in four enterprise employees bypass corporate AI tools entirely, using their personal ChatGPT because it has their memory and none of the guardrails. The episode closes with Sanders' most forward-looking prediction: within three years, AI agents will write the prompts, locate the software, and purchase it on behalf of businesses - with minimal human checkpoints - making the trust infrastructure G2 has built over more than a decade the most valuable asset in the AI-driven buying cycle. Subscribe to Eye on A. I. for weekly conversations with the people building and deploying the future of AI.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Craig S. Smith?

  For discovery, I think most people, I certainly have, have switched to asking one of the AI models before they go any further.

• 0:09Speaker 01

  You do that instead of going to Google and using keywords. You represent at least half the market according to our latest research. Like half of all B two B software buyers do it exactly how you do it, Craig. A lot is like the Oscars. Everybody talks about the movie. Not many people see the movie.

• 0:23Craig S. Smith?

  Certainly, there is an industry growing up around filling the training data with positive information about your company. So that when you're asked, when the model is asked, it'll come up with your, how big of a problem is that for you guys?

• 0:38Speaker 01

  When it comes to the purchase recommendation, ChatGPT and Gemini enter into a process called validation layer work at inference. What does that mean? They are validating that the entity that they're about to recommend is not going to lead to a regrettable purchase. I look in the future, Craig, three years from now, you express business goals and business problems. And the agent then will write the prompts and locate the software and perhaps buy it on your behalf with maybe a couple of checkpoints or guardrails. Um, I see that coming. Hi, I'm Tim Sanders. I came to G two in the fall of two thousand and twenty four. Today I serve as the company's chief innovation officer.

• 1:21Craig S. Smith?

  And you have a pretty interesting background. You are also part of the AI Institute at Harvard. Uh, and then you, you also have written a number of books, one, Love is the Killer App, which I've found fascinating. Can you just talk about those two aspects of your background before we get into G two?

• 1:50Speaker 01

  Yeah, glad to. Last three years, I've served as an executive fellow at the AI Institute at Harvard. And the charter of the institute is to democratize, democratize artificial intelligence for everyone, especially business leaders. Um, and they do that, um, with case level research. They do that with publications. They run executive outreach programs for a lot of companies you've heard of. To really help people understand the simplicity of AI and more important, how it can be leveraged for business outcomes. Um, Love is the Killer App. So that was my first book. I published it years ago when I was an executive at Yahoo.

• 2:29Speaker 01

  And the thesis of the book that is, um, that in a very high tech world, it's never been more important to be high touch. And you know, Craig, I wrote that with the advent of the internet and digitization, digital transformation. And here we are in two thousand and twenty six, high tech, AI is as high tech as tech gets. And high touch leadership and judgment and taste and all those other great human attributes around emotional intelligence have never been more important.

• 3:00Craig S. Smith?

  Yeah, yeah. And how did those two, your work at Harvard and, uh, the philosophy behind this book? Which is very generally, uh, that, that business should, should be conducted with, uh, empathy and compassion. Rather than sort of cutthroat, zero sum, uh, point of view. How do those, uh, help what you're doing at, uh, relate to what you're doing at G two?

• 3:33Speaker 01

  Love is the Killer App had, had this idea that when we talk about love in our business context, what it really means is the selfless promotion of the growth of the other. So, so when leaders show love to their teammates, they're growing them. When they show that customer love, they're promoting the success of the customer. And, and I also like that the way that we do this promotion to scale is by sharing the three intangibles every human, every organization has that actually grow as you give them away. Your knowledge, your network of relationships and your compassion. What appealed to me, uh, with Harvard's AI Institute is their charter. They wanted to share knowledge to promote the success of businesses.

• 4:16Speaker 01

  In light of this opaque thing called artificial intelligence, which is keeping too many people on the sidelines waiting, uh, not trusting it yet. I study trust a lot now. Um, I felt like there was just a real synergy between my longstanding belief around knowledge sharing is love. And the AI Institute's belief that we're going to share knowledge to democratize the most powerful technology in the history of civilization.

• 4:41Craig S. Smith?

  Okay, uh, and G two, I mean, its mission is to be the, uh, the arbiter. Is that a way to describe it of, uh, software, uh, particularly, uh, software as a service?

• 4:59Speaker 01

  Yeah.

• 4:59Craig S. Smith?

  Uh, so that, uh, you have, um, an impartial, uh, view of, of all of the software in a particular, uh, domain or for a particular task. Can you talk about that? How G two evolved what it is today?

• 5:19Speaker 01

  You know, G two was founded, um, around the idea of creating the highest form of intelligence for buyers of software. Um, G two stands for the highest form of military intelligence, for example. Um, the original name of the company was G two Crowd. The belief was the crowd, those people writing reviews. The community, the peers are the best way to generate the highest form of intelligence. At the time, our founders were struck by the, if you will, pay to play model of analyst firms like Gartner. And how difficult it was for a software company to establish trust with the buyer and actually become successful as they added more and more capabilities.

• 6:02Speaker 01

  So that was the driving motivation. And today, when you look at how the agents at Frontier Labs, like ChatGPT, so OpenAI, ChatGPT. And Google's Gemini, when you think about their logic today, um, they absolutely love human natural language as expressed in reviews. They love the idea that a company like G two has reviews to scale. And as a result, uh, we've become the dominant player, um, in driving AI search results when you're trying to buy software. We're often going to be the most cited third party, uh, when recommending the answer, which just illustrates over more than a decade of work, how far G two's come.

• 6:45Speaker 01

  I would say though, Craig, that, you know, we had the founding mission to provide this, this highest level of intelligence for the buyer. And let me tell you where we are today. I hear a lot about this in our meetings. G two's new mission is to become the trust layer for the age of artificial intelligence. And that's where we are right now. We want to help buyers and owners and users establish trust in artificial intelligence. Especially as agents and autonomous computing, um, come online and offer people a breakthrough in business results. Um, not thirty percent, not forty percent, but what we're hearing now, two X, five X, ten X in certain situations.

• 7:28Speaker 01

  Only if they trust it though, and trust is a hard thing.

• 7:32Craig S. Smith?

  Yeah, um, and, and the, uh, the, the reviews, are these reviews that are written exclusively on G two's platform? Or do you, uh, collate them from

• 7:45Speaker 01

  Nope.

• 7:45Craig S. Smith?

  Different sites around the world?

• 7:48Speaker 01

  They're written exclusively on G two's platform. Um, some are typed, uh, increasingly, uh, others are just spoken. We have a voice product now that actually leads to longer reviews. Um, we, I would say over half of what is submitted on the platform doesn't make it to the website. Because our trust and safety, um, approach is to not only verify the identity of who's writing the review. By the way, the agents love that for establishing trust. We also verify that you own the software and that creates a lot of friction, but at the same time, it differentiates us from others, uh, uh, from communities like Reddit. Um, or other review sites that don't go to that rigor in the spirit of trying to aggregate the highest quantity of reviews.

• 8:33Speaker 01

  We could have twice as many reviews as we have, um, but we feel like we have enough.

• 8:37Craig S. Smith?

  Yeah, uh, although it would be interesting to cross-reference your, uh, your reviews. I mean, in aggregate with reviews on, on Reddit or another platform to see if they

• 8:54Speaker 01

  I, I love that idea. It's something that I've been thinking about for a few years. I believe that especially for buzzy, hot, new things like a, a launch of a new language model. I, I truly believe there are, um, thousands, if not tens of thousands of unstructured reviews living in communities like Reddit. That really could be cross-indexed, AI sourced. So certainly I wouldn't be surprised, uh, to see if something like that would come online to even give more signals. Um, because what I would say is that, yes, Reddit is anonymous, but they have a wonderful redditor system, if you will, of uploading and other such things that can separate the quality from the spam.

• 9:31Speaker 01

  So I do believe there are high quality signals. Uh, by the way, Reddit is top three B two B software citations when you're looking to buy software. It's G two, YouTube, Reddit. That's top three right now.

• 9:44Craig S. Smith?

  Wow. Uh, in G two, you have over three million verified reviews now.

• 9:51Speaker 01

  That's right. That's right.

• 9:52Craig S. Smith?

  Ninety million, million annual visitors and, uh, two hundred million software buyers globally.

• 10:00Speaker 01

  We've added a little bit with the acquisition of, uh, Gartner Digital Markets, which involved the brands, uh, Capterra, Software Advice and GetApp.

• 10:10Craig S. Smith?

  Yeah. What are those three? And what is your relationship with Gartner? I mean, do they, uh, Use your reviews? Do they refer to your reviews?

• 10:22Speaker 01

  Nope. We, uh, we just made an acquisition of three of their properties, which were rolled up in a, in, in, in a property called Gartner Digital Markets. And we completed the acquisition February of 2026. Gartner continues to be an analyst firm. They still own the review site, Peer Insights, which their analysts use. But we bought the other elements of that business. Capterra's leading software review site, probably top three or top four prior to the acquisition, depending on how you want to count it. Software advice, fascinating service. If you're buying software, you can call a 1-800 number and have a conversation with an expert that will help configure for you the exact solution you're looking for. We think that's a, a great layer of intelligence to feed trust.

• 11:07Speaker 01

  And then GitApp really focuses on startups, very small businesses in terms of how they buy software to fuel their companies. And it also promotes startups and small businesses that are just coming online, trying to compete with the incumbents.

• 11:21Craig S. Smith?

  Yeah. And what's the business model? I mean, are you guys a subscription model? I mean, for example, GitApp, that's, I'm a small business and I'm always trying to figure out,

• 11:34Craig S. Smith?

  you know, what software I should be looking at. Uh, is it pay for, what's the term, you know, usage?

• 11:46Craig S. Smith?

  Payment or is it subscription?

• 11:47Speaker 01

  It's subscription. Um, and we can talk later on. I think the world is going to move to pay for consumption. But let's just talk about G2's business model, marketing services. So G2's business model is we provide a variety of marketing services. We provide platform access. You subscribe by claiming your profiles on G2 for your products. That gives you tools to generate higher quality reviews, have a higher quality presence on G2, which is often what cited your profile on G2. We offer review as a managed service to help companies augment what reviews they can naturally collect. So they can scale that voice. We offer buyer intent data, not only like who's looking at your products and services,

• 12:31Speaker 01

  but more interesting like which competitors are looking at, you know, their, their customers are looking at your products and services. So that, that data intelligence now available via MCP. And then finally, advertising. So we sell media opportunities for software brands to extend their reach in this evaluation stage of the buyer journey. And, and I would say that when you think about Google Ads, which is like the, the massive, massive, you know, business in advertising, that's more discovery, right? So you start on Google, but you end up at a review site, analyst firm reports, publishers, experts like yourself. That's called the evaluation stage.

• 13:11Speaker 01

  So we offer advertising solutions for companies that say, when that customer has that shortlist in hand, they're trying to evaluate. We want to show up there either to be picked off the shortlist or to disrupt the process and enter the conversation, even if we didn't make a ChatGPT shortlist.

• 13:28Craig S. Smith?

  These days, you said that G2 is, uh, G2, YouTube and Reddit are the, the top three places that people go for

• 13:38Craig S. Smith?

  to, to analyze software or understand users.

• 13:43Speaker 01

  They're the most cited. I think the way to think about that is if you are putting out a commercial intent prompt, and if you like, I'll share something with you. Um, if you're putting out a commercial intent prompt and you're saying, you know, I'm looking for a CRM software where I want it to be a medium-sized hospital, month-to-month payments, and it works great on mobile phones, then whatever the shortlist is it gives you, it will cite various sources that it relied on to make that recommendation for those three CRMs. That's what we track at G2. So we're currently number one by a margin. Number two, depending on the month, is going to be Reddit or YouTube as being the cited source behind the recommendation.

• 14:28Craig S. Smith?

  Right.

• 14:28Speaker 01

  So we track our influence since our business is marketing services.

• 14:32Craig S. Smith?

  For discovery, uh, I think most people, I certainly have, have switched to asking one of the AI models before they, uh, they, they go any further. You know, what are the top software for, for this particular use case?

• 14:55Speaker 01

  You do that instead of going to Google and using keywords. So you're writing a natural language prompt. You represent at least half the market, according to our latest research. Like half of all B2B software buyers do it exactly how you do it, Craig. This time last year, it was only 29%. If I'm looking into the tea leaves next year, probably two thirds of the people are going to behave like you do, starting in chat, maybe even generating the shortlist out of chat and then doing the evaluation work.

• 15:22Craig S. Smith?

  Yeah. And something that I've wondered, uh, and certainly there is an industry growing up around, filling the training data with positive, uh, information about your company. So that when you're asked, when, when the model is asked, it will come up with your, how, how big of a problem is that for you guys? Because I don't, and how, what is the mechanism for doing that?

• 15:51Speaker 01

  So, yeah, let's talk a little bit about that. So it's a huge AEO, answer engine optimization, huge industry. But I want to make a distinction here. The purchase recommendation use case is the only use case G2 is concerned with because we sell marketing services to help software companies drive purchase, right? There's many use cases on ChatGPT. You could do fact lookup, you could do document summary, you could be tool use. But purchase recommendation represents a unique risk to the user. OpenAI researchers call this your money, your life. So when it comes to things like medical advice or purchase recommendations, OpenAI, and for that matter, Google Gemini, they don't just rely on pattern recognition, matching the natural language in the prompt to natural language on a website.

• 16:36Speaker 01

  That's what many companies are retooling to do, is to have all that language on their website to win the pattern match. You may win in a variety of use cases, but when it comes to the purchase recommendation, both of the models I mentioned, ChatGPT and Gemini, enter into a process called validation layer work at inference. What does that mean? They are validating that the entity that they're about to recommend is not going to lead to a regrettable purchase. How do they do that? They look for trust signals from high authority sources, usually with strong underlying data. One thing I would share is there's a fantastic study out there by a company called First Page Sage.

• 17:17Speaker 01

  And First Page Sage, they broke down that algorithm for how the agent at ChatGPT is thinking at inference time about validation. And so as you can see here, 41% of the weights that drive a software company being recommended is their appearance and quite frankly, their ranking on authoritative lists. Another 18% is driven by awards accreditation, 16% by online reviews. All of those are third party. They can't be generated by a vendor. They must be acquired. Now, what are the authoritative lists? They could be review site lists like G2's Best of Software Awards. Last year, uh, we had an, we did an audit for where our citations came from.

• 18:01Speaker 01

  60% of our citations came from the Best of Software Awards and all the Best of Category pages across G2. Okay? Other sources could be authoritative lists published by analyst firms, Gartner, Forrester, right? If they're, if they're released to AI for crawling and then publishers, say TechCrunch, top 10 CRMs for hospitals. But again, the models are distinguishing verified identity of the user with sentiment expression. They're also looking for just like the number of people involved in the generation of that authority list. This is different than listicles. So let me come back to what you asked. Is there an industry helping software vendors create content to show up and have AI visibility?

• 18:46Speaker 01

  Absolutely. But AI visibility is not the same thing as winning the answer. Being cited means a link to your content is showing up in an AI return. In the case of that CRM prompt I gave you, the medical month-to-month, works great on iPhone one, that return from AI isn't just the three CRMs, it's several pages. Really breaking down CRM and hospital, the rise of it, mobile endpoints, different vendors that offer Android versus iOS, different pricing matrix, is the advantage of month-to-month versus a flat fee, for instance. And then at the end of that three or four page return, it makes three recommendations.

• 19:28Speaker 01

  You can use one of these services to generate all of your content to show up in the first three pages, get a little traffic from it. But that doesn't mean you're going to win the answer. So I always tell people being cited doesn't mean you are recommended. Those are different things. Being cited is how you generate traffic to your website, which hopefully you can convert. Winning the answer generates pipeline, and that's the business we're in at G2.

• 19:53Craig S. Smith?

  Yeah.

• 19:54Speaker 01

  And you can't do that with first party content. You can't win the answer with vendor claims.

• 19:59Craig S. Smith?

  Right. And, and, and you're talking about this evaluation layer, is that how you refer?

• 20:04Speaker 01

  Yeah, they call it the validation layer. Yeah, it's fascinating. And it's driven around like the regretability index. Like I'm, I'm trying to look for signals that predict, because that's what these machines do. They make predictions. I'm trying to predict that if Craig buys this CRM solution, he's going to renew in a year because if you buy it and you hate it, that leads to platform health decline. And that regretability index has been around since social media and regrettable minutes years ago.

• 20:30Craig S. Smith?

  Oh, and so that, that is something that the model builders are deliberately building into their models.

• 20:39Speaker 01

  Yeah, well, for, they're, they're building it into fine tuning for various use cases based on prompt categories. So I, I said, purchase recommendation. What triggers a model then to look for fine tuning related to purchase recommendations called a commercial intent prompt? It's where the natural language expresses the desire to purchase something. And that commercial intent then triggers the model to say, oops, we need to do validation layer work, your money, your life. Craig, it's the same if I ask medical advice. If I say, I want you to look at this picture and tell me if it's skin cancer rates. So I was, uh-oh, your money, your life. We need to go start to really identify high authority sources to gain confidence to tell you. And generally speaking, in a situation like that, it's going to be very iffy anyway.

• 21:24Speaker 01

  I mean, it's not going to be very middle. But anyway, yes, that, that, that's, that's, that's what we've learned over the last few years. That's just fascinating.

• 21:33Craig S. Smith?

  And, and you mentioned OpenAI and Google or Gemini.

• 21:37Speaker 01

  Those are the ones.

• 21:39Craig S. Smith?

  What about Claude, which increasingly is a fall?

• 21:44Speaker 01

  Well, we could talk about this. Claude is being adopted by a lot of companies. That doesn't mean their employees use it to research software. G2, we studied this, like, like we studied it every year. Like, what, what do you use the most when it comes to work and doing research for software? The answer is Claude is like the Oscars. Everybody talks about the movie. Not many people see the movie. Claude is being used for coding. It's being used for summarizing document slack conversation. It's used to automate various tasks that are usually weekly. Those are called cron jobs. And most popular these days, Claude is being used to create presentation decks, but it's not being used to do primary research. The primary research, and if you look here, the market shares 81% is ChatGPT and Gemini.

• 22:28Speaker 01

  I would also say Claude has the lowest net promoter score for the research use case of any of the ones I just showed you, including perplexity, because it's just not a priority. Labs devote tokens at inference to the things it thinks it's going to win by. And for Claude, that's coding and increasingly it's design. Their reasoning is mid. Their reasoning to me is not much better than deep sea.

• 22:52Craig S. Smith?

  Right. So they don't have this validation layer or validation.

• 22:57Speaker 01

  They may, they may, but they don't, they don't call upon the breadth of data that say OpenAI and Google do. Um, so for example, um, profound, a leading AEO researcher, Josh Blyss call revealed that less than 40% of the time when you're on Claude doing a research call, it does live retrieval, like less than 40% of the time. When I'm on OpenAI, a hundred percent of the time it does live retrieval. It changes the results. So even if the algorithm's the same, I just think that the users are seeing more performance for research. The other thing I'll just quickly say here too, Craig, is that a shocking number, more than one out of four B2B software buyers are using their own ChatGPT, their own phone, because it's got their memory.

• 23:42Speaker 01

  They don't want to use the corporate one with all the guardrails. They're very good at using it, et cetera, et cetera. Um, and, and then it's really the ChatGPT Gemini show, especially ChatGPT when we're talking about that.

• 23:56Craig S. Smith?

  Yeah. And so if you ask, uh, ChatGPT 5.6 or whatever the, the top model is.

• 24:06Speaker 01

  Yeah.

• 24:07Craig S. Smith?

  Uh, for a recommendation on a, on a particular software. And you ask Claude, are you likely to get different answers?

• 24:14Speaker 01

  Yes. Well, you're likely to get different answers from OpenAI if you ask it today and ask it tomorrow. That's why that slide I showed you that tracks G2 citations, it's not based on one prompt. We're profound's biggest customer. They run hundreds of thousands, if not millions of synthetic prompts every month to get that statistical body that generates real directionality. Like when you hire an AEO firm to build the dashboard for AI visibility, what they do, Craig, is they work with you to set up something called a prompt panel. Think of it as questions you're tracking. And then what they do with their software is take your questions and what they call fan them out to versions of your questions.

• 24:58Speaker 01

  So they can run thousands and thousands or hundreds of thousands, if you can afford it, synthetic prompts to generate synthetic returns. And by averaging them all, they can give you pretty good analysis of AI visibility. On the other hand, if you just went to ChatGPT and put the prompt in to see if your company shows up, it's going to vary every time you do it. But if you did it enough, you would reach a statistical average and have directionality.

• 25:23Craig S. Smith?

  Yeah. And, and again, just, uh, and because I want to get beyond this, but what are they doing? I mean, that's a good, uh, explanation of, of how they're, uh, seeing what the, uh, what the average is or, uh, across multiple sessions. But what are they doing to, to get a company, their client into the fine tune data so that, uh, they show up?

• 25:55Speaker 01

  Most of the companies are hiring AEO firms to recover the traffic they've lose because of the zero click tendency of users like yourself. So again, G2's laser focused on pipeline, like showing up on more short lists, winning more shootouts. That's where we play. Um, a lot of the AEO firms, they have more of a monolithic view of AEO, like companies need to get back their traffic. So if you're cited frequently, then theoretically you get clicked on. Now, the disappointing part here. Uh, Promptwatch has done some research around click through rates, Craig, and the average click through rate for the average citation and a return is somewhere between one tenth of one percent, one third of one percent.

• 26:36Craig S. Smith?

  Yeah.

• 26:36Speaker 01

  Now, if you're cited in relationship to a recommendation, Promptwatch found it's one to seven percent. It's significantly higher. That's why G2 got a million human visitors from OpenAI user bot just last year alone. My point is the AEO folks can help you create the type of content and the schema structure of your website that makes it very easy for the machines to consume where you have visibility, you're being mentioned, your content is being cited. And there's ever the chance that you'll see some traffic come back. Now, what we do know is the traffic you do get from chatbots is higher value traffic, because usually by the time they click through to you, they have more conviction.

• 27:19Speaker 01

  Um, they were coming based to do evaluation from a shortlist, um, or they're just further along the buyer journey. So, so the traffic that you do get is much higher quality. Some might say as high as two, three X. Um, and so I work with a lot of companies now that are beginning to partition off, like here's the traffic we're getting from the machines and it's very valuable. Here's the traffic we're getting from humans. But the reality is most companies are realizing organic traffic is going to start getting close to zero as time goes by, because we get our answers in the chat and there's no read to poke around. The one thing that I would say quickly though, the reason that the click through rate is highest, the deeper you get into a report where the recommendations live,

• 28:04Speaker 01

  and it's the lowest at the beginning of a chat report where all the explanations live, is because in 2026, we have this like, this trust, if you will, in chatbot results. So now we're at a point where we only click on what is risky to believe on its face. Your money, your life. So that's why when you see a citation next to a recommended purchase, you are the most likely to check on that and evaluate.

• 28:31Craig S. Smith?

  Yeah. And, and just again, that these AEO, uh, firms that are trying to boost your visibility in, in chatbot responses.

• 28:42Speaker 01

  Yeah.

• 28:43Craig S. Smith?

  There are, are they, and I understand they're creating relevant content. Uh, but then are they like building a million websites and spreading it around the internet? I mean, so that the, the crawler will, or the, uh, internal, uh, search uh, of the chatbot finds some, I mean, you know, just optimizing the, uh, content on your website.

• 29:10Craig S. Smith?

  May not do much if your website doesn't get a lot of

• 29:14Speaker 01

  foundational. Yeah, it's foundational, though. If if if your websites are optimized to be easy to crawl in training or easy to retrieve at inference, um, it's kind of game over. So, so, so there's companies, scrunch is a great example. They, they were just acquired by Sitecore. Their agents can create a website remediation, meaning literally spinning up new code for you to improve the, the machine readability across your site. Uh, different, well, different A E L vendors have different specialties. Air Ops, um, they're great at content engineering. They can have agents identify um, content gaps across your site where there's a market demand for a certain type of content you're not providing. You provide that content. They call it frontier content if you're first on the scene.

• 29:58Speaker 01

  And you're going to see an outsized number of citations against that content. I mentioned profound. They have general purpose marketing agents. Profound can do everything from doing constant reconnaissance and daily reporting to you on how your visibility may or may not be moving. They can generate either suggestions or actual content artifacts for you. They can also provide remediation of schema. I'm seeing a lot of the A E O companies really moving from just dashboards two years ago to always on agents today to help companies kind of catch up in this world of A E O, at least from a traffic generation standpoint.

• 30:32Craig S. Smith?

  Yeah, um, it's, it's, it's fascinating. And that really is

• 30:37Speaker 01

  I haven't seen anything like it since I was C S O at Yahoo twenty five years ago. Yogi Berra said it best. It's deja vu all over again.

• 30:45Craig S. Smith?

  Yeah, uh, the, you recently introduced, uh, you know, M C P capabilities, uh, a model context protocol which allows, uh, models to interact with your uh, software, uh, that let A I systems reference G two's verified reviews. Uh, tell us how that works and is that, uh, who turns that on?

• 31:15Speaker 01

  Well, M C P is a company-wide initiative. We're really taking advantage of it inside the company and how we retrieve information for the purpose of our own research and automation. We also offer M C P connectors to our customers. We're seeing an incredible take up of customers who are being able to access, I mentioned before, buyer intent data. They're able to access, um, buyer intent data headless now, um, where it automatically fits within their workflow. But we are also generating partnerships with Frontier Labs to set up M C Ps and we develop those connectors, if you will, those M C P servers. So that it's easier and frictionless for the agents to consume our verified reviews at inference because the thing Craig, when I keep talking about, like, you've got to be easy to work with.

• 32:00Speaker 01

  These frontier labs are, are compute constrained.

• 32:05Speaker 01

  And so, so they will rely on what is easiest to consume and not just compute constrained. They are discerning based on how much the user is paying them for inference. So I mentioned before, if more than one out of four users are using their own chat G P T, at best, it's a twenty dollar plan. That's buffet food. So, so, so whatever content they're getting better be just like super, super easy to consume. And that's why M C P, which we consider the universal U S B, uh, to connect data with any system or platform. We think it's really important. That being said, we continue to keep our eye on other connector solutions

• 32:47Speaker 01

  like command line interface, C L I. Uh, there are certain use cases where C L I's have very elegant solution. Not to get in the weeds here, but, um, researchers would tell you there is something called the tool tax that's involved with M C P. So when you use an M C P server and when you call it, it opens up every tool in the shed. Uh, whereas C L I is only going to open up the tool that's needed and called for at the time. However, again, M C P today, twenty twenty six, more general purpose, many, many more use cases and C L I. But both of them are a lot better than having to write custom A P I's every time. That's for sure.

• 33:22Craig S. Smith?

  Yeah. And, uh, does, so I'm a, a C T O, uh, at a major company and I want to research, uh, software. Do I ask the model to check G two? And then the model finds the M C P connection?

• 33:43Speaker 01

  Correct. That's how that works. And if you're a software seller, then the way you would use M C P at G two is to take a look at movement inside categories, whether your customers are looking at competitor reviews or whether your, uh, prospects in pipeline, those companies are looking at your reviews. You do all of that headless with M C P. So it'd fall right into your workflows, whether it's your system or record like Salesforce or, or whatever you're using.

• 34:07Craig S. Smith?

  Right, but you have to

• 34:09Speaker 01

  Our HubSpot's a great example where we have actual connectors directly into HubSpot for that marketing outreach, based on that intelligence that you're going to get through M C P.

• 34:19Craig S. Smith?

  But you have to explicitly ask the model or, or if it's, uh, if you ask, uh,

• 34:28Speaker 01

  Unless you're an agent.

• 34:30Craig S. Smith?

  Well, that's what I was gonna say.

• 34:31Speaker 01

  Like, you're a Claude coworker now, open A I work. You write an agent that just does that every week and creates like a downstream set of other actions. So you might have an agent, you know, uh, in cowork that uses the M C P to go to G two and gather data and then trigger something in Clay. We're now a Clay agent and might create a content cadence that might trigger, uh, an agent force agent to start Slack channel, if you will, to implement sales motions, alerting various account managers of changes in account. So I could see that. You know, we're, we're already seeing that, that customers are taking advantage of agentic workflows to really like get the real value out of M C P. Instead of it being like, we're the bottleneck, we're the bottleneck that has to actually write the prompts or build the instructions every single time.

• 35:16Craig S. Smith?

  But again, if you're working with an agentic, uh, framework or agentic system and you say, find the top software for X use case.

• 35:28Speaker 01

  Yeah, today that's still very manual.

• 35:30Craig S. Smith?

  Okay, it doesn't

• 35:31Speaker 01

  That is great manual today. So I think the sellers are using more agents than the buyers. We just did research on this. We think that the way that buyers currently use agents, I say that word loosely. I could talk to you about this, this gradient of what I consider agents from chat practice system of agents. But when the buyer employs agents, most of the time they keep them on a pretty short lease. They write a prompt, the agent does work, they consider the agents work, and then they usually evaluate before they make a purchase. However, we're starting to see a trend where companies are telling us, and not a small number, more than, more than ten percent in twenty twenty six, where agents then are advising, if not taking action on routine renewals.

• 36:14Speaker 01

  So we're starting to see that agents may go in and make decisions on contract renewal or escalating for churn consideration or, um, alternative uh, vendor recommendations. And I think that's going to be the canary in the coalmine. I look in the future, Craig, with three years from now. You express business goals and business problems and the agent then will write the prompts, locate the software and perhaps buy it on your behalf with maybe a couple of checkpoints or guardrails. Um, I see that coming. It's not there yet. Um, but for renewals, it's starting to show up.

• 36:49Craig S. Smith?

  Yeah, but my question is, does the agent find the M C P tunnel?

• 36:56Speaker 01

  Yeah, it's seamless. They don't even know it. The agent's just retrieving information. It just happens to skip a step. If the M C P is connected, it just doesn't have to go to a, you know, a bundling service like a, an aggregator, like a vendor that aggregates scrape search results.

• 37:10Craig S. Smith?

  But the user doesn't have to tell the agent, check G two. Okay.

• 37:15Speaker 01

  No, not, not when they're doing a prompt to buy software. They're just saying, you know, give me the three C R Ms and they give you all the conditions and then the agent knows to just go find the best information. It just happens to favor, if you will, the easiest information to consume. It also fits kind of like the specifications of validation layer work. The M C P just kind of makes you the easiest one to do business with today in twenty twenty six. The thing I would tell you is a lot of people think they're talking to a model when they talk to chat. They're not. You're talking to a harness. And so the chat, when it says retrieving, it's not going on Google and looking something up. It's literally retrieving from a database provided by a vendor who scraped Google.

• 37:55Speaker 01

  And, and people have to understand that's a step and it comes at an incremental expense. M C P solves for that with these models, making it even more preferred uh, than a scraper vendor that helps them, uh, with a real time retrieval.

• 38:08Craig S. Smith?

  Yeah. Uh, you, you talk about G two data being rooted in authentic customer feedback and real time buyer behavior. How do you maintain that authenticity as a plat, as the platform scales? To, to what's essentially a near monopoly position in software reviews?

• 38:29Speaker 01

  Yeah, I heard that comment from Prof G. So yeah, I recognize that we have a very strong trust and safety ethos at the company we have from the day we were founded. And so we have a process which is pretty bulletproof with respect to how we verify identity as well as how we verify ownership of software. Again, we're rejecting more than half of the reviews where a person does the work or whatever and submits to make sure these aren't written by bots, they're not generated by vendors through customer conduits, they're not too overtly recruited from customers. We've had situations, say, where a vendor might do an outreach program which is too

• 39:13Speaker 01

  aggressive with respect to trying to over incentivize with the, you know, but only for good reviews. We will shut that down and we will unpublish those reviews. That vendor could be suspended. So why do we do all of it? We do all of it initially for the buyer, but what we're learning is the agent values that it gives our reviews even more value at inference than than those review websites that don't go through this rigorous process. So yeah, it's true. We probably have eighty-ish percent or more of the market share for reviews. But for us to continue to grow and to certainly reach our goals and to reach our peaks, we have to maintain that integrity, not only for the buyer, but most importantly for the

• 39:58Speaker 01

  models. Because again, models are based on weights and right now verified reviews have the most weight. If if the model begin to believe that our reviews had less weight, then we would see a de-acceleration of citations like what we saw with Reddit after the entity update. Because I mean, I'll just give you an example of something that just keeps us up and motivates us at night. August of last year, 2025, OpenAI institutes the entity update. That's where they say, we're going to recommend fewer entities, we're going to have single ways to talk about companies, and we're going to do validation layer work and really begin to reward real verified human behavior to scale.

• 40:39Speaker 01

  And then we went from like competing neck and neck with Reddit for B to B software citations. The game changed overnight and this is something that we'll never forget. We don't want this to happen to us. We want to continue to accelerate, but we're only going to accelerate when we gain the confidence and the trust of the agent at the models at test time. So we have to be vigilant about it. And it's, it's, you know, it's, Craig, I've seen review websites, not just G2 over, you know, more than a decade, this goes all the way back to the early, early review websites and then the birth of Amazon and all the other places, Google reviews, et cetera. I mean, there, there are a lot of malicious, bad actors that want to create fake reviews

• 41:21Speaker 01

  to drive business and that's just a moving target for us, but not because we have other review sites we're competing with. It's because that's our value proposition in our competition with analyst firms and publishers and UGC sites. That's what makes us different. And if we are no longer different than by the sheer weight of, you know, review sites like Reddit or LinkedIn, we're going to lose.

• 41:44Craig S. Smith?

  How do you keep the marketing services from influencing the reviews? I mean, if you're offering marketing services, if I'm a company that has not not great reviews.

• 41:56Speaker 01

  We can't really help you get better reviews. We can help you get more reviews, but you're only going to get like the reviews you deserve. Meaning like when we do as reviews as a managed service, all that means is that we can take a customer list and be more proactive on reaching out to see if they want to write a review. We can't coach it. The questions don't limit it where we say like, well, if they write a one star review, we're not going to publish it. We're going to publish every review that we capture in our review managed service contract. So right now, that's not a marketing service that's on sale. So we can't help you win on G2, but we can help you generate review recency. Think about it as a form of content. So, so that for us is just not a product that's for sale.

• 42:37Speaker 01

  I mean, what you're getting when you subscribe is access to tools such as the ability to use an app layer like Pando so that your website can prompt happy customers to write reviews and make it very seamless for them to go directly to G2 and write it. But, but we're really not in the business of helping people improve their profile and reputation. I'm seeing that there are PR firms and certain AEO vendors that claim to be in that business. We think that, you know, again, I've been doing this for 25 years, going back to like pre-Google, like AltaVista, keyword stuffing. I, I think it's a moving target. I think doing things that are cute with respect to trying to win trust are short-lived.

• 43:17Craig S. Smith?

  Yeah. And so you, you sit atop a tremendous amount of data and can see trends in the market. Your winter 2026 report shows momentum in AI-driven categories. Which ones surprised you? Which ones are leading?

• 43:39Speaker 01

  We talked a lot about answer engine optimization. It's one of the fastest growing categories because it's just so much disruption going on, right? Software coding, growing really, really fast. Customer service agents, not just customer service chatbots, but agentic customer service like, you know, Finn by Intercom and Forthought and Zendesk. Agent Force, fast growing category. Agent Builder category, growing really fast. A lot of companies want to leverage these platforms. I mentioned Agent Force Service now, UiPath, et cetera, to build armies of agents across their site. That's growing really fast. Where, so, so the agent categories from, from, from horizontal to vertical, growing extremely fast. AEO, growing extremely fast. But we're also seeing other breakouts.

• 44:23Speaker 01

  We're seeing a lot going on in AI voice. So voice is becoming a breakout category that we're paying a lot of attention to. Orchestration software has become extremely popular over the last few years. And I'm not just talking about orchestrating agents, but I mean, like full stack orchestration, like how do you bring together rules-based AI, RPA, and then agents into a single workflow? That, that, that category has exploded for us. One thing that I am predicting and we're seeing it grow this year, but we expect it to really surge over the next few years, is the category of third-party agent guardrail services where companies will manage guardrails, permission requests, exceptions, rollbacks,

• 45:08Speaker 01

  remediations, QA for companies as they scale their use of agents, but their employees can't keep up with the risk profile and the permissions that are required to keep human in the loop. There's going to be a massive ecosystem of companies. They're not going to have the 75% SaaS profit margin. They're going to feel more like a services company at 30 to 40%, but that's going to be a $30 billion industry within five years.

• 45:34Craig S. Smith?

  Wow. And, and you're expanding buyer intent signals and letting vendors see which accounts are researching competitors or evaluating alternatives.

• 45:48Speaker 01

  That's right.

• 45:49Craig S. Smith?

  What are the, are there privacy or ethical concerns?

• 45:54Speaker 01

  Yeah, nothing's, yeah, nothing, they look at data, they look at data more from a signals base, it's not addressable. So they can't ask us to target specific users and send messages on their behalf. They can't address those users. We kind of keep it at the IP range. So what we do know for their buyer intent data is that people at Nike, for example, are researching your product right now, which you would see, wow, that's in pipeline for us. We're literally at the finish line. That becomes strategic or a lot of people from Nike are researching your top competitor right now, but we can't tell you who it is at Nike, not at a department, not at an individual level, because that's where you would kind of cross over.

• 46:37Craig S. Smith?

  What was the last thing you said?

• 46:38Speaker 01

  Well, it's where you would cross over from the privacy covenant you have with the end user. I mean, Zoom information, Six Sense others, they aggregate it to the difference between them and us is their third party and we're second party. Second party, meaning we're seeing the actual user versus inferring it by scraping third-party web data. But you still have to, you still have to provide that intelligence to customers at arms length to stay abreast of not only privacy restrictions here in the United States, but even stricter privacy restrictions you see over in Europe where it's really, really high bar.

• 47:10Craig S. Smith?

  Yeah, yeah, although that's incredibly valuable if you if you can see, identify who is evaluating your competitors.

• 47:22Speaker 01

  You can't at the company or the IP level, you just can't at the departmental or individual user level because again, that would not be within the privacy covenant of how that that arrangement works with buyers.

• 47:34Craig S. Smith?

  But you can at the company level, you can see that.

• 47:38Speaker 01

  Yeah, yeah, it's not, you know, the way we do it is not the same as Facebook like where I mean, I've heard some crazy, I don't know what you hear about Facebook and privacy, but I hear crazy stuff like you're in the kitchen, you know, and you're talking about a product and all of a sudden you see Facebook ads for that product. We don't offer that kind of service. Or you're on Instagram and you scroll over a picture and you stop and look at it for a second and all of a sudden you get an email from, I do see a lot of those things in the real world and for us, that's not a capability we have. It's not something we're really building. Is there a demand for it? Well, of course, people would love to buy that targeting. The problem for us is that if you ran expected value calculations on the risk and the reward of doing such a thing, especially with G2, we're a private company, you are, you can imagine we have aspirations to go public.

• 48:21Speaker 01

  That's not a very bankable business model based on where we are right now. When you take a look at organizations that are delivering that, they've long been public and they're scale organizations and they certainly can survive that risk reward.

• 48:34Craig S. Smith?

  Yeah, you you mentioned the voice reviews.

• 48:39Speaker 01

  Yeah, that's been exciting for us over the last year.

• 48:41Craig S. Smith?

  Yeah. Are you seeing a particular demographic or use case using that? I mean, is it a younger demographic?

• 48:51Speaker 01

  Not necessarily. Yeah, I asked the same question. No, not necessarily. As a matter of fact, I I was asking the founder at one of the leading vendors for voice keyboards. The voice keyboards are things like Whisper Flow or Willow Voice, et cetera, and I, it's what I use. I I press the function key, I speak, it comes out perfect, it's structured perfect, bullet points go where they are. It's fantastic. I'm three times faster, Craig, because I never, I I never touched the delete key, the return key or the spacebar. Like, think about how much time will we type? We we we touch those keys. So voice works the same way, but it seems to cut across demographics. I just think there is a personality that is productivity driven that love to use voice. Chat GPT, I talked to a researcher there, same thing with Chat GPT voice.

• 49:36Speaker 01

  It doesn't feel younger or older. It feels like more of a productivity profile issue with people that just feel really busy. That's why I started to use voice keyboards. It's why we're seeing people write voice reviews. But what we do know is when they write a voice review, it's longer than if they would have typed it. And it's got more tokens, it's got more content. Just like when we're hearing about prompts, prompts are a lot longer when you enter one via a voice keyboard than if you typed it.

• 50:02Craig S. Smith?

  Yeah, and I've got to ask, is it Whisper Flow that you're using just for my own information?

• 50:09Speaker 01

  Me personally, this is not a G2 endorsement. I use Willow Voice, but I've used them all. I personally use Willow Voice.

• 50:17Craig S. Smith?

  Okay, if you're advising a software vendor on how to win an AI-first buying journey, what what are the top three priorities they should be paying attention to?

• 50:29Speaker 01

  Make yourself easy to work with. So so check number one. You need an llms.txt file in the root of your website that tells the models what your policies are for crawling and retrieval and that you're open for business. You need to make sure that you're only blocking what you absolutely must block for info security or critical IP protection. You need to avoid gating and if you do gate, you need to age gate, meaning gate should disappear on high value content on the 91st day because most of your leads are generated in the first three months. You need to avoid hard to crawl formats like PDF and publish everything you can in HTML or if possible Markdown. That's rule number one.

• 51:10Speaker 01

  Rule number two is to earn trust signals from dependable lists. So you need to figure out which high authority lists awards that are AI visible, that you can be successful with and develop relationships. But you know, to do that, Craig, you better build outstanding products because like even with review sites, you don't get good reviews because you ask for them. You you get more reviews because you ask for them. They're good reviews because your product is excellent. I do think what that does bring up is that we're now in a world where this concept called product-led growth is truly possible that I don't think was possible before AI search. I believe incumbents with the sheer weight of their legacy

• 51:52Speaker 01

  made go-to-market growth the only, the only game in town and PLG was more of like a dream that every once in a while something like WhatsApp created.

• 52:01Craig S. Smith?

  Yeah. Is there a risk, though, that vendors will begin to see you not just as a review platform, but as a gatekeeper?

• 52:10Speaker 01

  Yeah, why? Yeah, a- anybody that that can influence your success is going to be seen as a gatekeeper. So, yeah, I mean, I I can't imagine that that some wouldn't think of that. If high authority lists were the biggest weight at test time, you know, to to to win the answer. And we know Gartner blocks AI. We know some of their winners post the magic hues. You know that a lot of publishers like tier one publishers, AP Bloomberg, they block AI. We don't. Then then then, yeah, we would be one of the ones that are going to influence your success. That being said, we're not charging you for a review to be put on your website. We're charging you for services that make it easier for you to scale the the volume and the recency of your reviews.

• 52:54Craig S. Smith?

  Yeah.

• 52:54Speaker 01

  It's a fine distinction. It's a fine distinction, but I would understand a vendor that would see us not as the G2 from from years ago that you just wrote a bunch of reviews, but now a G2 that has a lot of marketing services that can help good go great.

• 53:08Craig S. Smith?

  Right, right. And that's, I guess that's, you're saying that you keep the marketing very separate from the reviews. You know, being at the New York Times for years, I'm familiar with how ad sales is very separate.

• 53:23Speaker 01

  Editorial versus ad sales.

• 53:25Craig S. Smith?

  Yeah.

• 53:27Craig S. Smith?

  But but is, does that create a perception that that people should buy your marketing services? I mean, they they may think, well, it, you know, it it may not help, but it certainly won't hurt. Is is that a

• 53:47Craig S. Smith?

  something that's going on, do you think?

• 53:49Speaker 01

  I think I think what the customers realize is that as more and more people use AI search to find products, they've got to invest more and more in trust signals and be less and less dependent on, you know, onsite content to win. And that's been a real paradigm shift from the age of SEO. It really is. And I think it's part of a much bigger picture, Craig, where a lot of marketers are just really grappling with this zero click behavior that has emerged, started with AI overviews. Like, we always act like it was Chat GPT. It wasn't. It was AI overviews. And and I think that's the, if you want to talk about like the way people look darkly at at various companies, I think right now

• 54:30Speaker 01

  that the biggest resentment I see is like, where's my traffic and when is it coming back? And the answer is it's not. And I think emotionally, that's the most difficult thing that companies are having to grapple with when they look at G2. They they they look at G2 as one of many sources of trust signals. And I think they look at us like they probably look at at other new trust signal sources like Reddit and increasingly I mentioned YouTube as places where they have to make investments. I think the question is, do they have to add more money to the marketing budgets? That's where resentment could come in. Or really, Craig, is it a question of reallocation? And that's what I think a lot about. For every, I wrote a long piece about this, for every dollar companies are spending on AEO today, they're spending 60

• 55:14Speaker 01

  on SEO tools, services and SEM marketing like Google AdWords, okay? So there's a reallocation that needs to take place. I don't think you need to reallocate 50% because of 50% because AEO tools, they're all nascent. You don't know what works very well. Really, honestly, it's not the same as SEO, SEM where, you know, SEM, you put a dollar and you get a bag of chips. Right? You know this after 25 years, you're getting a smaller bag of chips than you used to get. But I digress. Yellow Pages best year ever for revenue. You want to guess what Yellow Pages best year ever was? Take a wild guess.

• 55:48Craig S. Smith?

  No, I have no idea. I would think.

• 55:51Speaker 01

  2007, 2007. So these things go great until they don't. So I guess what I'm saying is we've got to do some reallocation from the world of search, but it's not as dramatic as most people think. I think it's more like four to 6% this year. So as a result, when people look at like, this is how much I pay G2 and then I literally pay Google 20X. I think that that the resentment is more towards the ROAS at Google dropping than than G2 offering another service that's $75,000 a year for for data.

• 56:27Craig S. Smith?

  Yeah. Where do you see this going? And particularly, you were talking about agents taking over a lot of this work and and making decisions. I mean, is it, is it going to get to the point where there is a software layer, a agentic AI layer that's making these decisions and there are

• 56:52Craig S. Smith?

  you know, pockets of data like G2 that they're

• 56:58Craig S. Smith?

  contacting to to to get data on which to make those decisions and this is all going to take place.

• 57:06Speaker 01

  Yeah, I think that using agents to purchase things on your behalf will be one of the latter use cases, right? So I don't see that happening right away, but I do see everything eventually having more agentic workflows for the most successful companies in the world because they're going to win by brute force. But let me give you a different vision to close the show.

• 57:25Speaker 01

  SaaS companies, to quote Satya Nadella, a SaaS company is a cred database with business logic. And it has been for, gosh, 25, 30 years. Here's what's going to change. Within just a few years, faster than we expect, these SaaS companies will transform from being cred database business logic to becoming harnesses. What's a harness? Skills, context, governance, connector tools, those four elements, and that's a harness. And I'm already seeing it with companies. I mentioned several. I've profound. Every week I talk to more and more companies that are pivoting to a new model where it is more of a harness that sits on top of language models, either frontier or otherwise,

• 58:07Speaker 01

  and their value add is upcharging token spend. They're going to be paid based on consumption. The value of the harness is going to determine their markup. That's the future of the software side of the business. I have absolute conviction.

• 58:22Speaker 01

  Five years from now, seven out of 10 companies today that are successful then will have made the switch from SaaS to harness. And I think that's the biggest change we're going to see. The the way that we as as adopters use it will be happening in stages, one use case at a time, based on risk profile and trust.

• 58:40Craig S. Smith?

  Okay.