How to Optimize your Content for AI Search (GEO, AEO)

Content optimization for AI search. What you should start doing that you didn't before with traditional SEO, or what are the top myths? To discuss this topic, I have a new episode with the wonderful and so very well experienced Cindy Krumm, Gianluca Fiorelli, and Celeste Gonzalez to go through the top do's and don'ts, errors, mistakes, steps, and criteria that they have found to be useful to optimize your content in this new AI search era. Also, something important, tactics, recommendations, and actions that are scalable and well aligned with what you're already doing with SEO as well. So you can make the most out of it and leverage it rather than going against it. At the end of the day, we all need to leverage and maximize each one of the layers and channels in digital marketing, right? And before starting the conversation, I want to thank Similarweb, that is our today's sponsor. They have actually quite a few different AI optimization related features that I highly, highly enjoy as a user myself. So go and take a look at them and thank you very much for your today's sponsorship. So to start a little bit the conversation, this is one of the topics that I really wanted to address. I believe that because of query fanout, because of chunking, Because of all of this additional, let's say, shift that we're seeing with AI search, we have seen the highest number of tips, of new tactical things to do to optimize our sites, focus on our content in order to maximize the visibility in AI answers. And I would love to start with those myths, those top errors, mistakes that you still see today, people making something so very obvious that I have seen many people start doing in their websites, like leverage AI just to generate content at scale, supposedly to maximize the opportunities to get cited in LLMs, in AI platforms. Like, oh, really? I thought that we already knew about this and that it wouldn't necessarily end well. So anyway, we'd love to hear your thoughts. We'd love to hear about your experiences.

I'm an SEO and AI search strategist, consultant, and I work a lot with travel, for instance. All the things that I can say are backed up by experience with clients. For instance, mistakes. I'm going to be maybe obvious, but for instance, creating content without considering that every content, any content doesn't live in isolation. It's part of a bigger cluster of content on your website. So forgetting this is also a way to not consider the buyer journey, the search journey that can start or can land on your content and start from your content inside your website. Second mistake I experienced myself. The biggest mistakes are not when you are commanding AI to write something, but when you are briefing AI. And that is what sources you are giving to AI, what instruction you're giving to the AI, what role priming are you giving to the AI, and so on. And third, and sometimes people forget this a lot, using AI for creating content without considering the brand bible. So the tone of voice of the brand, how the brand addresses the user, all the things that brand says and what are the things that the brand never says, so on and so on and so on. So writing something which for someone visiting the website sounds clearly not written by someone knowing the brand voice.

You are tying this with good copywriting, good marketing, rather than trying to look for workarounds, that which I believe is one of the other top mistakes. Like we need to start pre-rendering and making this content available SSR, because we know that AI system cannot render JavaScript like the Googlebot can. And, and they were like, oh, why we, we don't leverage this markdown files that I see that are being shared about as a workaround in order to get your content to be ingested by AI platforms? Like, why invest in workarounds rather than making your actual content Good content that is good for your users, your customers, and any discovery platform out there or marketing channel out there at the end of the day.

And I just wanted to add to Gianluca's point about the content in isolation. I think the biggest issue is people focusing too much on tactics and not overall strategy at this point, and it's prioritizing those short-term wins. And just like you're saying with the markdown files, okay, cool, I got the boost right now. They're not thinking next year, they're not thinking 6 months, from now and what the actual priorities should be to get there. Because, you know, thinking long-term with this, we're going to have personalization come into play. And at that point, when you're thinking about content, it shouldn't just be trying to get those quick wins right now, but it needs to think of where search is developing in the future and how you can be prepared for that, which would mean including a lot more ICP focus, ideal customer profiles, market research, using customer surveys, review data, actually using what the users or customers are telling these brands or companies what is important to them and making sure that that's prevalent in their content.

Funnily enough, this type of mistakes that I completely agree happens because of the overfocus on tactical actions rather than seeing the broader picture. Unfortunately, it's a typical mistake that we have also been doing in SEO forever. And we have also, let's say, sort of like inherit in the AI search era. And another thing that you reminded me of is about trying to come up with new ways to, let's get this content cited by chunking it in this specific way, like bullet points, everything in bullet points. Like, honestly, do you think that everything should be a bullet point or a table? Sometimes it makes sense because it's good user experience. It's the best for the UI, for the design. It fits well. with the purpose of the content, of the information that you're displaying. In other cases, it's not, right? So you need to think about that as well.

I consult with a lot of big enterprise companies, and I think that the biggest mistake that they make, you know, adding on to what Gianluca and Celeste said, is focusing actually only on the website. I think SEO and AI optimization has wildly changed things where the website is not where you're going to get the most traction. The website is the very last part of anything after people have already learned about your brand from other things. And remember that AI systems and LLMs and algorithms, they trust human interaction and what humans say about you more than what you say about you. And so we need to be optimizing offsite more. I think onsite should be table stakes. That should have been done. AI didn't change that except We have more bots and we worry more about JavaScript rendering, but we should have been worrying about that anyway. But that's the basics. If you haven't done that, you're in massive trouble already. But to really be visible in the AIO and any kind of LLM, it's a lot about, especially with Google, because that's where most of the AI traffic still is. It's about what's being said about you off your site.

100%. Thank you very much for bringing that up because, uh, it's indeed like most of the citations sources are off page indeed are from other sites. And whenever we think about content optimization, we always think by default, like on site, how are you optimizing all of your presence across the whole web ecosystem out there? What you are saying about you, even in your social profiles, right? Many times about how, how do you communicate what you do in your social posts? I had a client a few months back that they wanted to beat over the top created a little bit, and it's— this is like one-liner sort of description. It doesn't say anything. It doesn't describe what this product is, of this campaign, of, of this launch, how the potential customer will understand this and how it will be differentiated versus your competitors. So everything counts 100%, Cindy. And in AI especially, it's a corroboration problem.

Consensus. The thing that AI does well as an LLM, or like the LLM background in the AI that we're using, is Like it does a good summary. And so that means it has to be taken from multiple places. And if only one website is talking about you and it's your website, they don't, they're not going to include it.

For instance, in a campaign for, let's say, a new product, you have to create all this content and then create all the content following the launch and be monitoring and being consistent in the message, because sometimes the biggest problem is the inconsistency, as you were saying, Aleyda, of the messages that you are presenting on your website when chatting in a Reddit thread or in another specialized forum, or when you are just simply spreading a press release, for instance.

Taking all of this into account, what would you say are those fundamental steps that you always tend to take now when optimizing for AI search?

I work in the local space, so reviews are huge. So we work with local enterprise companies at last-mile retail, and I'm on the product side now. Reviews are absolutely huge. That's where you're getting people who feel very strongly about whatever their experience was with your products or services, and That's where you can see what's being said about you, what's being said about your competitors. It's absolutely a huge place to start in getting that data and understanding what can be your potential unique selling points. What's important to the people who are leaving reviews that maybe you might not have thought was as important? What are the gaps between the reviews that you're seeing with that text about your clients versus the competitors? What the competitors are doing well, what they're falling short on, and sort of hitting hard on maybe what they're falling short of, but what you're doing well on, and making sure that that's prevalent in as many places as you can do that on, whether that is, you know, the website, videos, social media, and if it makes sense, absolutely pulling social media comments. I mean, now that's— Gianluca brought up Reddit. Now we have so many people trying to exploit Reddit to have nice things said about them, but it's a good place to at least do some research and see what's already being said about your So I believe that is fundamental to take especially third parties into account in a strategical way, right?

And to align with their social media team, community management team, digital PR team, to understand what they are already doing as well, rather than what we were saying before about the tactical plays that might be seen as spam even eventually that are not recommended at all. Like, oh, let's start like self-promoting in this or that. platform, just because a lot of the citations are coming from there now.

I'll build onto what Celeste said. One of the first steps that I'll always take is to do some test searches and see if any influencers are showing up that are naturally mentioning our brand so that we can find out what are real people that are, you know, have video platforms online talking about when they mention us. And how can we do that more, capitalize on that, find more people who care about that? Empower any influencers? Because what I'm seeing a lot, number one, is YouTube is one of the, the most cited sources in AI overviews. But number two, the way people are discovering products online is changing dramatically, especially the younger generation with TikTok. TikTok's algorithm is measuring people's interest in certain topics, and to a lesser degree, you know, Instagram and, and other platforms, but especially TikTok. They look at, at micro-engagement is what they call it, and what you like, what you share, what you comment on to figure out what you care about. And that's how people are getting exposed to new products and services and brands and going, then going down a rabbit hole there and then they'll search, right? So to get to the top of the funnel, you have to start where people are hanging out casually, not where they're actually searching.

I'm Gen Z, so all of my local searches start on TikTok. Absolutely. I mean, it'll just come up on my For You page and I'm going to save it and it's like, I'll go visit there. But if I'm looking, I'm traveling, I'm searching on TikTok first and the website usually is my last stop.

I feel like so many SEOs are missing that boat because that's where AI is really affecting our job is the AI of our feeds.

Yes.

Totally. I take borrowed from a classic of social media marketing. I create a social media, uh, let's say newsroom or something like this in order to always be noticing what people are saying about the brand. And when I say the brand, I say also the product naming of a brand and so on and so on and so on, on social and YouTube, etc., etc. And What I particularly like to do is diversifying the content in order to have it both on-site, proprietary properties like YouTube channel, etc., etc., but creating a lot of co-marketing collaboration eventually with not huge influencers, but the so-called micro-influencers. The followers they have, they are really engaged in order to to start creating the positive flywheel of discovery outside the website, outside of search, search eventually, or direct traffic. And then going on, going on, going on in order to also create those positive user signals that we all know from NAP Boost and so on and so on and so on.

To start giving like a, let's say, prioritized step, what I will highly, highly recommend viewers is that They start by analyzing a little bit and seeing where the citations are coming because of what we just mentioned here, right? Like you will assume it's like, let's start optimizing every single page of my site, which we know that it will be ideal in this ideal world where resources and times were unlimited, but we know that they're not, right? So what should we be optimizing content-wise to optimize in LLMs for those AI answers that actually matter based on those prompts that we're tracking that are actually representative of our customer journeys for every product line, service line, locations, and so on. Let's go and take a look where the citations are coming. Like, is it from our site? And we might be very surprised to see that these are the category pages are not the one-sided, but the product pages, the help pages, or the community pages, because a lot of user-generated firsthand content really experience content, factual content that AI loves to cite is coming from this sort of pages, right? Go and take a look at which are the pages that are getting cited for meaningful prompts from your competitors' presence, not only on-page but also off-page where you are not, and prioritize accordingly. If you do already have that content and somehow it's not getting cited, go and take a look at the why behind. Is it because Your content over-relies on JavaScript is because your content is not well organized or structured in a way that is realistically easy to cite. Go and take a look and compare it with the one that is actually getting cited, right? In order to understand which are those characteristics. I'm not saying to overthink it too much. It should be something, but like, will I trust these other players rather than me? Many times it's just because you're too ambiguous. In your wording, right? And you don't specify the actual stats, the actual data that will be used as the source to make it factual, to make it much more reliable and to make you the trustable source, right? A lot of this also, I love to continue using the E-A-T principles as an easy-to-follow reference for any copywriter at the end of the day that they can understand because it ties well with a good user experience as well. So I will highly, highly recommend you to do this. Tools that you will love to recommend to use and that you also like to use in your day-to-days for content optimization.

SparkToro for audience research, building out custom tools for reviews data. I created a Chrome extension for GBP reviews sentiment analysis, but you know, you can vibe code and do it on your own as well. I'd say VibeCoding to do some sentiment analysis for your own customer research data or any other proprietary data sources that you have. And SparkToro.

In the case of US and UK clients, surely SparkToro is super useful for the correct type of audience analysis or audience research. Unfortunately, just for US and UK because it's not as valid for other geography and markets. Then I create a systematic a multitask type of prompts based on a methodology that I always use even before AI was a thing, which is a deep observation of SERPs and now also extended to a deep observation of other things like being LLM answer and so on.

It's funny because I know that we have this excess, I will say, of new AI-oriented tools that are pretty much trackers. Right? However, there are a few tools that I really love and I find useful that actually helps you to optimize, to better understand how your content on page and off page is identified or is assessed by AI platforms to identify the gaps and the opportunities to change the messaging of how your brand is perceived. One of these is YK, for example, and I love indeed how it takes into account not only Your own content, but third-party content as well?

I use the basics. So I use Ahrefs and Semrush, SimilarWeb. I think the prompt exports, they started off a little rocky, I think, with weird prompts that probably weren't related. But I think they're getting better. And it's always a good place to start to see what are the prompts. You can weed out the ones that don't make sense, but then with the ones that do, who's cited? And it's a good way to show a client like, here are the prompts and Notice how we're not there, but all of these social medias are, all of these other platforms are to show like that SEO is more than just onsite.

This is great. Yes. I have to say that indeed you can still use your already existing tools. Many of them also have newer features that connect better your content optimization actions with those topics and prompts that you're also tracking with them as well. I, for example, thank you very much for bringing up SimilarWeb. They have actually this new functionality to help you topic-wise for combining the prompts or segmenting, grouping the prompts into topics, and then tying those with the citations where the citations are coming and the relevance of the content that you can create or optimize that you still don't have on the gap versus the one of the citations. So yes, 100%, you can still use and you should still use the existing tool because most of them nowadays have some sort of AI-related functionality that connects with content optimization indeed. So I would like to start wrapping up with your last tip, tip or action or resource even that we haven't mentioned yet that you believe is key in order for a good content optimization process. I believe that you actually have already highlighted this, that I believe is key and is somehow overlooked. Cindy, I think that it was you like. A lot of people are still optimizing for those prompts that are generated by default by the tool that they are using to track AI visibility rather than actually putting proactive care to select which are those prompts and topics that they are measuring and trying to optimize for because it all falls from there, right? And unfortunately, if you're just Importing by default what the tool says, unless the tool actually knows a lot about your context. And I haven't seen any tools so far asking you about your unique selling proposition and a lot of the messaging that is critical to understand. It should be something that you should do by yourself. In fact, I have created a doc to generate prompts, a good sampling, good representation. I have a guide for that. I have a Google Sheet for that. But it's something that I highly, highly recommend if you haven't taken a look at, go and double check.

Let's talk about prompting. I'm not a big lover of prompting, but okay. Whatever tool you're using for prompt tracking, so use really crafted prompts that are really resonating with the buyer persona that are, that you are targeting as a brand, but not do the mistake that we were already doing in the past of just considering one prompt the answer of a prompt. Sometimes the most important information in the prompt, apart from if you are visible or not, is what are the possible potential follow-up questions the AI answer is nudging at the end. Because if we already are blaming Google for zero-click SERP, LLMs also are zero-click. to the square. And so try always to find some sort of repetition in this sort of follow-up question that you have in the prompt that you're tracking, because so you can foresee and start designing the type of conversational search journey your potential customers are doing. So you can fix depending on onsite, offsite, both The content that you are creating, your content strategy in order to be always visible because visibility is what pays.

We're not reinventing the wheel here. I think just bringing up EAT, like you said, Aleyda, is super important and having that sort of mindset going into it with content, you're going to be fine considering what the users, potential customers are actually looking for to trust you and, you know, trust you to make a purchase, trust you as a source. Taking that into consideration and understanding that the website is not the only source of truth, and consistency is really what's key here in getting those things cited. So using the website is more of a place to echo or amplify what is already being said and making sure that it is consistent across other platforms as well. You should be in a pretty good spot.

Also, please don't overlook— again, take a look at the citations, where the citations are coming. Don't overlook the content that lives in your support or in your community section. Unfortunately, a lot of this content many, many times is related on a subdomain that is not necessarily as well optimized. Think about how you can leverage those insights, that content, that information in your main copy so it's even easier to reach, easier to access, and connect better with what your user sees, which at the end of the day Is what will end up mattering even more at the very end of the customer journey to convert them, right? And by the way, speaking about multi-format, the importance and the weight of video. Celeste, you mentioned about starting your searches in TikTok. Have you seen the weight, especially in SaaS, for example? I work a lot in SaaS, the weight in SaaS of a lot of YouTube videos. Talk with your influencer manager, with your community manager, and see what they're doing with videos and influencers there in your own channel as well. And Cindy, I know that she had been speaking a lot about this particular topic, isn't it?

So I have a client who had a subdomain for a community for years and they would never let us touch it from an SEO perspective. And it was kind of forgotten about. And then when AI started ranking and Google started focusing on LLM and AIO kinds of things, the community just took off and started ranking like crazy. And I think it was because it was clear that it was a long existing community and it had real humans participating. Not bots. And so I would say like, if you don't have a community but you think that you might be able to create one, it, it might be worth a try because real humans engaging. If you have kind of a lifestyle brand or something like that that lends itself well, that content, you don't have to create it, but it just ranks and it gets your brand known for that kind of lifestyle. But then the other thing is when you find out that you're mentioned offsite on a YouTube video or a TikTok video or whatever, make sure that you are linking to it, that you're featuring it, that you're engaging with it, because you can help that content continue to rank and continue to get more visibility in the AIO algorithm, as well as in the TikTok and YouTube algorithms. So understand how all the algorithms work and support the content that supports you. Offsite.

Love this. So thank you. Thank you all very much. Cindy, Celeste, Gianluca, your insights were amazing, were super actionable. If you have any questions, please follow Cindy, follow Gianluca, follow Celeste to see how experienced and knowledgeable they are indeed. And they're also sharing continuously about this topic in social media and their own sites, besides speaking on stages about this at conferences in our industry. And don't forget, if you haven't yet, to subscribe to the channel. There are more episodes coming, going through more AI search optimization topics. And I want to wrap up thanking again Similarweb for our today's sponsorship. I really appreciate it. Until the next edition. Bye-bye.