James Dooley: Hi, today I'm joined with Ben from ISO which is AISO or get AISO. Pleased to meet you Ben and let's jump straight in. So I want to get deep into AI and LLMs because you have a lot of data with the tool of what you have. So, what are people searching and asking in chat GPT and other LLMs?

Benjamin Tannenbaum: Yeah, this is this is a really good question. Um, and I think a lot of marketers are wondering currently, you know, if uh they should really pay attention to to AI search or or if they do pay attention, they they want to know if it's really worth investing uh a lot of money in it. And and I think a related question is what's new really about, you know, search. Is it is it uh is it just you know different apps uh you know that uh produces a little bit of a convers conversational uh you know content but really the the most you know most of the most of the rules of SEO still apply or online marketing digital PR still apply and there's many components to this question but one of the aspect of it if you start from the beginning in terms of the user journey the the the search journey online discovery journey. The very first step is to know okay what are people asking AI search you know is it similar to the keywords in Google uh with a few words before and after but really very similar same level of intent um is it similar to what people are discussing on social forums like Reddit you know like Reddit is conversational in in structure because people ask questions and then people answer um so really are people just asking to jpd what they used to ask Reddit And and that that that that that's part of the answer to should you care about AI search because if people are asking exactly the same things probably the method to be successful for uh you know uh social media and SEO and digital PR and branding still apply um you know it's more likely to still apply. However, if people ask very different things, then you need to take a step back and wonder, okay, I'm address I, you know, am I addressing these type of questions in my content? You know, do I is there everything on my website, everything on all the places where I distributed my content? So it's an important question for marketers and what we do as an AI search company. Um we we we try to do things a bit different from a lot of other companies that are providing super useful analytics, you know, with a lot of uh charts and see how you rank for some prompts. Um there's there's some great companies doing this out there and there's quite a lot of them now. So we decided to start maybe from from sort of before that and to to start to think okay can we find information on what people are asking AI and there are kind of hacks workarounds like some talented SEOs uh you know some talented SEO experts have described how you can find you know longtail queries in your Google search console which are likely to have triggered AI overview um you can make a mix of what people asking Reddit for your specific products because it's likely to be related But we tried to find like a more direct way to really have like okay what are people asking change like for real you know not an estimation not a proxy and we we we tried many different things. Uh our previous startup was actually doing this kind of data collection at scale in in uh in real estate actually. Um and we tried to do things that we learned before in this other industry and we tried to bring them to online marketing. Um and one of the things that worked was basically a system where users might get like some benefits. So it might be additional features compared to Chip app or it might be access to better models. Um because 95% of people use CHP for free. Um we didn't realize that at the beginning but that's the case. And that means they have a very limited they have access to very limited models. Um so the first time they try a model that is actually something you get you get access to in the in the in the plus version. It's mesmerizing. It's like wow, you know, because it's so different uh between CHPD 5.2 thinking and GPD 5.2, you know, without thinking, for instance. So, we tried different things and then there is there basically some things that that worked and um and and and and uh now we have access to this data set where we can see real ch conversations and a bunch of other models like Gemini. And um without going super deep into it, there's definitely differences between the things we discussed before. So if you start with the the keywords the Google you know the things you can see in samos or orref or in your Google search console for your website it's different right uh I mean this might not be a surprise anymore because we kind of used to AI search now but we didn't know at the beginning and so the definite answer is that it's not the same thing as keywords plus a few words before and after it's quite different um to to to bring arguments to this and I'll explain to you how different it is, but to bring a nuance to it. Um the the most of the conversations are actually one shot. We were a bit surprised by that. So there's not that many conversation where you know it's like full back and forth with like uh more and more details added. That happens sometimes and those are super interesting obviously for brands. Um but uh most of them are like one shot. I've got a question, I get an answer, and then I basically move out of THP uh and I go to like a normal uh Google search or or or you know, um it's AI mode and then I click on one of the links. That's that's actually most of the uh the journey at at this stage of the of the funnel like at the top discovery stage and so it's most of it's mostly one shot. But then if you actually look at the questions, they are much more specific. So it's very rarely like one keyword, right? like uh you know uh best uh pizza in New York. Nobody asks this to AI anymore. And of course this behavior is starting to also propagate into normal Google search because as you know Google is introducing a lot more you know gradually progressively because they want to you know be careful with the existing business but gradually is introducing more and more AI. So people are actually more and more using the Google search bar as an AI you know search bar. So it's more and more specific. So you will see you know for instance in the in the restaurants the three most popular categories by the way in our sample are uh travel consumer electronics and um uh travel consumer electronics and everything related to beauty like skincare uh you know health adjusted beauty um and for those topics um so if we say travel like let's say restaurant as a you know I'm wondering about restaurant in New York the person will typically say Hey, what's the best restaurant in New York? But one of the person has like this uh food allergy uh or you know uh my price range is this this much. So there will be you know you basically going to be much uh much moreformational uh instead of you know the things where you start starting to trying to get like a links very directly much more informationational and much more specific um much more personalized you know in the in the because people are searching for things that you know applies to them. Um and They don't feel like they have to summarize it to one keyword and then add more specificity as they go along the search. They feel they can put all in and then the AI will take care of the, you know, of working it out for them. Um, but that's really a big thing. You know, it's much more specific. It's not necessarily lower in the funnel. It can be quite informational, but it's much more specific. So, of course, you can think about, you know, um, is your content, you know, addressing all kind of, if you're a restaurant website, is your content addressing all kind of allergies? do you have a specific page for them etc. Um so that's that's um that's one really clear aspect. Um and now maybe I can go deeper into this maybe or you you let me know but the second thing on you know the second hypothesis we had or the the question was is it similar to what people asking on social media which is a fair question because media like social media is quite conversational in nature already especially things like Reddit. Um and the interesting thing there is that um and it's one of our users who put it this way is essentially people are share a lot more with AI than they would share publicly including in platforms like Reddit where they can be anonymous. Uh people feel like they can ask the GPT um things that are very intimate. Um so you know if we stay in the the case of the restaurant they they they won't be shy to share like all kind of you know sensitivities they might have with or you know any kind of any kind of budget constraint that they may not want you know their partner to know about but they still want to you know include it in the in the decision making. So, so that's going to be very different from things you see on on LinkedIn for instance where people share but at the same time they're also marketing themselves to the world. So they're careful not to expose their you know what might be perceived as uh you know awkward uh or or or inappropriate or anything of that sort right and and one of our users said okay basically chipity users are sharing more and this is particularly true for anything about beauty and health they're sharing more with chipity than with their romantic partners uh so so that's another reason why it's very different from you know uh uh both Google Google keywords and and uh and uh you know for social media was even though they're conventional in nature um so that's maybe the high high level picture I'll add maybe one thing which is that um most of the conversations are one shot but they are some very interesting conversations where um and I did a little video about this um that I posted on on on LinkedIn and YouTube where essentially uh it's in the topic of travel still um and you see someone asking about the trip to Italy And they ask they start by asking uh CHBT to help them with the destination. So actually picking you know the the places to go and then for one of the destinations they start to inquire about specific uh places for food and then it goes further and further down and at the end they ask for like you know specific budget and contact details. So it's fantastic because that's completely new you know compared to the world of Google. Um and we came up with a concept called like search density where really what's happening here is that in one chip conversation you have 15 Google searches. Um and of course what's interesting also is that they go lower and lower in the funnel. So you start by something veryformational. There's no brand you know it's like trips in Italy with some just like cities and days and then you start to ask about questions with a lot of specific actually that person had celiac you know allergies. Uh so you had like some specificity specific you know request and then you start to see some brands appearing in the answer but just at the kind of second stage and then the person ask for budget and contact details. So now you're like really done in the funnel and it's it's basically like 80% likely to to lead to transaction at this stage. This is fantastic for brands and it's very new. You know this this whole funnel in one place uh with such a high level of search density um this this is something completely new I think.

James Dooley: So with regards to let's say the search density of what you call it. So there's people getting from the top of the funnel down a lot faster down the funnel within one thread of going back and forth with let's say chat GPT or with Gemini. And what you're saying is that the queries now seem to be a lot longer where maybe ahrevs and seamrush might have came back and they might want to be likeword type queries. Now they're almost conversational, like longer sentences with like, I'm looking for the best restaurant in New York that does pepperoni pizzas and for this specific type of allergy that is under £50 for a partner or something like that. Um, what do you find that people then need to do who's a content marketer differently to what they're doing? like do they do they need to be doing more semantic content, deeper content, more pages? What you finding that they need to do to try to win top of the funnel and be able to keep get them from top down to bottom within the LLMs?

Benjamin Tannenbaum: Yeah, that's that's the the the million-dollar question and it's it has like multi multi-dimension and the there's a bunch of steps and again I find it useful to look at the sort of customer journey to kind of answer those questions and the the there's different things to do depending on where you look at whether like in terms of development of the user journey the the first um step as we said is understanding what people are asking and how it's different. And of course, once you have an understanding of this, the first thing to wonder is, and you know, it's it's a bit like uh when you were deciding which keyword to go after, it's a little bit of a strategic decision to think, okay, um the more specific I go in my content, the you know, the the the more long term I will go and and so the, you know, the smaller the volume of request I'm going to I'm going to hit. However, uh there's a trade-off, right? So, like if you hit like very specific request, the volume is smaller, but they're also like much more likely to convert. Um and in the world of CHP, you so much more likely to be picked up very simply by or and similarly for Gemini because if you answer very clearly and in a very relevant way that specific request and if you appear in the space of potential options to find the answer, which we can go deeper into, but if you if you're part of this sample and you ask her and you answer very in a very relevant way, something very specific, you're very likely to be picked up in the answer. It can be picked up at different levels as as a source or or straight in the answer that's that's that that depends on many factors. But of course, if you target those, you know, uh if you really hit that intent like in the nail, then uh you're much more likely to be picked up. Now, does that mean you want to have like 10,000 pages where you say, you know, my um how, you know, I'm how I have a good pizza for celiaak and lime disease and people don't eat nuts and people it's all of this. um knowing that your website is also read by human and you also want to you know satisfy all the constraint of classic SEO you know you you have kind of to to find to find where you want to sit in the in this trade-off and and probably probably create additional pages without going to you know parasitic things where you're just trying to cover too much but you may want to you know hit certain certain things and in most cases in many cases it's things that um people didn't really bother to explain on their website because nobody was asking Google about So sometimes it's just like updating content that already exists to just make sure you cover a couple of more specific uh you know requests and and that should be quite good to start with now and that's also relatively it's manageable in terms of budget because it's your website you can change it. Um now of course you know if you want to go further there's a question on where do you distribute your content because you know those AIs are also finding content they're finding content on your website and and and you can actually estimate depending on u the sort of the footprint of your brand um how big of this it is in in the sources but it's al also going to find you know information in uh third party to to call it that way uh which are which can be your competitor's website and can also be things like social media like Reddit and you know LinkedIn YouTube, you know, all of those things and also the classic digital PR channels, you know, like uh media that has like, you know, good authority. Um, so, you know, you can think of similarly to starting from okay, now I understand what people asking CHP and my website is covering those requests at least for the ones that are high priority. You can then decide, okay, do I want to distribute this content to to to uh to the rest of the world? If I think it's super important, you know, if I'm if I see like a request that is very specific, but I'm also also have an edge on it, right? Like I'm the only one who does pepperoni pizza um that is not spicy. Okay, I'm the only restaurant in New York that is doing non-spicy pepperoni pizza. So, I just want to, you know, really really double down on this because, you know, I I estimate the return on investment is going to be worth, right? Um so, I think it depends on your budget. It depends on, you know, how how how much you think you're going to get from it. But I would start by basically making sure your website is covering the slightly more specific request for your more popular pages, you know, but with some updates. I think that's an easy quick win and then you can think like deeper, you know, after that.

James Dooley: Yeah, for sure. Well, Ben, it's been an absolute pleasure. Um, for anyone who's listening to this, this is the first part in a series. Make certain that obviously on this is now talking about how people and users are searching in chat GPT Gemini and other large language models. In the other series we break down and I dig deeper into is GEO just SEO and he's got evidence to back this up. We dig deep on query fan out and what is query fan out and why is that important for SEOs to make certain what content marketers are doing. And then the last part of the series talks about the personalization of large language models. These are all in different episodes. Make sure you check them out. All the links are in the description. Ben, it's been an absolute pleasure.

Benjamin Tannenbaum: Likewise. Thank you.