}
Jojo Furnival
Sitebulb's Patrick Hathaway is joined by AI Consultant, Britney Muller, and Keyword Insights Co-founder, Andy Chadwick, to discuss how SEOs can optimize for AI outputs.
Watch the recording!
Please subscribe to the Sitebulb YouTube channel so we can keep you updated with future webinars. You can also sign up here to be the first to find out about upcoming webinars.
You might also like:
Webinar transcript
Patrick Hathaway:
So, let me introduce our guests today. Britney Muller is an AI consultant, has been speaking, blogging and sharing her expertise on data science, AI, and machine learning for years. She has a bunch of courses and content available to help make all of this stuff more accessible. Britney, would you like to just let everyone know where they can find out more about all the awesome things you're putting out there?
Britney Muller:
Yeah. So, most recently I joined up with Maven, an education platform, where I'm teaching a generative AI course, making this more accessible, helping you conceptualise how to use the tech. So, you can find that course there. And then, trying to continually upload different resources and things on datasci101.com.
Patrick Hathaway:
Amazing. I'm just going to put the link to your AI course in the chat. I think it's that one. If I've put the wrong one in, you can correct me.
Britney Muller:
[inaudible 00:02:21]. I have some students watching, so this is fun.
Andy Chadwick:
I went to one of your talks, Britney, and I think it was like the prelude to the course, and I can highly recommend it. So, it is a good course.
Britney Muller:
Thank you.
Patrick Hathaway:
Amazing.
Britney Muller:
Thank you.
Patrick Hathaway:
So, our other guest today is Andy Chadwick, who is an SEO specialist and co-founder of the SEO agency Snippet Digital. He's also the co-founder of Keyword Insights, which is an AI-driven tool used by over a thousand marketers and SEOs worldwide to enhance their content strategies. So, Andy, could you also share with us just where folks can find you and learn more about your products and services as well?
Andy Chadwick:
Yeah. I mean, to be honest, I tend to just shitpost everything on Twitter. So, there, normally. Or, our website, keywordinsights.ai, I tend to post a lot of guides on there, so if you go to the little resources section, there's some interesting things there.
Patrick Hathaway:
Awesome. Cool. Okay. So, I'm going to go on to our questions. Again, there will be time at the end for everyone else's questions, but I'll start with mine and we'll work through them.
So, I'd like to just start by setting the context. So, we could have called this webinar optimising for AI Overviews, but we didn't want to restrict it to just Google. So, what other AI outputs are there out there that we need to know about?
Andy Chadwick:
You go ahead, Britney. Oh, no. Oh, yeah. Okay.
Britney Muller:
Okay. I was going to say, I like how you asked the question, Patrick, because people kind of whitewash AI outputs and we have to get really specific as to are we talking about ChatGPT? Are we talking about AI Overviews or Perplexity? They're all slightly different, right? And there's nuances to each.
So, in terms of showing up and just more the general large language models like your ChatGPT, your Bard, it is a bit of a popularity contest with the sprinkle of randomness. Right? So, having more information online about your company, your product, your service, around the proper entities and contacts that you're trying to associate yourself with will be hugely advantageous in the long run to show up there. And maybe I'll just start with that. Andy, what do you want to...
Andy Chadwick:
Yeah. I mean, basically the same, but I guess, anecdotally we get clients asking us, "How do we appear on SGE?" And clients asking us, "How do we appear in ChatGPT answers just a really basic level?" And yeah, I guess to Britney's point, well, I guess we'll come onto it in a bit, but it's ChatGPT is essentially pulling from Bing, or Bing's pulling from ChatGPT and vice versa, and that's how they're training.
So, a lot of the things that we optimise for in normal SEO actually helps optimise for them as well. It's authority, it's links, it's brand mentions. All of the same things help you appear in those different channels. But yeah, we get clients asking us how to appear in all these different places. So, I'm glad you worded it like that as well. It's not just SGE, it's all these other platforms as well.
Britney Muller:
Yeah.
Patrick Hathaway:
Awesome. Okay. So, we'll dig into later, I guess, a bit more specifics, but how do you start the process of optimising for AI outputs then?
Andy Chadwick:
Again, I guess you're presuming talking just blanket-wide across all of them?
Patrick Hathaway:
Yeah. Let's not focus on any one at the moment, but if you were trying to sort of go, "Let's make something that's going to help us in the main," right? How would you approach that?
Andy Chadwick:
So, in tests we've been running for a lot of branded people, a lot of people concerned about that in the people we work with. If I go on ChatGPT and ask it, "Who would you recommend that sells steel doors in the UK?" Generally speaking, the top 10 answers it spits out, there or thereabout, it's the same 10 links that you see on Google. And same if you asked the SGE and it comes up with a list. It's almost the same results that you'd see. I'm just telling one example, we've done that with loads. If I asked it, "Who would you recommend that sells roller shutters or radiators?" I'm choosing really dull products here. But the answers it spits out are, generally speaking, the same ones we see ranking on the top search engines, Bing and Google.
So, the takeaway I have from that ... And I've gone to prove it myself, I wanted to set out getting my own name, I wanted to type in ChatGPT, "Who is Andy Chadwick?" And it now comes up with it. It didn't do that six months ago. It's because I put effort into making my name into an entity, building up the authority of it, and it works. So, to answer your question, from my point of view, it's a lot of the same stuff you would be doing anyway. It's driving links, it's building authority, it's making sure your name is an entity across various places. And just making sure that search engines, and by search engines, I also mean the bots that train the data on can find that information in the first place to train it, to bring it into that platform.
Patrick Hathaway:
What would you add to that, Britney?
Britney Muller:
So, I would optimise differently for different AI outputs. You can quite easily play the numbers game with Bard, with ChatGPT, because they're generalised outputs, right? They're kind of the view from nowhere, the average of everything. Something like AI Overviews has a level of information retrieval, a level of quality associated with the outputs because they're marrying that information retrieval and synthesising the output with a language model. You don't necessarily always have that with these other large language models. So those, again, are more of a popularity contest.
If I want to optimise for AI Overviews, and reading through Google's patent, it becomes more and more evident that you need to be showing up in other relevant queries. So, they're not just looking at the query results for whatever you're searching for. They're actually inspecting other ranking URLs for people that search your query, that searched this before or after, or in your area.
There's levels to this in terms of context, entities. This is a big reason why you see so many people like Mike King talking about the advantages of schema markup, right? Because we can sort of continue to spoonfeed search engines, "Hey, this is what my company is. This is what we're about."
But again, to Andy's point, it is like the SEO 101 traditional efforts of showing up and showing up high for that query, for other related queries. And it's interesting because I've been thinking a lot about how we tend to prioritise the high-volume keywords, right? The high-volume topics. But now we might have to revisit that a little bit because we're seeing so many results in AI Overviews that don't show up in the search results. The reason being is because again, Google is evaluating tangential topics and other related keywords. So, that's kind of an interesting aspect of this is you want to make sure that you're covering all your bases.
Patrick Hathaway:
So, just on optimising for the different platforms, would you go out and do something different for Perplexity, for instance.
Britney Muller:
So, I don't know exactly how Perplexity is put together. Right? It's obviously a married function of a large language model in some form of information retrieval. I've been keeping an eye on just global reach market share from Google, ChatGPT, Perplexity. And yes, these tools are growing, but they don't hold a sliver of the overall market share that Google does, and I think it's evident in that Google's been around for all these years. Right? They have all the smartest people in the world. They have doctorates that are experts at information retrieval and have been going about this, even though we love to harp on Google results, which I also understand, but they really do have the smartest people in the world working on the solution of providing the best results.
You can't surface information in a regular large language model like ChatGPT and I'd even go as far as to say Perplexity, without having that quality assurance behind you. These aren't truth engines. Right? These were not trained on, "This is fact. This is fiction." It was just fed everything. And so, it regurgitates things about dogs playing in the NBA, right? From Air Bud, or different blogs and stories.
It has no ground truth. We need information retrieval for that.
And so, that's why I continuously will say, I have my money on Google in terms of they have the strongest of both sides to put something really magical together. They're still trying to figure out how to do that. No one has successfully figured out how exactly to put all of this together, and there's a huge field of AI researchers that would argue LLMs have no business in information retrieval. So, there's a lot of complexities here that I don't think most marketers might be fully aware of, but it's important to talk about. Right? It's important to establish that these large language models, they're not search engines. Again, they also have no weighted information in the model. It's all contextual as things relate to each other. They're word predicting machines. They're not truth engines. Yeah.
Patrick Hathaway:
So, right, then if a lot of what we're talking about doing here is not that dissimilar consistency in terms of output, brand building, schema markup, all of those sorts of things, and a lot of that is much, much easier to do when you're a bigger brand, right? You've already got people out there talking about you as well. Do the small guys stand a chance of featuring for AI results?
Britney Muller:
Sorry, Andy, do you want to go ahead?
Andy Chadwick:
I was going to say, my opinion, and it is my opinion, is yes, I don't think anything has really changed. It's just more likely that everyone ... Well, I don't know. I haven't actually seen any up-to-date ... There's been two studies I've seen on do the AI search results take traffic away from your site? One was from Kevin Indig and one was from someone else and they both said opposing things. So, I think the jury is still out on that. But let's assume everyone gets less traffic maybe, because it just takes it all and summarises it, maybe with the little links. Assuming that, I think everything's still an equal playing ground. For me, it's the same things that have always been true. So, you could ask that question about SEO in general. Did the small brands ever have a chance of ranking number one against Amazon for this product? It's the same thing now.
So, to answer your question, I guess, pedantically, they still stood the same chance as they always have. So, it's still difficult, but for me, it hasn't changed. If anything, it's more exciting because the strategies that ... To come to a point that Britney said earlier, there's all these contextual queries you can start getting around and these longer tail ... I hate the word longer tail, because it's not really about that anymore. But these other angles you can come at offering it genuine expertise. I mean, we look at Forbes and it's got all these people chipping in and they're not experts. They've just got all this authority. I do think if you have genuine expertise, it's going to find that and pull that in. So, if anything, I think it's more exciting for them. But no, I don't think anything's really changed and it's maybe the same as it always has been. I don't know what you think, Britney. That is just my opinion.
Britney Muller:
Yeah. So, I have two different thoughts on this. So, I have done some wider scale AI Overview traffic impact research that unfortunately is kind of under a lock and key with an NDA. But I will say one thing that I've widely observed across some of these is that ... How do I say this? The higher your brand authority and brand popularity in terms of your top traffic driving keywords, the more protected you'll likely be with these AI Overviews. Because we don't see AI Overviews showing up for as many branded queries for good reason, right? Google doesn't want to get in lawsuits with all these companies spitting out wrong information.
So, brand authority really works as sort of this protecting mechanism from AI Overviews in general. Now, that being said, and something that I think that Andy was touching on in what he just said, is for the small players, there are certainly ways to get included into all of this stuff. And what Andy said about the expertise and providing unique information, I think this touches on such an important topic overall for AI overviews, for small teams, for small companies. Something I've already seen in the chat, Angie's saying that we're seeing so many LinkedIn results in AI Overviews, this all plays into this huge problem with language models around they have no opinions, right? They have no real world experience. That's why we're seeing Google pay Reddit $60 million a year to supplement their language model outputs with real human experience, with real human opinions.
To Angie's point, we're seeing them bring more LinkedIn in there, that they feel is representative of real people. So, I would go back to, I think, high quality content writing and unique perspectives, experiences, stories is going to be monumental in terms of these AI Overviews. That's what Google wants, right? We're going to see more and more of the AI content, but it's shit, right? It's generalised average of everything. In the long run, it's not going to be helpful for either people, it's not interesting to read. And again, Google's looking for the anecdotal, the quotes, the experiences of stories. And so, I think if you're able to really hone in on that and lift up the experts that you have access to, you can start to see a competitive edge pretty quickly.
Andy Chadwick:
To add another angle, I think it depends on your niche.
Britney Muller:
Yes.
Andy Chadwick:
So, we've now got in Keyword Insights, you can upload a load keywords and we'll track which ones produce the AI Overviews, and at the moment we're getting about 25 to 30% of people's queries are uploading, generate that. And the niches that we're seeing aren't generating that are financial, health, things that, for good reason, they don't really want to touch. They don't want to be liable for that. So again, to add another angle to your question, depending on your niche, it's going to have a lot less effect than other ones. Just to add another, I guess, angle.
Patrick Hathaway:
Nice. Okay. That's cool. We've covered that question really well. You kind of mentioned this before, maybe there is no right answer to this, but I wanted to dig into what impact AI Overviews can have on impressions, clicks, and traffic, right? And I can just imagine agency clients ringing up in a panic, "Why have all the graphs gone down?" And so, not necessarily what's the right and wrong of it, but how do you go about explaining this to clients?
Andy Chadwick:
It's difficult. Like I said, for me, the jury's still out. At the moment, we've got loads of clients worried about it happening, but it hasn't happened. I guess, a lot of our clients are financial or legal, so it's not really there yet. Like I said, I've seen two case studies. One said that it's taking a load of traffic. One said it hasn't. I don't have enough data to say which way it's going to go.
So, to answer your question, from my point of view, I've got no firsthand experience of having to explain that it has happened and what it means for them. I do have experience of putting together a strategy in case it happens and sort of caveating that, "Okay, we're going to lose a lot of traffic here..." But depending again on your niche, it's different advice. If it's e-commerce, we still need to rank your products. If I'm honest, it's still link building to articles and pushing that link equity to the product and category pages and ranking them. The AI's not going to steal all that.
For other niches, and I think it's maybe a question you might ask later going how this is going, but it's getting around the AI. So, "Okay, we're losing a load of traffic to this AI Overview. Let's find other ways." People are digesting more video content now. They're digesting these things in different places. So, "Okay. Let's not try and fight this AI or be visible in it or whatever. We'll find another way to get those eyeballs and it's maybe a different outlet."
So, for me, I've got no firsthand experience in explaining that, but that's the sort of strategies we're putting in place now is, "Well, you're an e-commerce site, so it doesn't matter, we'll still get these products ranking." Or, "Okay. You do have content, we'll find other ways to get the eyeballs on that content."
Patrick Hathaway:
I mean, obviously, Andy's just completely copped out of the question. Britney, how would you approach it?
Britney Muller:
Yeah. So, I think it's important to bring clients along this journey of how Google's features in general have impacted traffic over time. Right? So, hopefully there's been conversations previously about featured snippets and just growing features in your space. It's not all that difficult to come up with these custom click-through rate curves with Google Search Console data. And so, that's something that I know myself and many others have been doing for years. That is so powerful in these conversations because you can prove the change, you can prove that it's not on you. Right? The rank hasn't changed, but, "Here's been the impact on click-through rate because of X, Y, Z."
And so, I've been able to do that with different clients for the keywords that show featured snippets over the years, and now I've started doing it with AI Overviews, and it's fascinating. It's fascinating how it is impacting their traffic and how we're able to predict the percent of loss if they're number two, number three, number four, and just what that looks like. And then, we use that for predictions moving forward.
I also think, to Andy's point, about within Keyword Insights and other awesome tools, I do love Keyword Insights, I think it's amazing-
Andy Chadwick:
Thanks.
Britney Muller:
... but you're able to quickly create a pivot table, right? With AI Overview keywords and ranking URLs and see, "Okay, these are our most at-risk pages. What does that look like?" And I think for content teams, especially moving into 2025 planning right now, it's so important to really do a bit of a pivot. Right? We're seeing lots of AI Overviews hitting what is content? How to. How do. All the question informational queries. And knowing those pages that are being taken over by the AI Overviews, I would say, let's focus on the ones that it hasn't touched. Let's double down. Let's focus on how to expand your topic's space for the other areas that are a bit more protected while not losing those informational queries, but maybe just not putting quite as much into it.
And so, those sort of narratives, stories, benchmarks, clear benchmarks defined in the data can really help have those healthy conversations. Otherwise, my heart breaks for SEOs that people kind of point the finger at, like, "What are you doing? Our traffic is down." It's like, well, actually there's a lot that's out of their control. Right? So, having these conversations is actually so, so important, and it's great that, as an industry, we have some resources available to cultivate these in a way that makes sense.
Patrick Hathaway:
Yeah. So, when you mentioned doing this in the past for featured snippets, if you had a situation where a website used to show in the featured snippet, but there's now been replaced by an AI Overview and doesn't refer to the site or mention the site at all, is there anything you could do about that or would you not try? What would you do?
Britney Muller:
My first go-to would be to include more, again, those human experiences, the unique stories, the quotes. Evaluate what's ranking. Right? I also feel like that's far too often forgotten in terms of competing in search is, are you even reviewing the top 10 results and what they cover? Are you reviewing what results are within the AI Overview? What are the common themes here? What are the common topics? How can you cover that information better? Right? How can you maybe make it more succinct? Add in more anecdotal stuff. To Andy's point, is there a tonne of videos? Right? What's the medium in which these results are providing the information? So, taking all of that information and that will help you be more competitive.
Patrick Hathaway:
So, just when we've been talking about the AI Overviews and speaking with clients about the sort of SERPs and the makeup of the SERPs, it's kind of like the AI Overview is a bad thing. Like, you want to avoid it, or maybe you want to focus in areas that haven't yet been touched by them. Is that the way that we should be thinking about them? Or, should we be thinking about them as potentially a good thing that we can optimise for?
Andy Chadwick:
This is what I'm doing at the moment anyway, for, I guess, people who ask me. We're prioritising the queries first that don't have them, just because it comes back ... It's ignorance on my part. I keep referring back to these two studies. I still don't know how much traffic they're going to take away or not take away. I just assume they are taking it away. But I just don't know because again, anecdotally, I did a search the other day, AI Overview came up, and I did actually click some of those little links that came up, because I thought, "Well, that's interesting. I'll read more into it." So, I actually don't know, but my gut feeling is it does take traffic away. My gut feeling, it's not based on anything. In fact, as I said, my own experience is the opposite, but my gut feeling is it'll take it away.
So, I'm prioritising the ones that don't create that.
It's actually why we spun up that feature in Keyword Insights so quickly. We were doing a content plan for a client. I was like, "Shit, it'd be really handy if I knew which of these queries to prioritise first." That is to say, though, I wouldn't ignore the AI answers. If you read enough of them, and I have read enough of them, there starts to become a pattern, like you see with the featured snippets, it was always like, "Oh, to appear in a featured snippet, answer the questions succinctly in two sentences." There was a pattern. You can start to see a pattern with the how-to questions. And there's invariably a way to get into them, and it's still worth doing, 100%, because as I've said anecdotally, I'm clicking those links. So, for me, your question isn't, "Don't bother with them." It's more of a prioritisation thing. I'm prioritising the ones that don't, but then I'm still going to move on to those ones and optimise for them as well.
Patrick Hathaway:
Right. Yeah, that makes sense.
Britney Muller:
I love that. Yeah, that would be my recommendation too, is to prioritise by impact. So, we don't necessarily know what keywords are driving certain conversions that are your end goal, but to work backwards from your end goal, right? What are the most high-value pages that are converting X? Whatever that is. And then, evaluate the top traffic keywords for those pages and prioritise those whether or not there's an AI Overview. Right? You can double down on different types of content, but I would certainly prioritise by impact and also just wait to see what's going to happen. Because as I said, Google's still really starting to figure this out. Obviously, it didn't work very well at first when all of those crazy results came out and they had to pull it back a bit, right? They had to say, "Okay, we got to do something here. We have to pull in more qualitative search results versus letting a large language model go absolutely wild."
So, we're going to continue to see this metamorphosis of AI Overviews and what that means. But yeah, to Andy's point, prioritise by impact, certainly.
Patrick Hathaway:
Nice. Okay. So, what about a situation that is obviously bad? So, your brand or your product or whatever is appearing in an AI output, but conveying the inaccurate message. So, like if it claimed your product was more expensive than a competitor when actually it was cheaper or something like that, how do you handle that sort of a situation?
Andy Chadwick:
Yeah. I haven't come across this yet, but these are the risks I'm thinking about. Basically, any of the risk optimising for AI is you could have inaccurate results, brand defamation, all these things.
The only things I'm thinking about at the moment are things like active monitoring, like setting up some sort of monitoring event for them to come through. So, I think there'll be a massive space for AI tracking tools that pull that in. Couple that with some sentiment analysis. That is something that I ... Well, maybe we should just build that. That's something that I would want to be tracking right now.
Social listening tools to start picking up on ... Britney will know more about this than me, because she's an expert in the information retrieval and things, but I'm worried if a couple of bad reviews get overweighting in that LLM, that can affect it. So, some social listening tools, you can jump on it before the training data gets pulled through. I'm pretty sure that's how it works. But Britney, you'll tell me if I'm wrong. So, some social listening tools there.
I'm already thinking about these strategies, but yeah, that is one of my concerns. I made some little, I guess, notes here. So, is there any legal recourse as well, like people we can reach out to, to change these? I just don't know. But for me, it's all about social listening, tracking the AI Overviews on purpose to see if that's happening, and just monitoring it really.
Britney Muller:
Yeah. That's super interesting. I love the idea about tracking some of the social stuff and the reviews before it gets re-fed into a model to then be pumped out by these AI outputs.
I think something I struggle with when this kind of question comes about is we're really centering the conversation over the technology when I think it should be centred over people. I think there needs to be wide-scale public education about this technology. People need to understand that these things, they hallucinate, they lie. Those words don't even make sense, because again, we're talking about a binary system. They're not manipulative. People are being fooled by misinformation and being fooled by the technology because they don't have a strong enough understanding of what it does. So, I always struggle with the headlines that are like, "ChatGPT gets more and more manipulative." It's like, no, actually more and more people are overestimating capabilities or unaware of limitations. That's where I think a huge focus should be.
And shame on some of these AI companies that don't have clear enough statements as to what exactly is happening. It looks like a search bar. It looks like it should be some sort of information retrieval, and we know so clearly that it is not. I remember early on, I got a DM on Twitter about these quotes that ChatGPT had generated that I had said, and when I told this individual like, "Oh, that's really funny, ha ha ha." Something like, "I've never said those things." He couldn't believe it. He could not believe it.
Patrick Hathaway:
Wow.
Britney Muller:
He goes, "It just made that up about you?" I go, "Of course. That's what these things do." Right? But until you're familiar with that, you'll continually run into that.
Patrick Hathaway:
And there's easily a future where they are much more prevalent than they are even today. And the large majority of people don't have that perspective. That's probably the most likely outcome, is that just people are happy enough to believe that it's just sort of magic and ta-da, there's the answer. Right? That's what it means.
Britney Muller:
Yeah. That's the scary part because people are deploying this in ways that it should never be used in the first place. We're also talking about it in terms of deterministic models and generative, which are two completely different things. Right? And so, again, I think that lack of just 101, which I think should fall on AI big tech shoulders to do a better job, is just unfortunately missing.
Patrick Hathaway:
Is there anything that could be done or focused around consistency of message? So, we've talked about structured data as well, just making sure that everywhere you reference things ... I'm just going back to my example before where they've got the price wrong or something. Could you or should you start going, "Well, what messages are we responsible for that we can control that we are putting out there? Have they actually taken that data from something we've done?" Right? "Is there messages that we're not putting out there that are consistent that things are getting mixed up with?" Or, is that completely the wrong way to think about it?
Britney Muller:
I think that's likely so far out of your control because there's all these other external sites that can publish whatever the hell they want. Right? Yeah, and I mean, we're not reverse engineering Google. And I think us, as SEOs and marketers, we so badly want to flip this thing inside out and say, "Okay. How does this work? How can we correct this? Or how do we support this?" And it simply does not work like that. Right?
There's certainly the information retrieval aspect where Google's feeding some of the top results, we too often forget these large language models, they operate on a distribution output. So, they find the, "If you put in this, the most commonly associated things around this, the most likely next word are these." But then they're stochastic, which means random. And the reason why they sound so human and interesting and get a little creative is because they have this sprinkle of randomness in determining which of those words to produce. And that leads to all sorts of issues, right? Because it's not deterministic, because it's not just information retrieval. There's an element of stochastic or randomness that throws things askew every once in a while, and there's really nothing that can be done about it.
Hopefully Google would deploy some sort of reporting system so you can flag things, like we've seen with featured snippets and different things in the past. That's very likely something that will come about because Google can't cover all their bases either. And so, they're trying to figure all this out as well. Yeah, it's tricky.
Patrick Hathaway:
Cool.
Andy Chadwick:
Yeah. And to add to Britney's point, like you said, you can't control if another 50 sites is span up and says the opposite of you to purposely influence it, you can't fix that. But also, there's just so many queries that you will have never seen before, that you can't possibly account for. Like I said, the AI models are tending to stick away from anything health-related now, but if I have such a weird, "Can I give my parrot paracetamol?" Like a weird question that I wouldn't even think to make that content, and then it hallucinates and comes up with an answer. I can't control that. It's just a random question that someone's asked, that I wouldn't have even thought to make content about. So, there's just too many variables to, I think, even try and, I guess, mitigate against that to begin with.
Patrick Hathaway:
Yeah. Let's not even go down the potentials of negative AI SEO. I don't even want to think about that right now.
So, we've talked a lot about the present, I suppose, but this webinar is about future-proofing search, and in order to do that, you need to make some bets about where you think search is going in the first place. So, my question is how do you see the future of search over the next few years?
Britney Muller:
Ooh, good one.
Andy Chadwick:
So, I mean, I see it becoming a lot more platform split. I recently wrote an article, it's not a term that I've coined by the way, but it's a term that I like called search experience optimization rather than search engine optimization. And it's this idea of optimising in the right place at the right time. I'll give an anecdote. We've just had our bathrooms done. God, don't get them done. It's been stressful. But we've just had our bathrooms done. And there's a number of keywords that if I look at how my partner searched for this, there was a number of keywords that she searched that ... I'm going to use an example. Topps Tiles is a big bathroom supplier in the UK. These are all keywords they'd want. So, she started off by searching bathroom ideas. Now, that search started on Instagram just to get some ideas.
As she was going through it, she saw some cool gadgets. You know how you get these cool toothbrush holders that get rid of the UV and disinfect them? So, she was like, "Oh, that's cool. Let's go to TikTok and search for other bathroom gadgets." So, bathroom gadgets, another keyword that Topps Tiles would be interested in is now on TikTok. She comes across the gadgets, "Oh, that's so amazing. All right, let's start buying." So, bathroom tiles, that was back on Google again, "Let's buy the bathroom tiles." How to tile a bathroom was over to YouTube, and then something else was somewhere else. So, there's five queries there that would've been in my keyword research had I done a keyword research, but it was spread between five different platforms.
And so, for me, it's about working out what ... I'm calling it like 3D intent, and maybe that's a really shitty word, but it's this ... So, before we've had intent, it's just informational, commercial, whatever. I'd like another layer, a 3D, another layer of intent, which is, "Okay, this is the intent, but also what sort of platform do our users want it on?" I'm just adding that other layer of intent there. And so, we are coming up with strategies at the moment and ways ... I can't talk too much about it, but we're coming up with ways at the moment where when we do our keyword research, we're not just pulling through the intent, we're also pulling through the third dimensional intent, that platform that's most likely to be optimised for that given keyword.
And so, to answer your question, for me, the future of search is search experience optimization and splitting your keyword based on the platform that it's most likely to be available or searched on, if that makes sense.
Patrick Hathaway:
Yeah, I love that.
Britney Muller:
I love that. I think that is such a great example in the story of how multifaceted just information seeking is in general, that we so often forget until we hear or see a story like that. And to play off that a little bit, I think there is totally something there in terms of being more strategic moving forward by understanding, "Okay, this is the topic, this is the subject. What are the other platforms, websites that people talking about these things frequent?" Right? And what are those topics? I love SparkToro for surfacing a lot of that stuff. But then I'm able to take that and go a bit deeper, usually using AI.
So, when I see different subreddits that are really popular for this particular demographic, I will pull top questions, topics. I will start to immerse myself in the actual subreddit to understand how they even communicate. Each one is its own weird community, and you really have to deeply understand each and every unique one to compete in it. But I think understanding those platforms, getting really savvy in how you're providing value, how you're marrying both the content, the strategy, and current trending topics, is so brilliant in terms of future-proofing your strategy, getting included in things like Reddit, LinkedIn, different forums that, again, I think will have more weight moving forward. Because again, they provide Google and search engines what large language models will never offer, and that's that anecdotal human experience.
So yeah, I love that take on it. And I also see the future of search, it continues to get more and more personalised, right? And so, we'll continuously see some of that get pulled in as well. And so, to future-proof it by thinking about those different facets and trying to set yourself up for success when those things get brought in more and more.
Patrick Hathaway:
Awesome. Okay. Right. We know what SEOs are like. We know that if there's a thing that can be gamed, they will try. And I mean, a lot of the advice that you guys have shared today is nothing really to do with this, but I guess what I want to dig into a little bit is if we go back 20 years, early SEO advice would've been, "Get more links." Right? And then, that went well for a while, and then there were certainly ways which it did not go so well. And with the benefit of hindsight, we maybe would've done things differently back then.
So, what I want to dig into is what sort of risks are there? What things do we need to watch out for if you start thinking about optimising for AI?
Andy Chadwick:
I guess, for me, it's things like diminishing returns and opportunity costs. We saw it with the bloody, what was it? The voice search thing that everyone ... Do you know what? I never fell into that. You know when everyone was saying optimise for voice search, I'm going to tout my own trumpet here. I never fell for that. I was like, "I'm never going to ask Google Home for something and then go and buy it. It just doesn't..." But we saw it with that and we saw loads of companies and SEOs and whatever, putting resources into trying to optimise for voice search, which for me is just diminishing returns. Spend that time in resource, basically improving your products or services or other marketing strategies. So, I don't know if that's really a sit-on-the-fence answer, but for me the biggest risk is that. Yeah, I guess that's a big one for me.
Patrick Hathaway:
Cool.
Andy Chadwick:
Content quality, just churning out crap as well. That's a big risk. So, you start commoditizing all the content on your site and then you've just got as generic as everyone else. Because I mentioned earlier, and I'd probably be at risk of falling into my own trap here, there are patterns when you see these AI answers. If everyone starts optimising to all those patterns, everyone's going to start having the same content as well. So, there's that risk as well.
So, for me it's, yeah, commoditizing your own content, generalising your own content, and just it's a waste of time and money when you could spend that marketing doing something else.
Patrick Hathaway:
Awesome.
Britney Muller:
Yeah. I could not agree more, Andy. The comparison to voice search is so funny. Oh my gosh. It's so true though. It's so true.
Andy Chadwick:
It is.
Britney Muller:
It's like the shiny object syndrome in SEO and we all get fired up about it.
Andy Chadwick:
But it's not always us that's led, it's clients. It's clients that come to us because they read something, "I want to rank for this." And then you're like, "Oh yeah, let's go along with this thing."
Britney Muller:
Yeah. "What are we doing about AI Overviews? This is going to absolutely kill traffic." But it's affecting everyone. It's not one side or another being targeted.
So yeah, I couldn't agree more with what Andy said. I think the big risk is the distraction, a waste of time and trying to optimise for things that are really difficult to understand how to show up for. Certainly, experiment, but don't put all your eggs in that basket. And then, Andy also touched on, don't just crank out AI-generated crap that, to Andy's point, it's formatted so similarly, time and time again, it uses all the same words. Right? Like, unlock, unleash, revolutionise. I have a growing list with my students of just common words we see all the time, and it's insane. And anytime I see an email come in with one of those words in the subject line, I'm like, instant delete. You didn't even put the time in to try. I don't know.
So yeah, I would double down on high-quality content. And again, I think writers at the beginning of all this felt the most threatened, and I think incredible writers are going to be some of the most sought-after professionals in the next couple of years.
Patrick Hathaway:
So valuable. Yeah.
Britney Muller:
So, I'm excited about that.
Patrick Hathaway:
Yeah. That unique input, right? And that unique experience is just going to be so important. Yeah. All right. I have a few more questions, but we are coming up to quarter to, so we'll go through some of the submitted questions now.
Again, just final reminder that if you've got any questions, please put them in, and last chance to go in and upload anything. So, I'll be putting some on screen. So, we'll start off with Chris's one here. So, Chris has asked, "Amazon is using AI for product reviews to summarise the reviews of an item. Do you expect e-commerce stores to adopt this approach? And do you think AI product overviews are the next big thing in e-commerce SEO?" Anyone's ready to go.
Andy Chadwick:
I mean, I love it. I use the Amazon one all the time. And do you know, if there was one on Reddit threads, I think Reddit should bring that in because I find them so useful. Do I think they should adopt this approach? I mean, at the end of the day, to sound like a broken record of everyone, if you are offering better customer experience, do it. And for me, if your site has a load of reviews, it does offer a better experience, so do it. I don't know if I'd call it the next big thing. I don't think it's a strategy per se, I just think it's a good UI, UX user thing is my personal take.
Patrick Hathaway:
I think it really helps that you still have the real reviews underneath as well. It gives you a flavour, right? But you can still go, "Well, what did actual people actually say?" And that bit's not skipped out.
Andy Chadwick:
Sorry. I was just going to say it is open to being heavily abused by the website. I mean, I imagine if you are the web owner, you could easily influence that, which would be my worry. But that's not really the question. Sorry, Britney.
Britney Muller:
No, I liked your answer a lot. I think this is a awesome use case for things like large language models. Right? The outputs improve dramatically when you're able to feed it accurate, high-quality information. And in this case, if it's just pulling in all the reviews, it's really helpful to users who don't want to read through everything.
I would maybe double down on a use case like this for e-commerce businesses in general. Are you doing this to review what's going on generally with your different products or services? And I see hotels doing this more and more. Right? It's a really quick and easy, brilliant way to surface like, "Oh, we actually have a wait problem and the people that are staff aren't being as friendly as they should." It's a quick way to surface quick wins, in my opinion, like, "Oh, there's not enough towels in the room." Silly stuff like that can be surfaced like that. And I think we should all be using this technology in these ways to improve what it is that we're doing.
Patrick Hathaway:
Yeah. What bad things have people been saying? That's cool. I like that. I mean, in my experience, the best use case I've seen of AI is summarising. So, I have, for my sales calls, I'm talking to people about Sitebulb Cloud, the note-taker is absolute gold. It just comes back with, "These are their pain points. This is what they're using right now. This is how many seats." And it's just, it's perfect. It's really, really good. And that for me is a really handy use case for AI.
Right. What else have we got? Okay, so this one is not so much about optimization per se, but I guess we've talked a little bit about the pitfalls, but this is, "What about AI-generated content itself? So, do either of you recommend or support this?"
Andy Chadwick:
I love this question. I want to write a sort of thing on this. So, AI used right in your content is so useful and powerful. Again, I'll give an anecdote. I wrote a guide on Search Engine Journal. You can look it up. It's called Tangential SEO: How to Find the Keywords No One Else Is Finding. The reason I've been so specific about the title is because it's obviously a technique that we do. It's my idea, it's my methodology. It's very much a personal thing that I've written. However, that article, it's about 3,000 words. It would've taken me eight hours to write without AI. Using AI, it probably took me four to five hours.
Now, the ideas were all mine, but where AI came in really handy was things like, so I like to include metaphors when I'm giving someone an example. So, "Doing keyword research like this is like..." And then, insert funny metaphor here. Or, "Here are three examples of where this might help." And then, I try and think of three examples, because it really helps the user visualise it a lot more. Those metaphors used to take me ages, especially because I'm under some pressure to make it a bit funny, to keep the user engaged. Those examples took me ages because I don't just want to think of one, I want to think of three, because it helps the writer.
Also, I want to sound more intelligent. I used to spend ages poring over my words and making sure it was worded correctly. So, I use AI to help me write. I use it to write in metaphors. I use it to write similes. I use it to rephrase things to make me sound clever. I use it for that. Or, maybe I use it to generate a quick description if the word is something that everyone knows. So, if I talk about what is a page title, I might use it to just generate a paragraph because that's easy, but the ideas are all mine. It's just substantiating it, rephrasing it, metaphoring it, simile-ing it, they aren't verbs, but you know what I mean, to quickly come up with those things. And yeah, that's my take on it.
And when you can see that come through, for better or worse, on Keyword Insights, if you ever used our AI writer, I'll speak openly about it, it's not quite landing with our audience yet, because our audience are expecting what they get with, I'm not going to name names, but some of these AI tools that just generate the whole 3,000-word article. Our AI writer doesn't do that. It's designed to do exactly what I've just said, and it's not quite landing probably because we're not doing a good enough job at advertising that that's why and how you should use it. But because the people signing up and who googled AI writers and we're ranking, they're expecting whole articles. We are not about that. And that's not how I use it. So, I do support it when used correctly, if that makes sense.
Patrick Hathaway:
Yeah, totally does. Yeah. Britney, you're bound to have some-
Britney Muller:
I love that. I love that so much. I love hearing how you work with it to create something like the Tangential SEO article, Andy, because I think to your point, that is where it really can be an assistant of sorts to help you ideate or grow an idea for the perfect metaphor, right? It's probably not generating new stuff that you say, "Oh, I'm taking that verbatim first output and putting..." You're working with it. It's this iterative process and it probably sparked additional ideas.
Andy Chadwick:
Exactly. Yeah.
Britney Muller:
So, I think that is brilliant, brilliant. I personally struggle with transitions in my writing. I just don't think like that. I'm this sporadic thinker. And so, it really helps me with transitions. But I also think aspects of this technology are valuable in crafting outlines. And so, instead of, again, going to Google search, reviewing all of the URLs on page one, I could go to something like wordcrafter.ai, which I have no financial affiliation with, I just love it because it retrieves all of that information. It feeds it into a large language model and then it synthesises, "Here's what an outline or a brief would look like to be competitive in these current results we see today." That is taking so much work off my shoulders and off team's shoulders that are trying really hard to compete and search for different keywords.
It's accelerating that content creation process, to Andy's point, whether that be in the writing stage or this kind of brief research stage. It's super, super helpful. And so, that's where I see it working really well. But to Andy's point, we need Andy's unique voice in the Tangential SEO article, otherwise it would fall flat. Right? And so, he's providing all of the ideas, the anecdotes, the stories, why this stuff works and how he's experienced it, which is why people want to read it. Right?
And so, not losing sight of that, I like to call it human-centred AI, making sure that we're constantly having someone in the loop and providing the content. I think the worst thing you can possibly do is get AI to generate a bunch of generalised articles and hit publish on all of them on your website overnight. I think over time, that will not be good for all the obvious reasons. So, yeah.
Patrick Hathaway:
Awesome. I love those answers. Okay, we have a couple of questions here about tools. I don't know if there's going to be clear answers on these, but Simon's asked, "Are AI Overviews destroying ranking reports in tools?" What are your thoughts on this?
Andy Chadwick:
Again, I haven't had so much experience with it personally because a lot of our clients are in legal or health spaces. So, for us, the AI reviews aren't doing much to our ranking report. They're not taking anything up that they didn't used to before. I imagine they would do though. So, I don't know if Britney has any input.
Britney Muller:
I feel like, Andy, you would have a much better answer for this just than me.
I thought I would be able to spin up different instances of like, "Oh, a logged in state to look for AI Overviews." And I don't know how you tools do it, because I would've probably spent days trying to figure it out. But yeah, it definitely adds a layer of complexity, for sure. But again, AI Overviews, they're not showing up on every single search result. They're currently just in the US. And so, there's different aspects of that, that we can sort of mitigate and continue to look at the tried and true tools that facilitate that info.
Andy Chadwick:
The difficulty for me, for that, would be we are already pulling in if a keyword triggers an AI Overview and we are pulling in the rank. So, it wouldn't be difficult to create, do you rank in the AI Overview? That's not the difficult part. The difficult part is let's say it's got a question, it's got 10 links in it and some websites are on five of the links and some are two. So, how do you rank track that? Are you rank tracking whether you're just in it or not? I don't think it's that binary.
Britney Muller:
Yeah. That's a good point.
Andy Chadwick:
Is it, I have 50% of it, or I have 40% of it, or I'm third? It's so complicated. We could come up with a, "Are you in it or not?" Tomorrow. Done. But I don't think that's fair, because some sites have so much of it. It's a bit messy.
Patrick Hathaway:
It's too muddy. Yeah.
Britney Muller:
That's a good point.
Patrick Hathaway:
All right. We have another question about tools. Again, I don't know if we're going to know the answers to these, but Brendan asks, "Is there a way to track AI citation referrals to your website?"
Andy Chadwick:
So, that's the part we can do quite easily. If the citation is from SGE anyway, because we can scrape that. So yes, there is a way to track that fairly easily.
Patrick Hathaway:
Okay.
Britney Muller:
I guess, I'm a little confused because I'm wondering, is this also maybe asking, is there a way to track AI outputs of your website or referrals? Which would fall on the platforms themselves, right? Or, would there be a way to track that? I'm not sure.
Patrick Hathaway:
Yeah, I mean, I'm reading it as like hrefs being able to tell you that you've been mentioned or been used as a source, right? Within a AI citation somewhere.
Britney Muller:
But we're not able to track all of the ChatGPT outputs and we don't-
Andy Chadwick:
Yeah, I just meant SGE, we can, but yeah, ChatGPT, I wouldn't know how to. SGE, we can scrape that. When it comes to all the other platforms, no.
Britney Muller:
Yes.
Patrick Hathaway:
When you say we can, Andy, do you mean you can in Keyword Insights or is that something that you're doing manually yourselves?
Andy Chadwick:
I meant we collectively, like there are API things you could use to scrape them now, but within Keyword Insights as it stands, you can upload a big chunk of keywords. We don't currently tell you if ... We show you if there's an AI output and we show you where you rank. It wouldn't be too difficult for us to say that you do rank in an AI Overview, because we're already pulling that. So, I meant we collectively, but with not much work, we could put that in.
Patrick Hathaway:
Okay. I guess that's a feature request now. All right, guys, well, we're at time. So, I think we'll leave it there with the questions, but some really, really good questions.
I think the summary of today is you need to go out and buy Keyword Insights and start using that. And you need to go and sign up for Britney's courses in AI, so you can learn more about yourselves.
Andy Chadwick:
And Sitebulb.
Does that sound about right?
Britney Muller:
And Sitebulb.
Patrick Hathaway:
Oh, and Sitebulb. Yeah.
Forgot to say that. Yeah. Well, we don't have any AI features yet, so I can't really pretend that we've got AI stuff in there. But yeah, please do also use Sitebulb.
So, guys, thanks, everybody, for watching and for all the fantastic questions. And huge thanks, of course, go to Britney and Andy for so generously giving up their time and expertise. There has been a couple of questions about it. We will be publishing the webinar over the next few days on our website and emailing that out to everyone who registered. So, if you missed the beginning, don't worry, you can catch up.
