GEO? AEO? LLMO? What's with all this AI SEO stuff? | Patrick Stox

And we are live. Welcome everyone who has tuned in today. For everyone who has joined the webinar, you can ping in the chat where you are where you're joining the live stream from. What country? I see that we have 66 people joining us here today. That's amazing. That's awesome. Miami. Wow. I think it's quite early in there. Welcome everyone. Yeah. From all over the world. Wow. Awesome. From Malaysia, Ukraine, France, Venezuela, Canada. That's so cool. Okay, we're just going to go ahead and get started here. Welcome everyone to today's HRS webinar. I'm your host Constance and today our webinar is about geo AO LMO. What's with all this AI SEO stuff? Well, Patrick is here to help make sense of it all and what you should be focusing your time and energy on if you want to come out ahead. Patrick's our product advisor and technical SEO at HRS and he has been churning out study after study if you check on the blog on AI overviews, generative search, AI mode, all the different AI related stuff, monitoring how each change, announcement and finding is shifting the SEO space and the marketing landscape as a whole. So while he goes on with his presentation, if you have any questions for Patrick, you can ask him as he presents. But if it's like a deeper question you want him to explain something more complex, please hold your question until the end and then we'll do a full full Q&A and then we can spend more time on these like, you know, longer questions. Okay? And without further ado, I'll hand over the mic over to Patrick whenever you're ready. Patrick can begin. Yeah. Hey everyone. I see some friendly names in the chat. Josh, hey man. Uh yeah, I think uh we've been trying to figure out like what to call this. Um it's going to be interesting the shifts that happen over the next couple years. Uh and we're going to have study after study. Uh tons of data, tons of tactics for you. Uh all this might change though. It's going to be interesting times. First off, I just want to say join us at HS Evolve uh San Diego October 14th 15th. If you're in the US or even if you're out of the US, like come over. I went to the one in Singapore last year. Amazing time. I think even a lot of the speakers were like, "This is the best conference I've ever seen." Y'all probably know who I am, but just a bit about me. I'm one of the authors for the Href's uh SEO book for beginners. I write on the blog. Again, lots of data studies. I'm also on the product team here at Href. So, if you ever have any requests or anything productwise, feel free to reach out to me. Uh I was the lead author for the SEO chapter of the web almanac in 2021. I was a reviewer in 2022 for that. I was also the technical review editor for the art of SEO fourth edition. Um, I'm mostly known for technical SEO, but believe it or not, I think these days I mostly focus on content, programmatic and product work. Uh, less less in the technical space. Now, I do have a tech SEO Slack group. I moderate tech SEO uh on Reddit. Uh, I organize a tech SEO conference, techo connect and the triangle SEO meetup. And just a fun fact, I actually defined the role of SEO uh technically search marketing strategist. That wouldn't fit there. Uh for the US Department of Labor, uh the reason for that was to actually get it in curriculum at colleges. Like there you couldn't get a degree in SEO, digital marketing, PPC kind of stuff. And this was God, I don't even know how many years ago. I'm old. Sorry. I was When was that? Probably like 2013, 2014. So, more than a decade ago, I ran a poll, couple polls, trying to figure out what do we even call this now. Uh, seems like everyone wants to go with search everywhere optimization or still search engine optimization. Uh, I was really hoping EIIO would win, but no, too bad. That was going to be fun. What I have settled on is SEO for a this is SEO still SEO for AI search and AI assistance. AI assistants are actually the term that like chat GPT and all of them are calling themselves. So I'm going with that. Now traditional search engines Google fell below 90% for the first time uh in uh many many years. They absolutely dominate. What's listed for the AI assistant market share is that chat GPT is dominating uh nearly 80%. We see this in traffic too like they send a ton of traffic to the websites. I think that's actually misleading because if you count like Google AI overviews and now even Google AI mode is rolling out. uh it you know the usage statistics for it I I'm trying to remember like 1.5 billion a month or a day I can't remember the exact numbers but I I did look at the numbers when I built this chart and I would say that chat GPT is a fraction of the actual AI assistant market. I think AI overviews and now AI mode are going to be much more dominant and like I said we have AI mode now that's a whole new ball game I think we're in three countries now the US India and as of what yesterday uh the UK so that's awesome uh this will be global this will probably be the default search within a couple years uh if it even takes them that long to shift it ever traffic from all of these is growing. Again, we can only look at not all of them because you can't break out like AI mode. Uh AI overviews are messy. Like Google's not really letting us break that data out in search console. But if you look in analytics at all the traditional or the the newer AI sources like chat GPT, perplexity and all the traffic is growing. This was across 82,000 websites and it like 10xed in a year. That is crazy fast growth. Uh most of that is actually from Chad GPT. Again about 80 some percent of the traffic comes from Chad GPT. Now that doesn't mean that you should ignore Google because right now like in 20 at least in 2024 Google search was still growing like 20%. And according to the study from Rand chat uh it was 373 times more searches than chat GPT. So while they're growing they're still a very small amount. Do y'all see that? Like the 14 billion in Google versus the 37.5 million. It's a small fraction of search right now. I think SEOs have their attention wrong. Uh it's good to plan for the future, but they're focusing on what is less than 1% of the traffic and just a fraction of the searches. And I don't know that that's going to work out so well. Google's still sending a ton more traffic to websites, too. 210 times more traffic. So you got you got a bunch of SEOs focused over here on the new and shiny object and here's the thing that is like still driving pipeline revenue. Uh now again I think like it deserves a little more attention because it is futurep proofing. This is the way things are going. Um but I think right now priorities are kind of backwards for a lot of folks. The AI traffic is also a bit under reportported. So when I say it's like less than 1%, h maybe we'll we'll see. Uh we don't really know because they're not passing what's called the referral value. This is what lets you see like this came from google.com. Uh a lot of these are not passing the value for that. Now I I ran this study when was this like April, early May. Chad GPT actually fixed a lot of their referrals. is one of the reasons their traffic is like increasing suddenly or you're seeing more traffic from them is they actually fixed this so you're act they're getting the credit for it but what I want to say more than anything is keep calm and SEO on like right now this is still SEO you'll see as I go through this a lot of the things that are working for uh I I don't even want to call it geo sorry I really don't but a lot lot of the tactics are the same. A lot of people are already successful in the new AI assistants simply because they were good at SEO. An AI search is still search. I think it'll take SEOs a couple years to figure this out, but they will get there. Uh this is just the new normal. It's a change. Things change. SEO is not dead. Search engines are not dead. They're just different now. And we're even seeing more a mix like I don't know if yall have seen the new uh perplexity uh in the browser like it it's like a couple almost traditional search results a paragraph a couple traditional search results a paragraph. Uh they're becoming more blended and I don't know if the long chatbot answer I don't think that ever is going to win out. Um some kind of broken up thing like that maybe. Uh, I think it's kind of a good mixed experience, but a lot of them now, right now are also just only uh text content. What about images? You're starting to see some shopping. Uh, videos, not so much. Uh, all this stuff is part of search. YouTube's still the second largest search engine. Videos are important. So SEO is the worst it's ever been and it's still your best marketing channel. I stole that from Ryan Law because that really resonated with me. It's true. It's harder to attribute. The traffic is going down. Um and it's still there's no other channel that's as good. You're not looking at so many handwavy metrics, your billboard impressions, that kind of thing. like you can still track this to revenue. Now, you've probably all seen the charts like this, the great decar decoupling, which was coined by I think Darwin Santos, uh where impressions are going up, clicks are going down. It's the AI overviews that are causing that. Uh this is actually a site where there's almost no AI overviews and there's no decoupling. It's great, right? The AI overviews reduced clicks by 34 and a half% when we looked. I actually want to run a bigger updated study on this and see if that's changed, but I I think other studies are in line, so I don't know that the number is really going to change that much, but that's a big click-through rate drop. They're on about one of all queries now, and that number just keeps growing. Every time I go and pull this chart, it just grows more and more. Uh I don't know where it's going to end up. You know, back in uh early March it was almost nothing here. Like I mean not nothing but like now we're like four times that amount, five times that amount. They're growing crazy fast, which means yeah, clicks are going down. It's a bit offset though, uh which I'll show you in a minute by um some of the growth in the Google queries. like the study from Rand earlier had like 20% year-over-year growth in Google queries. Uh these are taking over other SER features. So featured snippets are going down as AI overviews are going up. Uh I think featured snippets will eventually go away completely. But it's not just featured snippets, it's lots of search features are actually like going away. shopping, uh, a ton of stuff. Now, they mostly show the AI overviews mostly show onformational queries for now, but I think this is going to shift again like we're seeing more shopping already. So, this commercial here is likely to go up much higher. Uh, I don't think it will go much higher like for navigational queries, but yeah, transactional commercial, you're probably going to see this this number explode. You're also not seeing it too much on branded queries. It's mostly unbranded. Uh, and it's not really that much on local yet. I am seeing more stuff in local. And I have a feeling like I've been using it more for local myself. Uh, or at least sorry, AI mode. I've been using AI mode more for local. So, I think like AI mode might be the the version that actually eats into the local pack. Uh it's really good actually recommending like restaurants and stuff. Kind of amazing. I've been using that a ton. Uh AIO's mostly show for longer search queries than the traditional search query. So people are typing entire sentences, paragraphs in Google used to have this hidden limit of uh 30 was it 32? Yeah, 32 words you could search before it actually cut off your search engine or or your search request. They got rid of that. So now if you want to type three paragraphs in there, it seems to take it and it'll come back with an answer. That is not the way it worked. Uh well, I don't know exactly when they made this change, but I think within the last year or so, cuz that uh that 32word limit has been there 15 20 years or maybe more. uh it it was always there as far as I know like as as far back as I can remember and then suddenly one day I was just doing this search and I I copied something I pasted it expecting it to be cut off and I was like that's not cut off anymore. That's interesting. So it like took an entire paragraph as a search. So no more 32word count limit. Going back to the great decoupling, we looked at this across 800,000 sites. You'll get a sneak peek at this because I haven't published this article yet. uh still going through review. One of the things I wanted to note though, the clicks when we looked across 800,000 properties here, the clicks for the year are actually up. So a lot of people are like, well, you know, impressions are going up, clicks are going down. Again, Google search itself grew. So clicks are like roughly stable or even up for a lot of websites in that time period. It's not that they all went away. The problem is as the decoupling becomes worse, like you can see after this data anomaly, that's when Google stopped sending data for a few days. So when I pulled this, we hadn't fully backfilled this. I think we have now, but when I originally pulled this, like um we we saw a huge increase in AI overviews after that. And you can see the impressions are starting to spike and the clicks seem to be either either staying about where they were going down. But more AI overviews means probably more lost clicks for queries. And it seems to just be getting worse. And AI mode, unfortunately, will be even worse. Uh I like I never click the links in AI mode. Very unless I'm testing something, it just doesn't happen. And I can't imagine I'm the only one that's that way. Uh same with with chat GPT. like we know the usage, we know uh all the things that show and when I did some data modeling for this uh based on market share and queries and and the traffic we're seeing in in analytics and everything, my best guess for that new normal is our our baseline clicks are going to be roughly about 25% of what they were. And I think this is going to happen over a period of 2 to 3 years like when when these systems are the default when there's no more Timble blue links. Uh that's why I'm really hopeful that something like Perplexity's new system will actually, you know, influence and and find that balance of like here are your kind of traditional links versus like let's just put stuff on the side or in this little uh footnote thing that no one clicks because that's not going to be good for anyone. There are entire website models that rely on this. Recipe sites, news publishers, they need that traffic. If your new baseline's 25%, you're going to see a lot less of those kinds of sites. They may just die off in general. Uh, but we're not going to go report on impressions. I mean, you can report on impressions, but clicks are still a better metric. Even if that baseline is lower, that is still a better metric to report on. Even better would be things like um uh you know any kind of conversion metrics or revenue metrics that kind of thing which I think is yeah sorry my next slide so revenue SQL there is a question in chat from Hans Graard sorry if I'm butchering your name uh he asked like what's your take on what's going on when Google says AI mode can make clicks go uh clicks go up when what we see is that clicks go down. I you know I I want to see their data on that because I don't believe them. Um they're saying you know the tra Yeah, they're saying things that third party studies are not able to corroborate. Uh now I do believe they are sending traffic to a ver wider variety of pages and maybe they're seeing that as a good thing. People are using it, searching more, adding follow-up queries. Uh, so to them like that looks good. Like it maybe their measure of success is that I want more people to search on Google. Well, job well done. But are people clicking through to the sites? No, it's not happening. So, I don't know where they're coming from with that. I I am still waiting for them to actually say like here's data on that because I don't believe them. There's no way. I actually have some some slides in here about like quality metrics and stuff that they've talked about and um yeah, the data doesn't doesn't align with what they're saying. Uh yeah, better to report on anything clo as close as you can get to revenue, leads, that kind of thing. Uh those are going to be better metrics. Impressions that a lot of people are like, I'm just going to report on impressions now. No, you don't get buyins from impressions. It's going to mix like yeah, billboard impressions, pay like a lot of channels relied on that. We can still do better by having better attribution and that is what we should do. But don't panic. It's not all bad. Uh, a large portion of these AIO reviews are on non-monetized queries. So, a lot of the queries you probably care about that you're willing to pay companies are willing to pay ads for, uh, they're not on those. It's awesome. They're also not always in position one. We're we're seeing an increase in this. Uh, I'm kind I what I suspect is that the AI overview answer is during the process called re-ranking where they like reorder pages, they're actually having to compete for their position. So they they like bid uh every result kind of has a bid for its own position based on like how good the answer is and some other factors and stuff. And what I think is happening is that they're making AI reviews actually bid on where they rank. So they're not always number one now. I think if you have a good enough website with a good enough answer, you know, lots of links, blah blah blah, you can actually rank ahead of them. Uh right now it's a little over 8% of queries, but we're again we're seeing this number grow. And I don't know if this is just a test though, they might roll it back and be like, "No, we always want AI overviews in position one." But right now, they're not. We're seeing them as low as position six. We haven't seen any in the data set lower than that yet. The traffic from the AI assistance is also very valuable. You know, I wrote up this uh I I actually looked at the HRES data and wrote this up. Half a percent of our visitors drove 12.1% of our signups, 23 times higher than our conversion rate from traditional organic search. Now, I've seen other numbers that put this at like four four times or like 11x. I haven't seen anyone that had as high as 23x, but that is what our data is. uh tons of people are converting from this traffic. I think they're getting all their like they're they're doing their research, refining queries, etc. before they even come to our website. They go to a bunch of different types of pages, too. Uh I'll have more data on this, but there's a couple things I want to point out here. Like they go to our free tools like crazy. Um, stillformational content is important, but the real question is how does this compare to the regular traffic, the normal organic search traffic? And it's this. We actually get way less traffic from free tools than we do in traditional organic traffic. So, while they look good, look like a great segment, it's actually underperforming in the new age for us. Um, programmatic is also down. like a the AI stuff is less likely to site programmatic. I Ryan Law just did a big study and uh most of the content that's cited is already like two three years old. Uh so a lot of programmatic is newer because yeah AI hasn't been out that long potentially lower value pages too. So while we were getting really good traffic from a lot of programmatic in the past that may not work so well in the future. Uh the things that are working well again product pages, homepages, uh that's great because I think again people have already done their research. It's not that we're like optimizing our product pages or homepages. We're just being recommended and people were ending up there. Um informationational pages are actually doing better as a percent of the traffic. I was super surprised by that because I thought like theformational queries would eat all that traffic. Uh so we're still getting a good amount. It's it's potentially overperforming compared to the traditional organic search, but the baseline is still still much less. Uh, hallucinated pages are also an interesting one. That was I think 3% or so of the p the traffic we get were to hallucinated pages. So, they just made them up. And that's mostly just Chad GPT with their non um non-web search version. Another thing I noticed is that international pages perform poorly. Even when you're searching in another language, uh the English content is more likely to be cited in results. Now, I think this is because the systems themselves, the LLMs, like they work across languages, and if they're just looking for a good page to site, maybe the English pages are stronger in general. Uh, but I really think they need some kind of system like href lang to properly swap and show the right pages for people's queries because right now that is not happening and that makes a terrible user experience in my opinion. Uh, going back to what we talked about with some of the traffic, it is not better quality or may not be better quality. You know, search engines are saying it is better quality. Yes, we saw more conversions. Uh, as far as like more clicks and stuff, I don't believe the search engines there. But for other metrics, like we've got now less pages visited from by AI, higher bounce rate, um, it actually won more time on site, like the AI searches won with the average time on site. So, people are spending time there. Uh, but again, lost like average pages per session. So, it's a it's kind of a mixed bag. They won one metric. I wish we had more data on conversions, but when we ran this study, I think we only had conversion tracking in uh HRS web analytics for like two weeks, so almost no one had set it up. Uh but this is data across 82,000 websites at the time. The traffic didn't really look better quality. Another sneak peek. Uh I thought this was interesting. And we were looking at the kind of like the uh different segments of the mobile of the search traffic. And traditionally for about a decade now like I think uh mobile search has been the dominant use for search. You know we had uh the whole like shift to mobile first mobile friendly blah blah blah. Uh it's not that case with LLMs. LLMs are still desktop first. It's it's you know 65% on desktop with only 34% on smartphone whereas on the traditional organic traffic it's 76 and a half% on smartphone. So a massive massive change here. Uh if I you know we've been we've been focused build your website for mobile blah blah blah. Well, now it's looking the opposite. Like I would I would be looking at that desktop version hard again by page type. Uh this was also interesting. Lots of AI visitors are still going toformational articles. Again, I thought this segment would be dead with these systems. I almost never click. I just ask like three more questions instead of clicking through. Uh, but for more in-depth queries, I still click through and I think that's what this is showing us. Um, there's there's still a ton of traffic there going to theseformational articles. And then core pages, that's like your product or service, whatever is being recommended, your company. Now compared to the traditional organic it's still it is less uh but for theformational type articles but you are getting sorry the AI search is getting less traffic forformational articles than the traditional but it's still getting traffic. It's not as bad as I thought it would be. The big shift I think is the the core pages, your product pages, your homepages, your service pages. They get a lot more traffic in the AI search by a percentage. Again, all of these numbers are going to be way lower in AI search because they're a fraction of the search and I don't think people will be clicking them as much. Every one of these AI assistants is also a bit different. Uh when I was looking at the overlap of citations, what I found was there's not a ton. I think there were seven sites uh that would showed up in the top 50 for all three of these. And I think a lot of the reason there is they have deals with different publishers and different syndication networks like they they've basically bought rights to data and are biasing towards a lot of those. So, Perplexity saw a lot of more uh international websites. Chat GPT lots of stuff that they had licensed for news and stuff. They were they were defaulting to site those websites. It's it's not going to be the same. like we've just mostly focused on Google and yeah being get some traffic and blah blah blah but like we might need actually different strategies for uh each different system and this is like the percent that's being cited for each one they have different biases I talked about this a little bit one of the interesting things though that I saw uh all of them despite Wikipedia having a getting a ton of traffic and a ton of citations from them. Wikipedia is actually being biased against like they actually should be showing more. Um we ran a big study uh across like 75,000 brands and the ranking factors the factors that correlate with the appearance and AI overviews they changed from traditional search. uh you know we used to see longer stronger correlations for links DR that kind of thing now it's all branded branded web mentions branded anchors branded search volume your brand matters for the AI overviews a lot but that's Google uh chat GPT showed a very low correlation when we ran the same study perplexity kind of in the middle uh Google since I don't even No. Uh I'm trying to remember. It was one of their former he CEO or VP. I think it was Eric Schmidt. Uh years and years ago, uh people were complaining about Google search results and that they were full of spam. And he actually said like brands are the solution, not the problem. So I think Google has a lot of systems in place to help clean up the results that these other ones haven't had time to build. they they may end up building something similar in the future, but right now Google bias is heavily towards brands. The others not so much. And then um the overviews are also like citing mostly websites that rank in the top 10. Again, heavy overlap in Google, heavy overlap in things that have traditionally worked with uh you know, traditional organic search with with SEO in general. Uh we actually have a breakdown by position. So again, you're if you rank at the top like higher organic rank, more likely to be cited in AI overviews. Another breakdown. Uh, I wanted to talk a little bit about how LLMs work because I need to explain some of this to be able to talk about what I want to talk about after that. Uh, they their next word predictors. So, all they're doing repeatedly is looking at what's already been said, context surrounding words, and then trying to determine like what what is the probability that the next word is this one or this one. So, they're over and over again, they're like, what is the next word? What is the next word? What is the next word? Because of that, they flatten originality. We've been telling people uh and search engines have been telling people unique content, unique ideas. Uh that doesn't really work with LLMs. If you're the only one saying it, you're not likely to be in that pattern to be able to be predicted. So, like if you're like um if you're saying this and you're the only one that's saying it and other people are saying the same thing over and over and over again, that's says something the opposite of what you say, the LLM is likely to say what the other people said. And that really sucks because again, this is this is what we should do. You should create unique content. You should uh provide valuable insights. It can still work, but until other people actually start repeating you, start saying that, it's not going to be in the response, but you still need to do it now if you want to be in that response later. So, I'm not saying don't create unique content, unique ideas. Please do it. But right now, the way that LLM's work, they just don't reward that. The probabilities in all this are randomness and you get multiple rand you get randomness on top of randomness on top of randomness. It's great. You get the LLM training data. So all the data that went in there from web pages or PDFs or potentially social media, any deals like Reddit that they have, all that is one set of probabilities. with rag models. Those are retrieval augmented retrieval augmented generation pages. Uh basically they go and pull a bunch of web pages before they create the content. And now you've got the probabilities from the content on those pages mixed in with the training data probabilities. And just to make it more fun, they throw in what's called temperature. And that's just to build in randomness. Uh a lot of them use somewhere like 7.8. So they just throw an additional randomness on top of the probabilities from the training data and the probabilities that are created by the retrieve pages. So every output becomes different because it's randomness on top of randomness on top of randomness. That's why if you're like tracking something and you just ask the same question again, you may get a different answer. Tomorrow you may get another answer. Um it's really interesting the way it works. Uh and again I think this is why like they can say that maybe site more sites are being cited because now they're pulling from other things. Uh they're actually even uh actually I think my next slide no um I was going to say uh there's a concept called fan out queries that it's probably a couple slides away but I'm going to talk about it now anyway where even the the queries that they're saying like let me let me search these five other things when I'm retrieving these pages. Uh what happens is they change those even so now you've got like random more randomness for the retrieved pages because like they're changing the queries that they find important. Uh but the these rag models almost everyone is using these now. Uh that is still search and there's where I'm talking about query fan out. Finally had that a little out of order I think. Oh no I just got off topic. Um the the way I describe query fan out to people is topic coverage. But like I say like it's pulling a few extra things usually when it's going out and retrieving these pages but this isn't consistent. Every time they're doing these fan out they seem to change the queries that are behind them and that changes the pages that are retrieved. That changes the probabilities in the response. So you get a different response, different pages cited, uh, and again a greater variety of stuff on the web that is potentially being cited in these, which might be why they're like this is better, but n people still aren't clicking. There's also a concept called chunking, which this is uh this is actually a breakdown from our AI content helper. Uh this is very similar to what we do in our search engine which is yep.com and this is how SEO should be thinking about this is basically what content is related to each topic. It doesn't have to all be together in your content. That is not the way that search engines read it when they create these chunks. It can be from multiple sections, multiple different paragraphs, that kind of thing. Uh this is more again on like the retrieval end. This is more on the like the scoring for information retrieval, the search side. What SEOs are doing is this. They're just basically like headline, couple sentences, headline, couple sentences, more headline, couple sentences. Let me throw in some bullet points. These are very annoying pages to read. They're not really going to rank better. In fact, they just look like spam. They look like something I don't want to read. Uh they're pretty okay for like skimming if I want to find like one thing. But like you wouldn't put 50 FAQs on a page. Like usually you see five, six FAQs that are useful, not 50 different things. It's like here's a thing, here's a paragraph, here's a thing, here's a paragraph. That is an annoying user experience. I hate it. Please don't do that. Uh, I don't think search engines will reward this. They may right now, they will not in the future because yeah, the user experience signals for that are just going to be terrible. As I said on the search side, it really is more like we figure it out. Um, so each topic here is different color, you know, shelf life of coffee beans. You got a little bit up here, a little bit down here, more down here. That is the way the chunking works. It doesn't all have to go together. Uh, and it's it's not the way that it's going to be read anyway. So, if you got even if you've got that 50 FAQ type setup and there's like three different related things, different parts, like that's still the chunk that's going to be retrieved and you're going to be measured against. And on top of all this, you've got personalization. So there's an entire history of things you like, things you do, companies you bias against, uh, all that is being stored now. And your results can change. Your results are probably going to be different from my results. If I'm searching about like SEO tools, mine are probably going to heavily favor hrefs, of course, whereas uh, some folks might favor like Simrush or Moz or whoever they're using. It depends on like, you know, if if I've asked a thousand questions, processed a bunch of hrefs data, if I'm asking about like what tools are best, it's probably going to tell me hrefs. It may not tell, you know, someone without any of these biases the same thing or someone with different biases the same thing. So more not really randomness, but more more things that look like randomness in that. These AI bots, they're going fast. All the crawlers are taking off. I think there were what 10 or 15 new AI bots last year alone and they're starting to crawl the web fast. They're also new and they're breaking things. Um, most bots when they start out have a lot of problems. Like they follow things they shouldn't. They they waste a lot of time. They waste a lot of server resources. Uh I'm curious if they're going to fix this or they're just going to be like faster or they're going to be blocked by more websites because we did see like they're they're starting to be blocked by a ton of websites. Uh some of that's like news and because they want to push licensing deals. Others are just because of server resources. They're taking down websites. Uh they're over overloading servers. They're they're using all the the resources and then you run into the monthly limits on some hosts. Uh, but it's it's going to be interesting to see how that develops. They don't render JavaScript. None of them yet render JavaScript, I should say. They may do this in the future. They may not. And the reason I say that is because like for the LLM, the training data, they may not need the pages that are rendered with JavaScript. Your so I'm seeing a lot of SEOs actually say if your content is not rendered by JavaScript, it won't be found. It won't be cited. in like AI mode, J GBT, etc. That's not accurate. Uh what can happen even if you're not in the training data, your content can be re retrieved by those rag models, which can be Google being a number of search engines. I think perplexity is now like building their own. Uh so if they have that content already able to be retrieved from like a Google index or a Bing index, I think we saw something recently chat GBT might be switching from Bing to Google. Then when they're retrieving the pages, retrieving the content, they get it from there. So even if the bots themselves aren't rendering it, you're a you might still show up. You might still be cited because they can still retrieve your page. They can see the content that was there that wasn't actually rendered. So I don't know how important this is actually going to be in the future because I don't think they're going to go back away from uh the augmented search. And yeah, like I said, they're they're all augmented by someone Google being uh potentially brave. Yep. A number of different ones. The probabilities are all like they're they're the randomness in all this. And I don't know like how that all is going to work um in the responses, but it's it's if you think about trying to monitor these search tools, all the all the tools that are being built to monitor AI search. There's so much randomness in all of these that I don't know how valuable it is to try and just choose one number or say there's one probability. Something you'll probably see in our brand radar is that we will try and split this out and say like this is the probability potentially from the training data. This is the probability from retrieve pages. Additional again like random probability. Um all of that will kind of impact how you need to go about doing things. So, if like if if you need to impact the training data, uh you basically just need to go on a bunch of websites and maybe wait a year, um you know, build a bunch more content, get it indexed, and potentially wait a year because right now I I think the update schedule is going to get less and less. Like for a while, it seemed like there was a new model every few months. Uh but now I think again I think they're getting longer, they're getting better, they don't need to do as many updates. So if you want to be have your what you say repeated, you have to be in that training data or the retrieve pages or both. Like it's it's kind of the way it works. I'll talk more in that in a minute. Uh LLMs.ext. I'm not a fan. If you've looked at my Twitter, LinkedIn, anything or X, whatever, uh I don't think this will be used. This is not a protocol um that's supported by any of the AI assistants. It's not used by anyone yet. Uh and it's got a lot of issues and I mean a lot of issues on top of right now just being like textual content, not images, videos, etc. Uh there are some actual like potential abuse vectors for this. And I'm recommending people not use this, but people seem to still be pushing it. I think it made its way into like Yoast and Rankmath and some WordPress plugins. And so people are pushing it forward even though it's not used. So it's a waste of time. Uh but at the same time, it's dangerous. And I'm going to have to write something about that up. And unfortunately, I think I'm going to have to tell people this is how you abuse it so that people actually won't adopt this. Um which sucks. I don't want to do that. uh we don't really like to tell people like this is what you should do to manipulate things but it is what it is. Uh schema is another thing that I think is overblown. It's not again these are the LLMs are next word predictors. They're not like trying to figure out all this stuff the schema relationships blah blah blah does not go in the training data. It works the same way with with search right now. Like it's mostly used for search features. So your all your shopping stuff uh uh breadcrumbs blah blah blah like all like there's a whole search gallery with the 20 different types of schema that Google uses and the all the different features it powers. And it's going to be the same with the AI search stuff. AI assistants are in their black hat era. Despina from our team wrote a whole article on this. Um, she didn't go into some things that I might though. Uh, it is really the wild west out There's a lot of spam. You parasite sites, hacked sites, expired domains, Reddit inquir spam. Uh, I've looked into so many examples just trying to figure out like how does all this work and like why did this this result change? uh or like why is you know all this completely different now than it was a month ago and oftentimes the answer is spam. I hate I hate to say that but it is uh that is not sustainable though. I I'll tell you more in a minute. There's data poisoning. People are actually putting false info about competitors in in these systems. So, they're leveraging their networks or or like their websites, their social media, their uh affiliates, their influencers, um or just spamming again like the parasite sites and stuff and actually putting false info about competitors out there. And this is crazy to me because there are going to be some interesting lawsuits related to this. There's a ton of legal liability for saying things that aren't true about your competitors, but that is what people are doing. And and again, they they basically they put that out there, the the data gets fed in and then repeated by the LLMs and now it's like the LLM telling you that like something's wrong and there may be lawsuits over that like companies might actually sue like Chad GPT Open AAI because it's repeating false information that actually came from like their competitors. So, black hat is is a thing right now with this. Um, it's not that everything goes into it, though. Like, there's already a lot of data cleaning. You know, there's a there's a saying in data science, garbage in, garbage out. Uh, they're going to get better at cleaning up the croft, cleaning up spam. Uh, they're already doing some of this, I will say. Give them credit, but they need to do a lot more. uh because yeah, unfortunately the nature of SEO is abused until it doesn't work anymore. I will warn you though, this is a very very strong warning. We got very lucky with Google Penguin. Now, it hurt for a few years from like what 2012 to 2016. It was penalty penalty penalty. Businesses went under, SEO agencies went under. a lot of the freelance work dried up or shifted to to cleanup work, fixing penalties. Um, by 2016, Google was ignoring the bad links. They're just like, "Yeah, whatever. We forgive you." I don't know that AI assistants will do the same thing with spam. So, if you spam now, these assistants may hold a grudge for a decade, two decades. they may penalize you for the next 20 30 years. We don't know. Um and unfortunately like that means a lot of people will probably get hit, will probably um suffer long-term consequences. So a short-term spam strategy may work. Now I would still warn people it is very very very risky. Uh there are some better things, better practices potentially, but they can still be abused. Uh right now a lot of SEOs create one article for things. Well, even Google changed this what two years ago, year and a half, I think a year and a half or so ago where you can actually rank like twice in the SER. I mean, you could always rank twice, but it would be like first and second or seventh and eighth. Now you can rank like first and 31st. It's like you can have two separate articles ranking. Well, now if you've got multiple articles on the same topic, you potentially are influencing probabilities, more chances of being retrieved. Um, I think we're going to see more of this. You might see more content on other websites, too. So, a lot of acquisitions, uh, a lot of leveraging additional websites. Micro sites basically are back. you know that that hasn't been a strategy in SEO for more than a decade but micro sites were a big tactic for a while I think for these it's probably back uh again all any of these can be abused so in moderation please because if you go and set up a 100 micro sites and you now have 97 of the top 100 ranking results when the rag model goes to retrieve stuff uh more than likely it's going to get some of your website say what you wanted it to say etc. But uh that's that's definitely abusive. I would say if you've got one or two others, you're probably fine. Uh but I won't make any guarantees of that. Even influencer marketing, you know, that's can be let me outreach to friends or pay people to change what's already said. Those recommended product lists, don't trust them anyway. Like they're all paid. Uh they're all someone's affiliate. They're all they're all paid in some way in Uh same same with affiliates like leverage your entire affiliate network. Create content if you've got an affiliate program. You can you can make these things say what you want to say by putting that what you want to say on a lot of pages. Like that's how the probabilities work. Basically, uh it's only going to work to a certain extent though. Like if I'm uh if I'm in the US and I'm AT&T and people start saying my network coverage suddenly sucks. It doesn't. It's a great network. Uh but let's say they start saying that I don't think any hundreds of millions of dollars of campaigns would even influence the results there if that's what everyone is saying everywhere. But for a smaller site, yeah, probably this stuff in moderation will will work to influence the results. Uh yeah, paid outreach, paid content, content syndication, digital PR. Again, all these I think can be abused at scale. Uh so we'll see what lasts. But in in some kind of normal business function, normal moderation, this is I think right now the way to influence the results with with caveats. It's all more than SEO though. Like I know a lot of folks are like, "It's still just SEO." No, no, it's it's not. you're getting into other business functions that SEO is probably not going to be able to take over. Now there a lot of times you'll see you know performance marketing teams with like paid and organic maybe even analytics or something rolled in but I've never seen a single team set up where like you're working with the brand team also or like you're working maybe PR but generally even that's kind of separate u social media all this stuff your affiliates your influencer marketing in general like these are different business functions SEO is not going to be able to take all over all these you're going to have to work with these people in general to make some of this stuff happen. Uh you know the other websites that might be a mergers and acquisitions team within a company. It's it's more the same as what we've always done in that like we don't own all this and you're not going to own all this but you can work with all these teams. You can influence things. You can give them best practices. you can make suggestions on like what sites to buy and uh what kind of content to build on those and you know who you should be which pages you should be outreaching to or trying to like pay to change the narrative that kind of thing. Um but it's not just SEO just SEO is not going to be enough anymore. There's even more teams product brand product teams. Uh if someone's complaining about again AT&T's network coverage that's probably on their engineers and stuff to actually fix it to their business teams. Where do we need to put towers? Is what's the problem? Is it our signal isn't strong enough? It's not our network's not fast enough. Um these are things that are way beyond what SEOs do. We might have to try and influence it. like I've said repeatedly, but it's beyond just SEO, product marketing, comparison content. Like, we're seeing a ton of stuff for HF. So, in comparison content, it's working well for us. Uh, I would say some other companies are actually doing this much better though, and they're doing it at scale. They're doing it I've seen them leveraging affiliates. I've seen them leveraging other websites they own. And in many cases they are controlling the narrative in their markets by doing this. That is if you take one thing away that is that is what you should take away. Control or influence the narrative. You may not always be able to control it but you can influence the output of all these based on all those probabilities. So your top AI search tactics are not even SEO in my mind. Build a better product, be a better business, fix your stuff. These are what are going to matter. If someone's if if everyone online is complaining about something, you can't create SEO articles to get out of that. Yeah, you can do some reputation management, blah blah blah. Still a problem. People are still going to be saying it on social, on blah blah, on other platforms, potentially video, everything. You're not going to SEO your way out of that. So fix whatever the problem is. That and again now we're now now we're back to like business functions. People are like your service is terrible. Well, what's the problem? Is it the customer interactions? Is it my support? Like fix that. That is going to be your best way to influence the results here. You also might want to start diversifying your traffic. You know a lot of people have already moved social media. YouTube, I think, is my number one uh recommendation for people. Well, not just YouTube, but like video content in general. Go make uh Tik Tok videos and YouTube shorts and and everything like those are huge traffic driving channels. Also, uh as we see the clicks go down in search, if you still want to see traffic to your website, that's the way to do it until the AI starts taking all that over, too. You can actually see how you're mentioned in HF's brand radar. Uh I love this tool. The new UI, this that's what this is a picture of is launching I think next week. Should be out. Uh but you can actually see how you're mentioned. You can see how your competitors are mentioned, where they're mentioned that you're not. Like these are going to be important things to figure out like what do I need to fix? What do I need my content to say? How can I like change or influence that narrative? Cuz yeah, whatever is there, whatever is is showing in the results now, you need to change that. It's not just ranking your page. It's now a lot of off-site stuff, a lot of potentially other articles, other websites, social media, Reddit. I need to be able to change how I'm perceived. And the easiest way to do that again is to actually like fix your stuff. Whatever the problem is, fix that. Not create content that says you're better when you're not or whatever. But that's what SEOs are going to do, I'm sure. By the way, I just want to quickly interrupt Patrick to say that what you guys saw earlier was a sneak peek of what brand radar will look like soon. Yep. Like if you're looking at your brand radar right now on your uh in the HS platform, you're saying, "Oh, Patrick, it doesn't match that." So you that's a sneak peek. that brand radar might look a little different very soon. Uh on a page itself, like if you're focusing on the content of your site, what I would say is write like you would for feature snippets. Keep it simple, specific parts of the text. Again, think about those chunks. That's what's being retrieved. That's what you're being measured against with like cosign similarity. Like that's what AI content helper is doing. It's basically saying for this topic or this chunk like what are the relevant sections and then the the topic score we have is is basically cosign similarity in the back and that's saying like this is how related this is this search is to my content my that specific part of the content though. So, cover your topics well. Cover all the topics. Cover the things that you should talk about within them. Work on like if you if you've got this, work on getting these scores up because you're likely going to show up for more queries in more of the AI searches and you're going to be cited more. Uh, I still recommend like citable content statistics. You've probably seen us even in this uh I I used a couple of them but like the 34 and a half% less clicks from AI overviews the search traffic converting 23 times. These quote these two statistics are very heavy traffic drivers to those pages from AI search already because people are like how do AIO's impact my clicks and it's like oh yeah this from hrefs. Uh so this is working very well for us. I highly recommend it if you can do it. Uh redirect to hallucinated URLs. I talked about this briefly earlier, but the 3.6% of the traffic from AI search uh uh to hrefs were to hallucinated URLs. It's not that they used to exist and were redirected or like not redirected or anything. These did not exist period ever on our website. Chad GPT mostly just made them up and you can still reclaim that. I was talking with uh with Tim our CMO and I was like we need to put in some some business logic in this like I don't want to redirect everything but like if it's you know more than 10 visitors a month for like three months or whatever we should probably go redirect these pages. So do that. Put that in. Go go go every once in a while and be like just figure out where that was supposed to go and go redirect those pages. uh creative approach I saw online um I'm going to butcher his name Metahan uh he was doing social share buttons and he asked AI to remember or sorry not social share but like the AI share so it takes it to like the AI he asked it to remember a certain page now it goes into your personalization your history it's probably more likely to uh site your website in the future and I think that's a really cool creative uh hack for this. We ran a content survey and a lot of folks are using AI to create content now. 87%. That's almost everyone now. There's still some old outs. Uh companies that are using it publish 42% more content. Now, we we've even seen this with our own blog team or even a lot of the data studies I've published. I mean I'm using it uh even for like processing the data for uh a lot of mockups of images for some of the short text like but it's a very iterative process. It's not like I'm like AI go write me this that doesn't work. This doesn't work well. Uh but I' I've been able to put out a lot more content, a lot more studies because I am using this like AI assisted approach that's iterative. of the ones that are doing it, most of them are actually editing and reviewing the content, which is good. Um, if you just put out out a bunch and you don't review it, more than likely your traffic's going to go to zero. Uh, I've seen that time and time again. Thousands, literally thousands of examples we've looked at and we're monitoring. Uh, it doesn't work. If it's AI plus human, it's usually fine. Some edits, some extra sentences, some details. uh add your own unique expertise, your insights like uh AI just isn't good for everything. But for base level content, writing some some things like I used it to even try and capture some featured snippets or be mentioned in AI overviews. So I was like, well, if I want a good answer that's short and uh not too much fluff or too wordy, then that's what the summarization is like a great use case for these AI assistants. And it worked pretty well. I think in my first test I got like 13 out of 39 uh that I was trying for. And we've seen also that like lots of people are using AI in their content. Again, I don't recommend like pure AI or even even if you're at like 90% AI, like that's probably too much. But if you've got some mix, and for me that's usually like 20 to 75% I would say. I I think writing as just me now would be a waste of time. I there are still some holdouts that are like it needs to be 100% human. No. Uh you can still use AI in different ways to come up with better content. Uh but I wouldn't have it all AI or all human anymore. And that's kind of what the web is showing us now. Like it's not one or the other. Um the AI overviews in fact are citing AI generated content more than human writing. We ran this study because like after my study I was like well that seemed to work like can I just use the AI assistant to write something better write a better definition you know or uh create a better list than I could create? And the answer was yeah seems seems to work and that is seems to be what is happening overall too is that these systems are do tend to site the AI generated content more than the human generated content or human written content. There's only a minor minor preference for human written content. I think this was like less than one position on average. Uh, now that was like versus AI plus like human, not completely AI because completely AI, I'm telling you, like it'll rank for a while and then it'll just tank. It will not continue to rank. But overall, I think this is the path forward for all of us. Human content is not going to be the best use of your time. AI content is not going to rank. some combination means you're going to make better content in less time. And that is what SEOs uh writers and everyone I think are going to do. Again, my process isn't just like uh I mean I'll use it brainstorming iterative. What if I add this? What if I do this? Blah blah blah. Like it's it's not a oneandone generation and edits. Um it's it's I I don't know. I think it works and I think this is what we're going to see because again you you will get better content in less time and that's that's important. Uh you're going to have to create more content than ever. It is just the way that it is like now it's not just your site, your one page that you should be focused on. It is a bunch of stuff and not not even just like blog content or anything, but could be social media or you probably need stuff set up to monitor questions and Reddit. There's a lot of different things that we need to be doing and we don't have extra time. So, this is one way to get some of that time back is to actually like use AI for to assist not fully for your content. Thank you. That is all, but I want to give one more shout out to Hrefs Evolve. This is going to be an amazing conference. I'll be out there. Um, it's it's going to be a fun time. The afterparty, the sessions last year was amazing. Seriously, best conference I ever went to. There was this giant like 220 foot screen or something. It was ridiculous. and just every detail was handled by our events team in such an amazing way. Thank you all for coming. All right, so we are now going to move into the Q&A part of this um today's webinar. So we are past an hour now. If you have to head out um we're happy you guys stayed until the end. And for those that asked certain questions, now is the time to get them answered by Patrick. I'm going to start go ahead and start off with a question slash comment Patrick Herbert uh made earlier. I'm going to ahead and pin it there on the chat. He said in summary, spamming works, but spamming is bad. Please don't spam, but it does work for now. Uh and eventually it'll get hammered in a short-term question mark. Yeah, I don't think it is a sustainable path, but unfortunately what we are seeing that influences the results uh in a quick and fast way is a lot of spam, right? Okay, awesome. And the next question that was asked um Garrett Susman asked a question about the AI share button hack. He's saying, wouldn't this be like penalized in the long term? Like, should we even bother? I don't know if it would be penalized, but it would likely just stop recording in your personalization history. But then again, like you're the one kind of asking them to do it. So, I don't know. Uh, everything is up in the air. Everything is new. We We will see. I guess there's there's potentially it could be, though. Yes. Good call out, Garrett. Okay. And there was another question about um someone was not able to see a number. Okay. Um I think Hans, if you still aren't wanting to find out what that number is, let us know. Uh next question is, what kind of patterns are you seeing in the PPC world? Haven't looked. Uh, I don't think any of these systems even have paid ads. Like they just put them sort of around the AIO reviews finally. They were saying for like a year they had them and we're like, "We don't see them. No one is seeing them. Where are they?" And then they're like, "Oh, yeah, like we finally added them." Like, "What?" So, it was funny because even in some of their uh financial reporting, they're like, "Oh, it doesn't influence the click." And I'm like, "No one has seen one. How did they know this?" And uh yeah, now they're finally there, but I still haven't seen any in AI mode and chat GPT. In fact, I think ChatGpt is going to try like an affiliate model with their agent where like if it checks out, it gets like a percent of the revenue or something. Uh I don't know when these systems will have ads. I hope sooner rather than later because that means we'll probably get some keyword data and stuff and search volume type things. I think overall we're just hoping for more data. Like a lot of these uh AI players are not very upfront with the data that they're willing to share. They have a lot of claims about what is working and not working. And we're just on the other side just trying to collect our own data to see whether or not um what they're claiming has any merit at all. So, we hope that these players would be able to share more information about that sooner. Okay. Uh, next question. Are there any changes you recommend for team structure? I don't know that again teams I've seen all kinds of things like I've seen SEO sit in product. I've seen SEO in performance. I've seen SEO in an analytics organization, which I thought was weird. Um, I don't think I don't I don't know. Every company's going to choose their own team structure. Uh, but it's going to have to be more close collaboration. Brand teams I think are going to be like SEO is almost merging with brand. It probably should have years ago uh because we're a big driver of brand traffic and everything. Um, but now with like impressions, the way the reporting changes, I just don't know. I some people will try and take over everything, but like you're not going to take over actual product teams. More like more than likely like they would bring SEO in, but then now you're separated from other things that might influence. So, it's all up in the air. We'll see. Uh, people will try different things and we'll see what works. Okay, great. And sorry, Robert, I missed your question earlier. What is the best way to discover hallucinated URLs? I'm not sure if you can quickly pull up our platform, Patrick, to show how you found those. I believe it's web analytics. Yeah, in HS web analytics, there's an AI search, but that's like all the traffic to all the pages. Um, right now there's not a way that you can see. I can see it on my end. Uh, but I can't show you that. But we basically are pulling like the page titles in now. So, if it's a 404 not found or whatever you use for that title, you'll actually be able to filter to that soon. Otherwise, pull the list, crawl the list, and see what returns a 404. Yeah. basically broken pages uh in in your website traffic that like you've never seen before. Yeah. Well, again, like if you just pull the AI search stuff because your other stuff isn't likely to be 404, but could be. Uh but pull the AI search pages, crawl that list, see what it returns to 404, which is very easy to do on a web analytics, HS web analytics. It's just a click away. Okay. Um, Patrick asks, "Do you see any significant behavior differences between LLM traffic right now? Uh, people from Perplexity versus Chat GPT rather than AI visitors as a whole." Uh, you know, to be honest, I didn't really look into it. Perplexity is such a small amount of the traffic. I I just pull it all cuz yeah, chat GPT, what was it? 83.6% 6% of all the AI search traffic is chat GPT. So I just haven't gone into the the smaller uh platforms to look for any differences. Yeah. Okay. Um next question is by Mark. Um you spoke about how brands will have a strong presence. Can you link the connection between brand and your discussion on the importance of social media's presence for brand? Well, unfortunately the way these systems work right now is they're basically a popularity contest. It's it's sort of the same uh in Google search always has been. But like, you know, if you've got a higher DR website, more well-known, more people visiting your pages, more people talking about you, like you're more likely to rank in organic. And that is the way that it seems to be working with LLMs as well. So, I'm trying to remember, I looked at numbers once and I think it was uh something like the top 50 sites had like 20ome% of all the the citations. And that's a very heav topheavy bias. Uh so really like the best way to influence these these systems is probably just be more known. Uh be a be a more popular product, a more popular service, one that people talk about that uh do interesting things that fix any issues that come up. Um, I think that is probably the better way of going about it than just trying to manipulate the results with with a lot of the SEO hacks and spam that people are going to do. Those are only going to work for again businesses of a certain size anyway. Once you get to a certain level, like you're you're just adding a little noise. Like, yeah, you might influence 1% of the output or something. Uh but how much difference will that really make with all the other probabilities? Okay. Uh next question. Would you still do Google search ads? Yeah, right now. Absolutely. I mean we do currently. Yeah. Plenty of other companies still do it. again, everyone's still searching on Google. It's still like the massive market share. uh and still sending what I think when we looked like 40 some percent of the traffic was still from uh Google or traditional organic search versus like the less than 1% from the AI search stuff and and with in regards to like paid ads and AI search and AI overviews, we really just have to wait and see because I think they're still playing with the numbers. Um, Mark Anthony Staria asked, "Hi, can you please explain in simple words AI bots are breaking things?" Yeah, it's uh basically they're crawling a lot of things that other systems have figured out they shouldn't crawl. So if you've got like an infinite uh path full of parameterized URLs and everything uh they may just go and crawl like 500 million pages whereas you know Google or HFS might crawl like 20 of them and be like well this is all a waste of time and we don't need to crawl this. So they're wasting resources crawling things that they really shouldn't. Uh and again they're they're eating people's servers resources. they're not necessarily backing down. Like I've seen reports of them taking down websites. There's also just so many of them now. Like it's additional uh server load. So now it's like eating like you can only have so many parallel requests and so many pages being loaded at once. Uh so it's it's multiple ways that they're breaking things and it's it's really that um I guess the main two are probably that it is overloading the servers. they're not backing. Like typically servers, if they start showing like a 500 error or different error codes, then bots that have been doing this a while, we know to like slow down or like let's just stop for now. Let's come back in a few hours. Let's see if like there's more resources available to us now. But the newer bots, they're always they don't do that. Uh it's it's a learning process that every crawler goes through, I think. Um, so again, crawling faster than they should, crawling things they shouldn't, uh, eating entire res like some ser some, uh, hosts are limited and they're like, you only have this many, you know, uh, gigabytes or terabytes of, um, of, um, how do I put that? Bandwidth that can be served each month. And like they're just eating a good chunk of that. And so I'm starting to see a lot of people block them because they are aggressive. Okay. But even even search engines went through that same process. Like Bing, what probably when was that? Maybe 2018 SEO started getting on Bing because their bot was like really bad. And now like they fixed it like with index now changes to the bots, a bunch of stuff like that. They're probably the mo one of the most efficient crawlers on the web. Okay. And uh next question is again from Patrick. In a world where AI search is 50% of traffic or something. So prediction time. What do you think is the change that will come from that? Uh big future prediction prediction on how marketing SEO content changes. You shared something earlier uh previously. Yeah. clicks will be less. Um, attribution may be even harder than it currently is. Uh, it's hard to say like what all the implications are. Different different page types may work better than others. Again, like I we're already seeing our programmatic stuff is like, well, maybe we don't want don't want to invest as much there in the future. Uh, because like it looks like there's there's going to be calculations. Everything now becomes harder. It's going to be harder to justify budgets. It's going to be harder to justify new content, page creations. Um, everything goes back to like a costbenefit analysis. Can I justify even putting resources towards this? And it it's sort of going to be the same with uh other types of sites like I'm again I I talked about a bit about like the relationships and the different types of sites being uh cited in different systems. We very well could end up back in like a 1990s walled garden where sites are cutting off search engines. Like if they're big enough, they might be like, "I don't need you anymore." Like, "I'm just going to take my private audience over here." Others may still give away their audience or their their material and try and grow. Um, but it could also be like a case of, well, if chat GPT licenses this and Google licenses this and this other one licenses this, but what if I'm a fourth competitor there? I'm dead. Uh, so there's going to be a lot of interesting changes there as far as like who's allowed to crawl what, who, uh, is sourcing what, what kind of deals can be made. Uh, Cloudflare has this new thing where they're trying to get them to like license the crawling for like all the different AI search bots. I don't know that that's going to be successful where that's the answer. But it's a it's a good start of the conversation because yeah, when when all these backroom licensing deals are being made like I don't know, Google probably doesn't even have the resources to go and make it with every type of website. Um, so there needs to be something where like yeah, they'll probably make a few uh private deals and then a few that they just got to grab a bunch of stuff in bulk. But it's it yeah, I I don't know. I think every everything is going to consolidate a bit. Um, some stuff is just going to go private, just be taken off the web in general. It's going to be a wild time to see how all that changes. Okay. And I think we're just going to do one last question here again from Patrick. Uh, I mean, we were just speaking about Cloudflare. So, are you or do you have anything to set up to monitor bot traffic specifically? I guess specifically crawling your website from AI bots. Yeah, we we we use Cloudflare. We can see it. Yeah. So, that's the easiest without having to deal with server logs or setting up an elk stack or anything like that. Yeah. So basically a solution like Cloudflare that helps to filter traffic, lets you see and manage the kind of visitors that you get. Yeah. Okay. Um I'm so happy that you guys stayed all the way to the end. For those that did, um you'll get an email um after this session is over to share some feedback. Please share your feedback for today's session. is the first time that we did uh a webinar together with a a guest. Um I mean but like Patrick I think a lot of you know about Patrick already from many other events or or from uh his uh talks in other places. So be sure to check out for the next one. The next webinar that will feature a separate guest will be in August. And uh thank you so much for your questions. I think you guys asked a lot of really interesting questions, really cool questions um that I think are in the back of the minds for a lot of marketers. Uh thank you so much everyone once again for joining today. Uh we'll end today's webinar for today and we'll see you in the next one. Take care. Bye everyone. Thanks for coming. See you. Okay, stream has ended. All good.