Jessica Redman: How to Optimise Your Content for AI Models - Get Found, Be Visible
About this talk
This talk explores how to optimize content for AI models, emphasizing the shift from traditional SEO to Generative AI Optimization (GEO). The speaker discusses significant drops in organic traffic for many brands, notably a 70-80% decline for HubSpot, due to the rise of AI-generated summaries and recommendations. They argue that the focus should shift from attracting clicks to ensuring brand recommendations within AI interfaces. The session includes strategies like crafting comparison pages, improving user-generated content, and utilizing structured data to enhance visibility in AI responses. The speaker also highlights the importance of understanding customer personas and adjusting content strategy accordingly, while also cautioning against relying solely on keyword research.
Full transcript
[applause] >> Um, cuz I'm feeling a bit nervous. Let's go. >> [applause] >> Thank you. Appreciate that. Um, how to optimize your content for AI models. So, a little bit about me really quickly. I'm an SEO, GO, and content consultant. So, I freelance. I'm also co-founder of local SEO agency Ditch Heads. And also do a bit of fractional um digital marketing director at B2B SaaS brand. There's
a lot going on, I know. Um, AI is helping me a lot. So, this is probably you guys watching another talk about content and AI and LLMs. The schedule's packed full of it, which I love. But, come on, we must admit sometimes it's a bit like, "Oh, okay, another one." Hopefully, I'm going to make this interesting for a good 45 Mic is I know it's back. Um,
to keep you engaged 5 minutes. Do we need a handheld mic? Maybe? I don't know. Um, I want to play a game. what we're going to do is I'm going to ask a question at the end of this talk. Um, and the first person to put their hand up, I'm going to ask the answer to the question. If they get it right, they get I put Oh,
I changed it to 50 quid cuz I thought you guys would think I'm tight. Um, so, oh, the first correct hand up will get a 50 quid Amazon voucher. My co-founder and colleague here, if you go to her at the end if you win, she will sort you out. So, pay attention cuz the answer is in this presentation. Um, 58% where most of us are marketers in
this room, that is statistically significant. I'm sure we can all agree. If I said you had a 50% chance of dying if you touch that window, or you could the other percent the the remaining percentage you'd win a million quid, you wouldn't touch the window, right? Um, is how much clicks dropped off in the last 12 months from Google search. Um and this is for keywords that
now kind of render the AI overviews at the top. That number is crazy for marketers. We've literally spent the vast majority of our careers worrying about clicks, and we just lost 58% of them. That is concerning. It's something we really do need to recognize. And this is SEOs at the minute. We're like, "Yeah, this is This is not a problem. This is fine." Like, "Clicks are still
important. SEO is never dying." But like, we need to recognize this. We need to figure it out. We need to We need a strategy. We need a plan. So, here's a little bit about why you should care about my talk and why I think you've made a good decision coming here. Um so, I'm going to start with a few stats because I was on a panel yesterday.
A few of you I think came to that panel. I saw one of two of you, but there was a lot of debate about whether GEO is actually a thing or if it's just SEO repackaged. I am You'll soon to find out on the side of GEO is a separate thing. Um but I've come with stats, so you know I'm not just trying to sell you Um
37% of consumers start searches with AI instead of Google. Great. We know that. Um Usage of AI is now 56% the size of search worldwide. That growth curve is mad. It does tail off at the end a little bit, which is really interesting cuz it's almost like I wouldn't say it's early adopters. It's gone beyond that, but we're kind of seeing a bit of a stabilization in
terms of adoption. Um I think that's probably going to go through another spike sometime soon, but 56% the size of search worldwide, massive. But the interesting thing is that doesn't mean that people have stopped using Google. They're now just adding AI models to that kind of that that customer journey. So, 70% 77% of consumers use AI and search together, and only 4% have abandoned search entirely. So,
the SEO is not dead crowd are very much right in that sense. Um I really like this. It's really interesting. So, there's a bit of a theory that basically, if you're a massive brand, like a Nike or something like that, you'll just automatically rank in AI models and AI overviews. Um historically in SEO, we've seen that happen because we've had backlinks have a massive weighting, right? And
domain domain authority. So, those big brands would rank well in Google. AI is a little bit different AI models. So, this basically looks at um the market share of a brand and weighs it out relative to the share of voice in AI answers. So, these brands here are over performing quite significantly, 60% more than their actual share of the market, and these brands are underperforming. So, basically,
doing GEO is going to help you over perform according to the this kind of share of the market you have as a brand. Um 62% of consumers now trust AI to guide their brand decisions. Um that's on par with traditional search. So, the data we're seeing is that that trust layer of AI a lot of marketers are skeptics, right? But we're kind of in a bubble. So,
we're like, "Oh, would you really trust ChatGPT to tell you data?" Blah blah blah. But we're talking about the general public here, and what we're seeing is an added layer of trust. because ultimately, AI has a different way of answering questions. It's not just giving you a list of blue links. It's correlating lots of third-party information on the internet, and it's a recommendation engine. So, it's giving
you a recommendation. And this is the old journey. So, you'll see noticeably, there's no click. Um so, typically we'd go keyword results page results page list of blue links and then it would be like, "Okay, user, now it's up to you to go through each blue link and compare and decide what product you want to buy or solution. You've got to do multiple clicks to get to
each blog and figure that out before you decide where you want to spend your money." We don't have that anymore. There is no need for those multiple clicks, right? In the in the in the top of funnel, mid funnel, lower funnel, yes, but my opinion that will be erased soon when the likes of ChatGPT and Claude will allow consumers to purchase and subscribe within the AI model
interface. Um so, we've got prompt refinement, summary, and a short list. You want to be in that short list because people trust those short lists very much. And this is a point that I really want to nail home when it's like GEO is different to is not different to SEO. AI models that haven't reinvented search, they've changed the interface through which that demand is captured. And as
a result, our behavior has changed. When you change an interface, when you launch something like Uber, the way that you get a taxi changes. You are influenced by emerging tech. And that means less clicks. It means more trust. And the data shows that it means when people do click, they are much more likely to convert. Um this is the HubSpot CEO. So, I want to talk through
an example, HubSpot. How many people use HubSpot in this room? It is massive. It's massive. You got CRM, email, it is HubSpot is like the default, right? HubSpot experienced a 70 to 80% decline in organic traffic in 1 year. Absolutely fell off a cliff, right? Um Ah, slides, they were in a Google thing, that's why. HubSpot have built a strategy and you see this a lot in
SaaS, B2B specifically. They build these massive blogs and the upper funnel drives tons of pipeline because they're educating people. It's been a really, really big strategy and it's worked for a lot of us over the years. That strategy is being wiped out by AI overview. So, the vast majority of informational searches are the ones that we're losing ground on. Things like this for HubSpot and the other
point to make I would argue famous sales quotes, cover letter examples, just useless. Useless content that a brand like that doesn't really need. It's not related specifically to their product. It's not product-led SEO. HubSpot went in a complete panic because [clears throat] the vast majority of their pipeline came from their blog. It's crazy. Crazy. The vast majority of a pipeline of of a company like that from
their blog. It's tanked to just 10% of leads coming from their blog. If that is not an emergency alarm bell for the marketers in this room, I don't know what is. The blog format is dying in terms of lead capture, okay, in a lot of instances. I know there are people that are sitting here probably going, "Oh, that's rubbish. My blog." Yeah, but majority, right? 73% of
B2B websites lost significant organic traffic um and 58% of US Google searches now end with zero clicks. So, it's not just a HubSpot problem. It's happening across the whole industry. Meanwhile, in the SEO industry, we're still talking about clicks. this is not the job anymore with AI The job is not clicks. The job is to get your brand recommended by AI. And then what typically happens, what
we're seeing with user behavior, there's no click, they will then go back to Google search and search for your brand or they will go direct. So, what I'm seeing and probably a lot of you might be seeing in GA4 is a spike in direct traffic. And you're going, "Where's that coming from? Cuz we haven't done any above the line." It's coming from AI they are driving that
increase in direct and that is again a reason why clicks are not the job anymore. Clicks are not our job. Okay? GEO isn't a traffic strategy. Yeah, it's a recommendation strategy. The question now is not how can we rank? The The question of rankings in AI models, pointless. It's how do we show up when someone needs help? Let's forget this kind of wishy-washy high volume, oh let's
talk about subjects vaguely relevant. When somebody needs help, when my ICP needs help, how can we make sure that we are the brand that formulates that narrative? Okay? Um coming back to this stat, 23 times conversion rate vs. traditional organic. I'm going to come over to this side of the room cuz I feel like a change. Um 23 times conversion rate. So, the reason we're seeing that
is back to the trust layer, right? Is because people trust the recommendations AI giving them. So, when they do click, they are converting. And if we've been paying attention to how ChatGPT and that are evolving, we're seeing e-com launched within the ChatGPT interface in the States. It's going to come to Europe soon. We're going to see that in Claude. We're going to see a subscription model. I'm
going to take a big bet in front of all of you so you can hold me accountable. I think that websites are the thing that are going to die. I think websites as they exist now will not exist in 5 years time. What I think we're going to have is just Does anyone know what markdown is? Yeah, okay. So, simply markdown is like what we see with
H with websites is like a HTML version, right? It makes it look pretty. It is branded, colors, all that fun stuff. You can click, you can interact, blah blah blah. With markdown, it's basically just a text file with a bunch of information. It's like on your Mac or whatever, that's what it looks like. It's just text. But that is what AI models do really well when they're
reading and digesting information. When you serve an AI model like a web page, it has to convert that into markdown. So, it takes a little bit of time and a bit more effort. If that web page is already in markdown, it's super easy for the AI model to read that information. So, I think we're going to get to a point where we only have markdown versions of
websites. Um and all the brand activity is going to happen in social and other areas of the internet. I'm so excited for people to come and be really annoyed with me after about that. Um so, stop creating content for clicks, start creating it for conversations. Okay? We're moving away from this idea of keyword research. Honestly, as an SEO and I'm probably going to see people again tag
me in this on LinkedIn. don't do that much keyword research I know, shocking. What I do instead is start with your ICP, your personas. Once you know them, you know what they'll be searching for. You should anyway. Any Any good marketer should know what their ideal customer profile and personas are going to search for without having to validate that in Ahrefs. Cuz if they're not searching for
it, you don't have a business, right? So, figure out exactly what their problems are. Why do they choose you? Why do they need what you offer? And then you create content for that journey. Marcus says we should be working really closely with customer services, with sales, and asking them, "What are people asking you? What are the problems people are bringing up day-to-day?" That's how we create a
content strategy for AI. It's not Ahrefs and SEMrush. Sorry if you're in the room. Um that's maybe a little bit more relevant advice to sort of B2B or SaaS. E-commerce is a different story. Um I don't work with a lot of e-commerce brands, but if you do, I would say very much keyword research is still relevant. You need to know what products people are looking for. SEO
versus GEO. SEO is a retrieval engine. GEO is a recommendation engine. Um SEO is can your page be found? Can it rank? Can it satisfy the query? GEO, does the model understand you and what you're saying? Can it classify who you are for? Coming back to ICP and personas, who are you for? Because the other thing we're going to see is personalization across the internet is going
to skyrocket. So, we're doing things with HubSpot at which is really about um dynamic landing pages, right? That change according to who is viewing them. So, they'll see different content based on who they are and and different details around ICP and persona. That is going to skyrocket. So, we need to be thinking about AI. Can it classify who you're for? Is it really clear everywhere online who
your product or solution is made for? And as a result, will recommend you. Lazy argument. I Sorry if the guy's in here that said this yesterday, but I disagree with you. You said that AI overviews are just a rehashing of the top 10 search results in Google. Not true. Not supported by data. The data shows that 60% of AI overview citation come from URLs that don't rank
in the top 20 because GEO is not SEO. It's different. You can't expect that because you rank in Google, you will be recommended by ChatGPT and You can rank and still be invisible. So, with AI models, there's a lot of stuff going on in the background that we don't know about. There's things like parametric knowledge, so that will be like the brain of knowledge that the AI
models have that kind of just stays there. It's like it's training data. Fan out queries. So, when you search for something in Google, what happens is a technique called query fan out. You you've probably heard a a lot about that in other talks if you've been to them, but ultimately, Google will decide a bunch of other queries that are relevant to that first query. It will fan
them out. It will rank different articles. And abstraction and waiting, so that's just like the math side. We don't we don't care about that. Boring. I'm not an engineer. I don't know. Um some researchers believe that AI models have already made their mind up about what to recommend before retrieval even happens. LLMs are basically predictor models. They predict the next com- the next word in a sentence
according to their answer. That is fundamentally different to how Google works. Um it's not just an opinion. Princeton, Georgia Tech, and Allen Institute, they found that GEO tactics increase visibility 40%. Um the tactics that work tactics that SEOs default to. We're not just GEO conspiracy theorists. It's real. Um here are the top three tactics that we found move the needle, all data backed. One, adding statistics. If
you can add statistics in your content, if that's proprietary data, even better. Um we did a big exercise recently. If any of you have heard of Opinium, we did a a of a survey. A thousand people in the UK is a representative sample size. So, we asked about five questions to the general public. Cost us about a grand, not that much. Got a bunch of stats back,
sent that out. We saw about around about 25% uplift in citations across our entire After that, it got a bunch of backlinks as well. Um two, quotation edition. So, adding quotes. If you publish a blog, blah blah blah, if you do press releases, stuff on social, quote your CEO, quote your CMO, quote people that matter. If you work with experts, quote them, too. And then three, citing
credible sources. So, your content should be citing other websites and brands that are credible. That helps. Um GEO is seven jobs. Um one, prompt research. So, it's not keyword research, it's prompt research. Understanding that prompting is a conversation. It's not just two little keywords together. Um recommendation research. Find out who's being recommended. Figure out why you're not being recommended. Diagnose the gaps. I use Peak AI, P-E-E-C.
That will help you understand where the gaps are. It will show you where the AI models are pulling the recommendation from, Reddit threads, blah blah blah. If you're not in that Reddit thread, find a way to get in the Reddit thread. Um build recommendation-ready assets, data-backed quotes, infographics that actually are researched. Again, proprietary data. Reduce ambiguity. So, if you create a piece of content that has a
bunch of quotes or stats in it, share it everywhere. Share that blog everywhere. After this talk, I'm going to take this presentation, I'm going to put it in Claude, it's going to generate five articles for me. I'm going to put them on LinkedIn, my website, I'm going to share it on my Instagram, I'm going to drop them in Reddit. That's just telling everyone, "Hey, Jess spoke at
this conference. She must be good at GEO." I mean, can't be sure if it's true, but whatever. Um would you measure properly? So, again, Peak AI is really good. A lot of people use profound as kind of the more expensive version. Peak AI is quite affordable. going back to the gaps. So, I really love a tool called Air Ops at the minute. I'm also using N A
N. Air Ops is a bit more user-friendly. They have like their own co-pilot built in. So, with Air Ops, you with N A N, sorry, you have to kind of build out these workflows and if any of you try to use it, if you're a beginner with that sort of stuff, it can be quite intimidating. With Air Ops, you can just tell the AI the workflow you
want to build. So, you just drop it in, I want to build a workflow that does a Google search, searches for that thing in Claude, identifies my gaps. That's what I did. I've got a keyword that I want this a commercial page I want my client to rank for, safety data sheets. I did this workflow that does a Google search, it does a ChatGPT search, does Claude.
It looks at competitor URLs, so other people that are getting recommended. Oh. it iterates the competitor URLs, scrapes, looks at my page, and then tells me where my gaps are. So, it will literally do it for me. It will say, "These are all the things you don't have that your competitors do. Go and build something like that as a workflow. Um prioritize these pages. So, data shows
that home pages, commercial pages, are where AI models mostly pull information from about you. Um so, if you're going to do a workflow like that, start here. my arrows. Mid-funnel content as well, so competitor pages. Um start there. Don't worry about this stuff on the on the left because this is the stuff getting eaten up by your AI overviews and um whatnot. Um this workflow had on
the pages that we refreshed drove a 54% uplift in share of voice for Enhancer. These lines represent our competitors. Just doing that refresh exercise, you can see how much we closed the gap. Just ignore this little drop Um, no, just kidding. This This is another experiment I'm doing at the moment, so I don't know if that's what's caused it, but we'll see. 85% of brand mentions in
AI come from third-party sources. I know there are few UGC people in here, um, so this is massive. You can tell AI models until the cows come home who you are and what you do, and that they can trust you and that you're the best at something if the rest of the internet doesn't say the same, you've got a problem. Um, reviews, Reddit, YouTube, directories, your author
bios. If someone's writing a blog, author bio, that's old school SEO, right? Um, I kind of contradicted myself there, whatever. Um, partner pages and press mentions. So, this is all the third-party stuff you need to be thinking about as a brand. And the really big one that I really want to drive home is reviews. So, um, working for Enhancer, which is the B2B SaaS company I told
you about, they didn't have a presence on G2. They didn't have a presence on Capterra. It's the first thing I got budget for when I joined. Um, it's a decent investment, but absolutely worth it. If your brand, product, like whatever it is, is not on a third-party review site, make it a priority that you go and sort that out. Then what we did is we got a
list of like our top 20-25 customers. We asked our, um, sales reps which ones are the most happy with our service, and then we asked them for reviews, and we incentivized it. So, we gave them a £50 Amazon voucher. I promise that's not the £50 Amazon voucher I'm giving the winner. >> [laughter] >> I just I haven't just taken it from So, incentivize it. If you have
a problem getting reviews, it's so worth it. Absolutely is worth it. So, this is another test that I ran. I used to be a personal trainer before I came back to SEO, and I had like a personal training website. And I wanted to rank for local queries cuz I was a face-to-face personal trainer. So, I built loads of local citations. Who knows what a local citation is?
Okay, so I'll explain. So, a local citation is things like directories. So, if you have a business in North London, a place let's say Enfield, which is near where I'm from, you want to look at all the directories around that area, and and also around London, and you want to add information about your address, where you operate, your services, what brand name, stuff like that on lots
of different directories and websites. It's not It's not to do with links. It's not a backlink strategy. So, it's a little bit different from that old black hat technique of just adding loads of links to directories. This is a well-known local SEO strategy. So, I added hundreds of local citations, and then I deleted them all. And you can see this is clicks from all the different AI
models. By the way, a lot of people are asking how they report on how successful their GEO has been. Simply for me, yes, the the tools like Peak AI are helpful, but just build a report in GA4 for for the clicks coming from AI models. It's the best way to do it because ultimately what you want to look at is conversions and revenue. That's our job, right?
So, citations definitely had an impact. Let's talk about UGC a little bit more. So, it's a trust layer for AI search. 50% of AI citations come from UGC and community platforms. Absolutely massive. Um as we know Reddit's a big one, but there's so many different UGC platforms. I'm just walking over here to see how much time I've got left. Okay, cool. Um is that counting down or
up? I've only been talking for that long. That can't be right. Is it? Okay. Um trust layer for AI search. community platforms. Um a lot of people say, "Well, Reddit, like you know, I'm a like e-com brand, doesn't really make sense for me. Isn't that like a B2B SaaS thing?" Absolutely not. Look at Sephora. They've got 482,000 weekly visitors on their Reddit thread. Um create a subreddit
for your brand. And direct customers to that subreddit. So, you need to link up with your email marketing team. Tell them about the subreddit. Tell them to include it in your next newsletter and say, "Hey, if you've got any questions about any of the products we offer, you're having any problems." It's like a back in the day how Twitter used to be the customer service thing. Let's
do it on Reddit. Um they can share insights. Again, incentivize it. Okay, if you share a video of you doing X, Y, Z, you can be entered into a competition to receive X, Y, Z. This is a really good one and it will you're basically getting other people to do the work that will help you show up Um review platforms rule. So, voice reviews contain three times
more content than form-based reviews. So, a really good way of doing this is for example, actually G2 now offer an option where people can leave like reviews uh videos on their reviews. So, rather than just typing them, they actually record a video. the more content the data shows that the more content in the review, the more language patterns for AI to learn. So, basically if you're in
set the more crap people say about you, the better your representation in AI answers. And recency matters. The more recent, the higher the impact on AEO. Oh, and another thing I was just going to say is I mean this was from a guy called Alex David who does work at G2, but that doesn't mean the data is not real. So, there's a three times higher chance of
ChatGPT citation if you were on any of these platforms. Um offsite is more important than ever. I cannot kind of hammer this home enough I feel like as marketers we used to have this view where it was like, "Well, this is not really relevant in B2B. TikTok, why would we want to be on TikTok? We're a software." Or, you know, this type of thing. Reddit, we're not
we're not we're an e-com site. Like why would we want to be on Reddit? That for less true than it ever has been. So, that presence again build a workflow. There's so many ways if you create one piece of content, if you run a webinar that you can use NAN to turn that into something that publishes automatically across all of these sites. There's a really good video
AI tool I use that's called veed.io. v e e d dot io. That is a video platform that can basically turn any content you can make scripts, it can get an AI avatar and turn it into a piece of content. And if you are an international company, it can even translate it into multiple different languages, which is great. Absolutely fantastic. I highly recommend Um so, in our
awareness stage we've got trends, recipes, awareness days. This is great social great for social, right? What I've talked about a little bit before is the awareness play not really hitting home on website as it once was. Consideration comparison pages. So, I um of an experiment recently. I searched for competitor um for Enhesa, a bunch of different competitors. And the information that came back about us as a
brand was quite off. It wasn't really how we'd speak about ourselves as a brand and it wasn't the message we would want to get across to potential clients. Um we didn't have any say in that narrative. So, we just built out a bunch of comparison pages, which I'll talk about a little bit more. Um but first of all, the thing we did when we built them out
was we looked at structure. So, structured content earns 2.8 times more AI citations. What does structure mean in practice though? So, comparison pages with three or more tables got over 25% more citations. So, tables, charts, visuals, things like that. AI models love them because they can digest them really easily. Um validation pages with eight list sections. So, there's a lot of lists, I know, but again, AI
models love lists because if you think about it, when you ask AI stuff, how much of the time does it come back with just a list of stuff for you? Because it loves them and that that's where it's getting it from. Just going to have a drink. Sorry, one second. Um sentences 10 words or less um on short list pages. keep stuff punchy. Um don't drone on
too much. Um if you're using Oh, actually, question. Who here is using chat GPT as their main LLM? Hands Who's using Claude? Oh, okay, more Claude. Who's using something else I haven't mentioned? I just canceled my chat GPT subscription >> [gasps] >> I much prefer Claude. If you are writing content, use Claude. It's excellent. Um yeah, you feel me? Yeah. Um and include banned words in there.
So, if you start to find bad like stuff that just keeps repeating, include those in there. Include it in your skills. Um who's using skills in Claude? Yes, love that you guys. You guys are skillful. I'm going to share a skill with you at the end that I really, really love. Um but I would recommend if you are using chat GPT as your main LLM, try Claude,
please. It's fantastic. You can connect it to all sorts of different things. It It's just incredible. The writing is excellent. The content is much better. Um content types that drive AI um G2 do really well from their best of guides. So, you can do that. You can look at doing that programmatically. So, if you have lots of different topics that you want to recommend best of, again,
build a workflow. Publish 20, 30, 40 of them. A lot of people say quality better than quantity. Yes, if you have to choose one or the other, quantity is important. Don't underestimate quantity. We're in a world where everyone is ramping up the amount of content they're creating constantly. So, if you're not delivering volume, you're going to get left behind. Um best of guides, product reviews, comparison pages
we've talked about, learning hub articles, and category pages. Best of content drives 56% of all AI citations. So, another thing to do, which I love this workflow. Um, what typically happens is people think best of content, okay, I'm going to go and create a list on my own blog telling everyone that I am the best. you might gain AI for a bit, but how happy are your
customers going to be about that? Is that going to build trust? Probably not. Better other people to say you're the best. How do you do it? So, something I did recently, I bought a bunch of blogs that were recommending best of stuff. And what you can do with software like Clay or any of the kind of email um data software is then just export a list of
all the people that write in those blogs, whether it's the journalists or whether it's site owners, founders. Get as many emails as you can that correspond to those blogs and get in touch with them. Don't ask them, "Hey, can I you say I'm the best?" and not tell them why. Provide them with information, this is why we're the best, we're not in your list right now, I
think we should be. >> [snorts] >> Um, this is what we did. I mentioned it earlier, so I I noticed that we weren't owning the narrative. So, we built out some comparison pages. And what basically the highlighted area is just where I've clicked on where we're ranking in the AI overview. It highlights where it's taken the content from. Interestingly, it took the content from the most struct-
the kind of structured part of the blog, which literally is nothing fancy, it's just two lists saying Spherovelo and Anki. And this really changed the narrative completely. AI overview is pulling everything about the competitor from our Which is great for us. find the gap, use the stats. Use stats to build trust. This is a workflow that I would look at. Structure it well, and you own the
narrative. So, we've got stats in there. We're not just making claims that we do this and that. We're kind of explaining what we're doing and why we're doing it. this is the result, actually. So, this is us versus the two competitors that we did these comparison pages about. And you can see over here share of voice, how they dropped 4.29%, 5.14%, and how we gained pretty much
the exact same. So, we actively took share of voice from our competitors just by creating better comparison pages um than theirs. And that what I will say about comparison type content is be honest. Don't say, "We are better for everybody." because AI won't trust you. So, what we did in our content was actually, we're not good for this type of thing. These guys are better. Actually, they're
better at doing this, and we're better at doing that. So, that honesty and that neutrality, neutral language is so important in terms of AI. Um another workflow that we did. So, um I'm going to mention AirOps again cuz I really love it. I swear I'm not being paid by them. Um one thing that the research was so showing is that adding like a TLDR summary at the
top of your content is really, really great for um AI citations. So, I built a workflow in AirOps. All it does, it crawls our blog, and it lists every single blog, and it will analyze the content with and it will write the summary for us. I connected it directly to our WordPress cuz we've got about 700 blogs. We couldn't possibly do that manually. And it literally writes
it and just inserts it into the blog at the right place. And doing that across our blogs, 32% improvement in citation position. So, citation position is is for those of you might not know, it's just like the ranking order of which you're mentioned when AI recommends you. So, this is a really good workflow and I highly recommend it. Um markdown version of URL. So, this is a
funny one because um a lot of people will argue that it's a load of rubbish. So, um markdown is what I mentioned before. A few of you put your hands up knowing what it is. Um if you're not a developer or an engineer or something like that, don't worry. It's not really your job to know what it is. But it is your job to tell your dev
team that you want this for your website. I am not standing here and recommending it wholeheartedly. I'll be clear about that. I don't want you all going tanking your websites and then blaming me. Um this is a test that I'm running and this is a test um that G2 specifically they put markdown versions of every URL on their website. They put them live. And across their site,
they saw a 100% increase in citations. And on some pages, they showed a 300% increase in Um so, I just want to reiterate, the reason that markdown is good is because AI models can read it very very easily. They don't have to take HTML and convert it into a language they understand. This is their language. The issue that I'm not the reason I don't want to necessarily
recommend it right now is because I'm not sure if creating those pages will create a duplicate content issue for SEO. So, if I'm creating two versions of every page, is that going to just tell Google that I've got two two versions of everything? I need to still figure that out. What we've seen initially in our tests is we've got a 5% increase. I wouldn't necessarily say it's
statistically significant at this stage cuz there's a lot of other stuff we're doing. So, this one is a bit of a TBC. We'll see how it goes. Um the singest big single biggest predictor of AI visibility. Sorry, just got to check make sure I'm on time. Um the single biggest predictor of AI visibility, pages updated within 30 days. So, freshness is so important. Please stop writing blogs
a year ago and then just leaving them forever. Um it it's not a good thing to do. Um at the moment with Enhancer, I'm going through a whole content audit where we're just filtering out all the URLs that get less than 100 clicks, and we're just getting rid of them. Just get Absolutely pointless being there. If anything, they're just going to degrade the helpfulness of your domain.
Obviously, that is down to your discretion which ones you remove. If they're case studies, keep them, but um pages not updated quarterly are three times more likely to lose citation. So, if you're not updating your content, your competitors are, you're in trouble. For commercial queries, 60% of AI citations come from pages refreshed in the last 6 months. Read that sentence again. It says for commercial queries. I'm
not talking about blogs. I'm talking about when people want to buy stuff that you sell. Um if you refresh those pages, you'll win. A lot of people will just default to I'm just refreshing my blog. Look at the commercial pages, product descriptions. Can you change them up? Make them more detailed. look at what your competitors have got that you haven't. Your review widgets, are you feeding through
reviews to your product pages often enough? Is there any news related to your product or any regulation changes that you need to add there? >> The single signal that AI models care about more than any other is recency. Build a workflow. So, I use NAN. I've got a really good workflow for this exactly. So, every week it will crawl our blog. And in again, I'm not a
developer, but there is a little little thing in like the HTML that will that it basically knows when it was published. So, it's like a date published something whatever. It crawls that. It knows when the blogs were published. It will filter out stale posts that are 6 months or more old. And then if if it can figure out, okay, actually this was more than 6 months old,
it will just notify me in Slack and be like, "Hey, you need to go refresh this blog." Um What I will say is there's been some debate about some blogs that I've seen that have date published and then they also have date updated on and a lot of questions about which should we do, one or the other. Studies show that Google gets confused if you have both.
So, I would keep like date refreshed on as the only thing. Um And yeah, get your content team on a workflow that makes sense here. ultimately, optimizing for the keyword best CRM software pointless. Well, not pointless, but not really the way we're going to do it now. Instead, we're comparison page with three tables, eight list sections, fresh data, and five third-party citations. Not a lot of work
really, is it? Um We are moving into a period of time we have to start learning workflows, how to use Claude, how to connect everything, because we will get left behind as marketers if we don't. You should be learning about MCPs, you should be learning about skills, you should be learning how If you use HubSpot, connect it to Claude, build interactive dashboards. As marketers right now, it
is our job to be on top of all of this, because there are people out there that are on top, and if we're in this, not, we're going to get left behind. And that's horrible to hear, and it's horrible to say. And I don't want to create anxiety and have you going away thinking, "Oh my god, all this stuff, blah blah blah." It's not that. It's just
learning the basic way that this stuff works, so that you can go and tell the tech people to do it. We just give them more work, but we need to know as marketers what we're recommending, right? Um I built a skill that All of the evidence-based stuff I've got in this presentation, I've put in this skill. Um so, if you scan this, you should be able to
get the skill, which you will then, if you're the people using ChatGPT, maybe you want to move to Claude. You can put this skill in Claude, and it will basically give you that research-backed tactics, all the things I've talked through, all the things that are going to help you increase your chance of being referenced. You just need to add that skill. When you, um, sign up to
get this, it will tell you the instructions of how to do it. And you will then just go in your Claude and say, "Hey, we've written this piece of content. Can you tell me how I can make it better to get, you you um, in AI answers?" And it will tell you, "Hey, um, what is the best way for us to get visibility for blah, blah, blah?"
And it will answer. So, it's all baked into
More from this event
See all 5 talks →
Janina Bak: Why Smart Organizations Make Stupid Decisions? Business lessons
38:51
Stefan Chiriacescu: Email Automation on Steroids: AI That Drives Profits
45:56
Andre Alpar: GEO - Frameworks and Case Studies to Get More Visibility in LLMs
33:05
Giulia Panozzo: Fundamentals of Neuromarketing: How Brains Decide - and Brands Win
40:23