Transcription
Good afternoon, Grow. Oh. Participation already. I wasn't even expecting that. Thank you.
>> [gasps]
I want to start off by saying something that's maybe obvious. AI usage, it's growing. But do you know how much it's growing? According to this analysis from March 2026, monthly sessions of AI are now 56% the size of search worldwide. And to put that into perspective, Google has been around for nearly 30 years. ChatGPT has been around for 3 and 1/2. Bonkers.
And notably, this data comes from Ethan Smith at Graphic.io. This time last year, Ethan and his team were pretty skeptical of claims around how quickly AI usage was growing. But now, he and his team expect the use of AI to surpass search in the near future. So what does this all mean? When you ask customers, "How did you hear about us?" the answer is becoming ChatGPT. Are you ready for that shift?
By 2:30 p.m., my goal is to get everyone in this room feeling like they are. We can do it. Because you are going to walk out of here with these seven things you need to know about how to do AEO, a 30-day checklist so you can start implementing these insights immediately, and a renewed sense of purpose and a new lease on life. [laughter] Just kidding. Wrong presentation. But I do hope you find that.
I'm Major Frost. I like bad jokes. I'm senior director of global growth and paid advertising at HubSpot. And over the last few years, I have transformed our SEO team into an answer engine optimization team. And I've led the strategy that has made HubSpot the most visible CRM in answer engines. But HubSpot hasn't just been busy building our own visibility. We've been busy figuring out how to help our customers build theirs.
In October, HubSpot announced that it was acquiring xFunnel, answer engine optimization software that helped companies track brand visibility and citations. And on that xFunnel foundation, we built HubSpot AEO, which I am super excited to say went into public beta last week. This tool gives marketers full visibility into how their brand shows up in AI answers. And probably more importantly, what to do about it. You can try it free today by going to hubspot.com/a-e-o.
The xFunnel acquisition also gave us something very important. It gave us access to many millions of AI answers. And this is really powerful data. You know what they say about sharing? It absolutely crushes on LinkedIn. So today, I am sharing this data with you. These seven most important things you should know about how to show up in answer engines.
Before we begin, a quick vocab lesson. AEO, or answer engine optimization, is how you get your business to show up in AI answers. Visibility is how often your business shows up. It's a reflection of your mentions, or when your brand appears in AI answers without a link. Citations, meanwhile, are clickable sources. And both mentions and citations matter. Mentions build awareness. Citations drive traffic. Okay. Those are the terms. Now we can dive in.
The first insight: the blog is not dead. You might have heard that with the decline of organic traffic, the blog has lost its place in your marketing portfolio. But do not divest yet. We looked at 14 million citations. And blog posts and listicles were by far the most commonly cited page type, making up a whopping 62.1% of citations. Why? Because AI is looking for content that's easy to understand and extract. It's looking for content that helps it explain trends, opinion, analysis. And guess where that information typically comes from? It typically comes from your blog.
But it's not just citations. We can look at HubSpot's logs, which are like a guestbook for your site. When a bot comes to your website, it has to sign where it's going and when. And 19% of the time that LLMs are visiting HubSpot sites simply to learn and train, they are visiting the blog. 19% of the time. So what are the implications of this for your strategy? It means even if your blog is getting less traffic than it ever has, it's a very important channel for helping LLMs understand what your product is, the value that it offers, and the people that it is a good fit for.
In fact, one of the most successful AEO tactics that we ran at HubSpot over the last year, when we saw that ChatGPT was quoting old pricing details, we published a series of pricing blog posts. ChatGPT visited these almost immediately after we published them. I'm talking hours. And now it gives people the right prices. So simple, so straightforward, so important. Because if someone is asking about HubSpot's costs, or ChatGPT is proactively giving them, they are a hot, hot lead, and we need them to have accurate information. So here is what I want you to take away. Don't spin down your blog. If there is something AI should know about your brand and products, put it there. But it would be a mistake to approach blogging the same way that you did pre-AI.
Which brings me to insight number two. There is a shift happening underneath all of this. As more and more people adopt LLMs, search is going from keywords to context. And for years, SEO has been about short-tail queries. And HubSpot has targeted a metric ton of these. Yes, that is me, uh, indicating that the metric system is superior to imperial, but don't tell my friends at home. Uh, some of the keywords that HubSpot has targeted: business name ideas, best free CRM, TikTok trends 2026. But in answer engines, queries or prompts, bonus vocab for you, they look completely different.
One analysis of 13,000-plus ChatGPT conversations found that the average conversation was approximately 350 words, compared to about five words in a typical Google search. So rather than "business name ideas," users are typing, "I don't think that name will work, actually. There's another company with that name. Can you please come up with more name options like that previous one, but keep it to one or two words max and two syllables?" And rather than "best free CRM," they are typing, "Are there any CRMs I can use that aren't specific to one business, so I can use the same one for my car dealership and credit repair and home rental businesses?" This person is really busy. And rather than "TikTok trends 2026," they're saying, "Okay, this person is just really, really mad that ChatGPT is still using questions." The AI revolts. I am not too sure about their chances. Uh, but I share these examples to illustrate that models are receiving a ton of context. And when AI gets that much context, it is looking for content that directly answers a specific situation. When queries get more specific, content needs to get more specific, too. Because generic content, it's just not going to answer the question the way specific, contextual content will. The content that shows up in AI answers, it's tailored to a specific persona and or a use case. CRM persona? Car dealership. Use case? Generating leads and closing business.
But specificity is not the only important factor. What do the pages that get cited actually look like? It turns out their structure matters a lot. There are a few patterns that are clear across cited URLs. First, they are likelier to use question-based headings. And that one thing, formatting headers as questions, showed the strongest correlation with citations across almost every engine. And this makes a lot of sense if you think about how these systems work. They are literally trying to answer questions. So, title sections: "What is AEO?" instead of "AEO definition," and "How to do AEO?" instead of "AEO best practices." When your page already mirrors the structure of the question, it is significantly easier for the model to extract and reuse your content.
Now, a few other patterns also showed up. Having an FAQ section correlates with higher citations. And an FAQ section is exactly what it sounds like: a block of common questions and clear, direct answers all in one place on your page. You probably already have a few on your website. This works really well because you are pre-packaging content in the exact format that AI is looking for. Again, question, answer. And people like this, too. You're not just writing for humans or AI. This is actually a structure that works really well for both. Adding schema markup is going to give you an additional boost across most engines. An FAQ schema is just a way of labeling those questions and answers behind the scenes so that machines can read them clearly. It's kind of like subtitles for AI. You're removing any ambiguity about what's a question and what's an answer.
This is your picture moment. This is everything on one slide. Having 20-plus external links, a descriptive page title, block quotes, a TLDR section – those are all factors that are positively associated with citations as well. And the last citation pattern that I want to double-check into is statistics. LLMs are constantly trying to decide, "Is this specific? Is this credible? Is this worth repeating?" Which are all questions we should probably be asking ourselves about content online more often. Stats are going to accomplish all three of those things. They're going to make your content more concrete, more trustworthy, easier to quote. But execution is pretty important here. You are going to get the most citations with unique data versus data that's pulled from somewhere else online. That is what's going to make your content uniquely citable. So think business benchmarks, customer outcomes, case studies. This is also what you care most about an LLM repeating to a potential customer. Think about how powerful it is for ChatGPT or Claude to tell a user, "HubSpot customers using Customer Agent with Help Desk saw a 25% boost in ticket resolution." That's compelling. And you don't need a fancy market research survey to collect it. I think the pattern here is pretty clear. AI models prefer pages that are structured, scannable, and information/data dense.
Now, the next insight might sound familiar, but the degree of it is new. In SEO, we have always said it's a good idea to refresh your old content. But AEO takes the importance of this tactic way, way up. Because insight number four: freshness matters more than ever. Here's what we are seeing in the data. Answer engines rely even more heavily than Google on recent web signals. And this data actually comes from our friends at Ahrefs, but it's so good I have to share it. The average age of ranking Google content: 3.9 years. The average age of content in major LLMs: 2.9 years. That means that the content AI sites is, on average, 25% more recent. And the difference is most dramatic for ChatGPT. Content cited on ChatGPT is 1.2 years more recent. So, TLDR, or maybe "too old, didn't read." When an engine decides which sources to cite, a recently updated page is a very strong signal. It's telling the model this information is likely still accurate.
And one of the highest return AEO tactics that we have seen is very simple: consistently update existing content. Now, I don't mean making it 10% longer, slapping a new published date on it. Doesn't work. I mean refreshing statistics, adding more recent examples, getting a quote from an expert on the latest information, then republishing. Because with answer engines, recency can push you back into the citation pool. So, if you remember that pricing blog post series that I mentioned, you better believe we are constantly updating that because we want to be the most authoritative and current people talking about HubSpot's pricing. And there is a reason that there is an arrow pointing to that last updated tag. That's important, too. In our data, we found that having a visible last updated date, just having one, correlated with more citations across nearly every engine.
Now, the next insight is unexpected. And if anyone is still arguing that AEO is just SEO with a new letter, actually, I know they are because I see them on LinkedIn, you can share this data with them. Insight number five: backlinks don't predict AI visibility. For 30 years, SEO has had a dominant currency: backlinks. And the relationship was very simple. More backlinks, higher rankings. But when we looked at AI visibility, that relationship basically disappeared. Pages with massive backlink profiles were not consistently cited more often by AI. In fact, some of the most cited sources have little to zero link authority. This "Top CRMs Tools of 2026" article is cited in AI responses 85 times for questions relevant to HubSpot's business, but it has just one backlink. Just one.
When I saw this, I thought, "How could this be?" But it goes back to the outcome that answer engines are optimizing for. It's fundamentally different than what a search engine is optimizing for. Google is looking for the most authoritative page on the web. Answer engines are looking for the best snippet to answer the question. So, instead of asking, "How many backlinks does this page have?" the model is asking something more like, "Does this paragraph answer the question clearly?" And that means that content clarity and specificity often outweigh domain authority. If you're a smaller company sitting in the audience right now, or maybe watching this on YouTube, this is great news. Because this means there is a much more level playing field.
But if backlinks aren't the strongest predictor of AI visibility, the next logical question is, what does AI trust? And at a high level, answer engines are doing two things. First, they're finding a clear answer. And then they're deciding whether they trust it. And to make that trust decision, they're looking beyond your site. We looked at the sources that show up again and again in AI answers, and three platforms dominate. Which brings us to insight number six. Answer engines look for consensus. When forming an opinion, they're asking, "Are multiple independent sources saying the same thing?" And they tend to look for that consensus on LinkedIn, Reddit, and YouTube. If you are choosing where to build your AI authority, those should be your priorities.
There are a few reasons LLMs return to these platforms again and again. First, answer engines love user-generated content. It's authentic, it's usually available in mass quantities, constantly refreshed. LinkedIn, Reddit, and YouTube are, of course, rich libraries of UGC. And on top of that, OpenAI and Google have partnerships with Reddit. Microsoft has invested tens of billions of dollars into OpenAI and owns LinkedIn. And Google owns YouTube. So, there's some symbiosis happening here. For product comparisons and recommendations, these are especially important platforms. 55% of citations for those final research questions go to these three platforms. So, if your brand never appears in YouTube transcripts or Reddit discussions or LinkedIn posts, you are missing three of the most influential ecosystems feeding AI's opinions.
But importantly, this does not mean spamming Reddit threads. Please do not do that. It looks like creating an account for your company that can interact in relevant forums. Ryan Gunn is a Reddit HubSpot goat. He is a top-cited Redditor for us, and he also works at a partner agency, so this is really good for his business, too. It looks like posting customer stories and demos and walkthroughs to YouTube, like this HubSpot customer, Travelers Autobarn. And it looks like encouraging your C-suite, your sales reps, your marketers to post on LinkedIn. Now, you might be thinking this feels like a lot of work, and it is a lot of work, but the good news is that we have found when we are active in these communities, mentions of HubSpot go up across the board. By reminding people of your presence, they are more likely to talk about you. So, you create this compounding effect where your activity has an exponential impact.
Now, the final insight is probably the biggest mindset shift of all. Because most of us, myself certainly included, were trained to think about performance like a snapshot. Rankings today, traffic this week, conversions this month. And that mental model really breaks in AI search. Because visibility is much less fixed. Answer engines aren't retrieving a fixed list of results. They're generating new answers each time. And that means volatility.
Here's what that volatility actually looks like. Half of cited pages change every single month on average. Nearly 60% show up once and then do not show up again the next month. And within 6 months, the majority of citations have been replaced. Okay, that's citations, but you remember your vocabulary lesson, so you're thinking, "What about mentions?" Even if the sources that LLMs are citing are changing really quickly, do the businesses that they're highlighting stay constant? There's a little more stability there, but not much. Ahrefs did a similar analysis, but for mentions, and from one answer to the next, only 30% of brands stay visible. In other words, the same question is producing different answers with different sources every time.
In a volatile system, moments, they're meaningless. Trends are what matter. So, the real question is not, "Did we show up today?" It's, "Are we showing up more over time?" And that is exactly what tools like HubSpot AEO are built for. Tracking your visibility across engines over time so you can see what's working. So, yes, AI search is volatile. The teams that win are the ones that measure it correctly, optimize accordingly, and improve over time.
Okay. I've given you a lot of data. Seven insights, 32 data points, eight charts, to be exact. And now I want to make it really simple and give you a checklist. Exactly what I would do if I were you over the next 30 days.
First, if you don't already have an AEO tool, hopefully I've convinced you that you need one. I'm obviously partial to HubSpot's. Upload the prompts that your buyers are asking AI. And remember, context, not keywords. If you're not sure how to format those questions or what they should be, HubSpot will analyze your website and your competitors to help you find the best ones.
Second, create or update five blog posts. And these pages should include the information you most want AI to understand and cite, like why you are the best choice in your category. And maybe your pricing. Turn your headers into questions, add an FAQ section, mark it up with FAQ schema, add a last updated tag, incorporate unique data, and make sure everything is as recent as possible. You are serving your content up to AI on a silver platter.
Third, publish five use case/pages. Again, the shift here is from keywords to context. So, instead of "top CRM," it's "top CRM for real estate investors." Instead of "best rental camper van," it's "best rental camper van for traveling the Gold Coast with your family." I... This was a rabbit hole that I went down after I came up with that example. I hope someone out there is doing that. Generic content, it doesn't match the question anymore. Specific content does.
Fourth, show up where AI looks for consensus. Pick YouTube, Reddit, or LinkedIn, and create a content plan. Start posting.
And finally, measure visibility over time. Train yourself, and maybe more importantly, train your leadership to think about visibility directionally. Because with answer engines, you're not winning by being number one once. You win by being consistently present.
Okay. Here it is, all in one slide. Your 30-day checklist. The shift to answer engines, it's already happening. The question is not whether your customers are going to use them. It's whether they find you when they do. So, be the answer.
All right, Grow. Thank you so much. I hope you have a fantastic rest of your day. Thank you.
>> [applause]