Information gain is the anti-slop ingredient for AI visibility: net-new knowledge models have not already ingested. On episode 19 of The GEO Show, a Friday LinkedIn Live, Paris Childress and Simon cover GEOforge 0-100 information gain score, corroboration, reviews, YouTube AI Overviews citations, and influencer go-to-market.
Information gain is the anti-slop ingredient for AI visibility. Models chase net-new knowledge they have not already ingested.
On episode 19 of The GEO Show, Paris Childress and Simon cover Google Trends acronym fights, the Brighton SEO keynote on information gain, the GEOforge 0-100 score, e-commerce experience paths, Reddit black-hat aged accounts, YouTube AI Overviews citations, SiteForge earned-media pitching, and influencer go-to-market.
Net-new knowledge public models have not already trained on or discovered.
It is really the key ingredient, the key anti-slop ingredient to ensure that you are giving AI what AI is hungry for, which is new information.
No. Information gain guarantees novelty, not truth. Models still need corroboration.
Through lived experience, verified reviews, use-case stories, and product video.
Human traffic falls while bot visits rise. Structure for retrieval, then earn relevant mentions.
18:15. Information gain score 0-100 as forecast metric.
39:07. YouTube share of AI Overviews citations.
48:20. Influencer and UGC expansion of GEO.
Paris: Hi, everybody. Simon and I are back for another exciting LinkedIn Live. Simon, how are you doing on this lovely Friday?
Simon: Ready for the weekend. I think, it's been one of those weeks where I haven't stopped, so yeah, definitely ready for a little break. Yeah, same here. Yeah. But exciting things are happening, and things are, things are moving fast, so I'm not, I'm not complaining. It's been one of the busy- Well, I was- ... the busiest summers for me, by far. I was looking at some videos today of me talking about answer-engine optimization back in 2019.
Simon: So I'm building a case study up for all of those videos, the strategy I was talking about then. I d- did I t- did I tell you that I wrote a book as well? So- I think you did mention that.
Paris: Yeah, so the- And I do wanna remind the audience that's listening that your LinkedIn handle is linkedin.com/aeo. Is that right? Yeah, f- yeah, /in/aeo. Oh, okay. Yeah. They don't allow stuff in the direct- But your handle- Yeah ... your LinkedIn vanity handle matches that older label. So you have no choice but to, to be on the AE- in the answer-engine optimization camp, opposing me, who's in the GEO camp.
Simon: Yeah, but I'm, I'm not, I'm not opposing you, because I personally believe they are two disciplines. albeit- Okay ... you know, working on the same thing. I think my, the answer-engine optimization side is more the marketing, and yours is more the structural and core tools, you know? So I think both of those things work together. Yeah. I think that the disciplines work together, but will the industry be able to sustain two acronyms for two disciplines that are so closely related, that are Well, I'll, I'll te- ... gonna ...
Simon: I'll give you the answer to that, because the SEO industry are making up about 30 different acronyms every day to go along with it. So- What's the latest one? AI? Well, they're now, they are now trying to go with AI SEO. - ... or there's AI optimization, or I think- ... 'cause I even put into a chat yesterday VEO. Why don't we have voice engine optimization? 'Cause that's, that's where we're gonna go next.
Simon: Yeah, there's a lot of, there's a lot of acronyms flying. Did, did, did SEO ever go through this? I don't think it really did. It was SEO and SEM for a while, but then SEO won out and then SEM faded away. Yeah, but- But SEM Rush as a product still somehow, that didn't hurt them apparently. No I think we'll see. I think SEM certainly- We, you know, I don't, I don't think there's a huge, there's not a huge push into, you know, they're not trying to get rid of answer-engine optimization and GEO, and the SEO guys are putting it in as a, a skill, which, you know, that's fine.
Simon: I think you and I both know that the whole search p- you know, world of search is gonna change to AI. How it does that will be- Yeah ... really interesting to see. you know, I, I don't, I don't like the word, you know, LLM. are the general public gonna talk about LLMs? I don't think so. No, no. You know, I, I think we- Well, it's already much broader than, than language.
Simon: I mean, there's so much- Yeah ... multimodal output now that they're, they're much more than LLMs. And, you may have seen that ChatGPT, ChatGPT-6 Astra just dropped today, and, the president of OpenAI, Greg Brockman, has claimed AGI has arrived, and that history is gonna look back and say, "ChatGPT-6 Astra is AGI, AGI's first arrival." And, I don't know, maybe it's a big milestone, but, Fable 5.1, which was, which is awesome, I've tested it already, awesome model, got blown away on the benchmarks, I mean blown away by Astra.
Simon: So, I think, I mean, just the biggest, the biggest news that's, just hovering over everything in AI right now is, is Astra, I think. I mean, has ChatGPT now not only reclaimed the lead, but m- but seriously leapfrogged Anthropic, who it has trailed for at least a year, a year and a half, I think. I know. I've seen, I've seen a lot of my, what I would call, well, say clients and people that I work alongside with using...
Simon: So I've always been, well, I started out with ChatGPT, I'm sure we all did. and then- ... gone into using Claude, Claude Code, and sort of got ingrained in that. but a lot of the guys that I've been speaking to have been using Chat and then working in the models there. I think it's gonna be, it's a whole playground that everyone can learn every single day.
Simon: Yeah. I think there's now, the pendulum, the pendulum is starting to swing back to OpenAI, I feel like, in terms of the momentum and just the buzz I'm, I'm hearing, and Astra is certainly gonna keep, keep that pendulum swinging. But it has been at least, I think when, Anthropic basically proved that coding agents was the, was the way to go, that was the best use of this powerful AI, and from December, I mean, probably December of last year all the way through, I don't know, June, July this year, so pro- probably six to eight months, I think it was clear that this is when most people went full, full on with, with, Claude, including our company.
Simon: We got Claude licenses for everybody in the agency. But, now, now I think the pendulum is swinging. And you also have Grokbot, which is so interesting.
Paris: I've been using Grokbot really heavily and just little by little giving it more routines, giving it more work, 'cause, what I wanna do is test the limits of the, of the usage. I've only been scratching the surface, but that chat experience is, it's really powerful. And it's not task-based, like with, ChatGPT and, and Claude. It's role-based, so you don't really ... I, I just communicate with my chief of staff, and it figures out how to convert whatever I'm asking for into the right task for the right agent, and then it, and then it, delegates it down the, down the chain.
Paris: but anyway, I'm digressing. We were talking about the acronym, acronym soup in this industry we're in. but who knows? I think it's still gonna be answer-engine optimization, GEO. I'm always checking Google Trends, by the way. And do you wanna see? I mean, I might as well show you here. Yeah, that'd be really interesting, yeah. - Give me a second to pull it up And, By the way, Google Trends ha- had a, had a major facelift, which is really nice Let me get these things up so I can have the big wow effect when I share my screen.
Paris: So what do, which one do you think is more popular worldwide? Which one, which one would you bet on before I share my screen? If, if you had to bet on it, I'd probably go with a lot of...
Simon: There's so much talk around AI SEO, which is what all the SEO guys are gonna- ... like, they're not, they're not gonna go away from SEO. the book I wrote in 2021, there's some really key quotes that I put in there around answer-engine optimization, right? Because I was running, I was doing a lot of work with Google, answers, and telling people, "You've got to be building the question and answer," and talking about schema and everything that you should be talking about.
Simon: but I was talking more so about voice at the time, you know, voice and video- Yeah ... mainly because Google was saying, "We're going video. We're going full video, right? It's all gonna be in video." So I was talk- talking to people about building the question and answers in video to come up in search. And as it worked, g- you know, that still works. It works incredibly well, but just people don't wanna make the video, or they haven't got the time, or you're asking people to do something that they don't wanna do.
Simon: fundamentally, we are still typing into these search engines or agents. it's got to move to voice. Like, the pocket device that I've got now, which is here, right? So this is, this is listening to our call now. - I better, I better watch what I say now. No, of course, yeah. Normally I don't have a problem with me being live, but when I'm being recorded, then I'm getting nervous.
Simon: No, I'm just kidding. That thing looks like a credit card. Well, I won't comment. You put it in your wallet or something? Yeah. Yeah, so it just sticks on, on the back of your phone like... There you go. Oh, okay. Yeah. So then once it's finished- Love it ... it'll transcribe it. But the interesting thing is it doesn't just transcribe. It does a great summary. It cuts it down into meaningful pieces, then it'll give you the action points.
Simon: I've got it connected to Claude, so it'll book my meetings. It'll publish blogs for me, etc., etc. It can do everything, so- Can, can you talk to it? Can you give it tasks? Can you just say after a meeting, "Hey, b- book the next meeting with Paris- Yeah ... tomorrow at 5:00?" Yeah. Wow. And what, what LLM does it do that through? Or, or does it do it on its own?
Simon: Well, you can... There's multiple connectors, so Claude being one. Yeah. Okay. But to get back to the point of, you know, where do you think it ends up? Surely it ends up in voice, right? It's got, got to end up- ... with us all just asking questions. Yeah, totally. Yeah. I think, I think voice, voice will dominate once the screens disappear, I think, and the screens will disappear.
Simon: So I think these two things have to coincide. Your thing that you talk to, the speakers we've been talking about, the clips, the, all the accessories, they're all screenless, and that means that the only way... Nobody's gonna type into- You wouldn't type something into that credit card device you've got. I mean, you're definitely gonna speak. No. But- Yeah. So- Look, I'm, I've been holding back, Simon.
Simon: I've got... I, I can't wait anymore. I have an itchy trigger finger. So the answer is, who's winning? Let me, add this to the stage. Boom.
Paris: Generative engine optimization still winning for the last, how many, how many months? You wanna look at a little bit lo- a little longer? That's the past year. Let's look at the past five years. Boom. What about the- Well, that shows us we're in the right, we're in the right industry, but that's- What about the, the acronym themselves, though? As in the- Oh, you can't, because GEO spoils it.
Paris: You just can't. It's GEO. It's, you know, too, too many GEOS. Oh, I see. I see. Too m- too many, too generic, the G- It's too generic. You can't, you can't compare GEO and answer-engine optimization 'cause GEO will, will be off the charts. But I tested something else. Are you ready for this? Let's try AI search. What do you think? Is it gonna even show up? Boom.
Paris: It overwhelms the other two. So here we are debating answer-engine optimization versus GEO, where, answer-engine optimization search is completely smothering them. Although it's on a downward trend- Yeah ... as of the last, month How's, how's AI slop performing this year? Well, well, the, we can't compare it in this set because that's not fair. No, no I mean, we're, we're... But it ne- it needs a competitor, in its own, at its own level.
Paris: I mean, a contextual competitor. Well, what, what about SEO then? Fair enough. See if, see if that's falling off at all. No, look at that. SEO dominates AI search, which dominates A- G- GEO, which, which has the lead on answer-engine optimization. But, this, this really puts things into perspective I think also you, the SEO has, ha- has an intent that evolves over time.
Simon: I think a lot of the intent behind someone searching for SEO and SEO terms, they may als- the, the, the intent may be now for AI SEO, but they just don't know how to, how to phrase it. Let's put AI SEO- Yeah ... in here, because you said some people are pushing for AI SEO. Yeah. No, that's interesting. Here it is with the orange line you see in the middle.
Simon: So l- let's take out SEO now, 'cause that's Okay. So now AI SEO is significantly ahead of both GEO and answer-engine optimization, but still way underneath AI search But is, is trending downwards? I don't know. You can't, I can't really tell. well, if you wanna see the trend line, let's remove AI search and let's get these, just these three players. Now AI SEO is trending down.
Simon: Yeah. But the, the, the interesting thing that we can't do here would be to say if... And the interest for me is what, what are the interest in these terms in terms of marketing leaders, you know, marketing CMOs, marketing directors, people in marketing. What's their level of interest in these tools rather than everyone? 'Cause I would bet that most, you know, most marketing people are still SEO, right?
Simon: And are not really talking GEO or they might, you know, they might phrase it LLM visibility or I'll tell you the one- ... the one that we should search is share of voice. I don't know whether that, you know, that's one that's crossing over into other- Well, yeah, that's, I mean, I think, that, th- that's, that not even gonna appear here, just relative. Oh, no, look.
Simon: There it is. Okay. Share of voice. So I think if we wanna see s- then we have to get rid of AI SEO 'cause I, I wanna look at tightly bound, competitors. Now this is interesting. So basically th- the, there's a downward trend for everything. maybe answer AI, I'm sorry, answer-engine optimization, answer engine optimization is, is still holding it a little bit, but GEO is in a downward trend and share of voice is in a downward trend.
Simon: So does that, is that telling us that the overall popularity is, it's going down for the space? 'Cause I wouldn't, I don't think I would believe that. No, it's not. That's gotta be impossible for the, you know, unless, unless the SEO bros are winning and pushing, you know- ... pushing the, you know, pushing the interest in the prop- what we would call the proper acronyms downwards.
Simon: But- Yeah. Yeah ... you know, and I know we can't measure, you know... Well, I know you can't do GEO 'cause it'll go off the charts. But what about the just answer-engine optimization? Because does that, you know, that, does that tell us anything? 'Cause I don't think anyone's- Well, I, I, I no longer think this is apples and apples, but let's, let's do it and explain.
Simon: No, no, no. I, and I agree. Yeah. So, okay. There you go. - But if there was a way to, to, to put GEO without geo, if you know what I mean, then- Yeah. By the way- I think we could- By the way, answer-engine optimization also stands for American Eagle Outfitters Authorized Economic Operator. so we can't say answer-engine optimization. I know, I know it's, it's probably mostly answer engine optimization, but, can't...
Simon: We can't do it. I mean, and by the way, we also have to make sure we're not comparing search term to topic. We have to compare either topic to topic or search term to search term. So let me get rid of, I'm gonna get rid of share of voice for now. And now if I, if I compare topic to topic, we see the same thing, but even a bigger gap.
Simon: And topic means that it's a broader, I guess it's the equivalent of a, a broad match keyword versus a, a phrase match. - Yeah ... but I think topic is probably a better indicator rather than the actual search terms. So what, what about if we look at not what we would search, what... So if I was a marketing manager, I might search, increase or rank in, ChatGPT.
Simon: ChatGPT. Yeah Yeah, I mean, I think we're, we're, we're going down the longer tail, but, - I think that's more of a meaningful... You know, that's what- Yeah ... somebody in that sort of- It's not gonna be on par with... Yeah, it's a lot smaller. Yeah. Yeah, so there's few, very few actually type in that in. Yeah. And by the way, I also am, it, it's, I'm not doing apples and apples because it doesn't give me a topic.
Simon: I can only select search term, but now I'm comparing a search term to two, to two topics, and that's not... It's apples and oranges again. Yeah. So think we gotta... Well, look, I think we should just agree that I, I'm winning right now, and that, that's it. Case closed. GEO for the win. - And we do need to get onto our main topic. Of course. Which was about, what was it about?
Paris: Information gain. Yeah. so just a little bit of a, a shameless plug. I'm gonna be speaking at Brighton SEO in about a month, on October 8th and 9th, and my topic is information gain, and this is something, we, we've kicked it around in a previous, previous episode, Simon. I think it was even the first or second, LinkedIn Live we did when we talked about the kno- the importance of the knowledge base, and that's really extremely connected to the concept of information gain.
Paris: I think we even brought the term up, information gain, in that episode. But I am now, starting to build out the, the narrative of this, this keynote for information gain, and, I'd like to hear your thoughts. Right now what I'm thinking of here is that it is really the key ingredient, it's, it's the key anti-slop ingredient to ensure that you are giving, you're giving AI what AI is hungry for, which is new information, net new information that it has not already ingested, trained on, slurped up and it, it cannot find anywhere else.
Paris: And, we have developed this scoring system called an information gain score from zero to 100 inside of GEOforge, and I know you've seen it, Simon. Yeah. And that is actually the key metric for deciding, or let's say for forecasting the success of a piece of content, or even the attractiveness of a content topic recommendation. We look to that more than anything else. What is the information gain score?
Paris: If it's low, probably means that AI is, is gonna be bored with it and say, "already been there, seen that. Next." Or, you know, pro- I mean, it certainly wouldn't be cited for something that AI, has already learned from, from other sources. So when we see a high information gain score, we think, all right, this is something now where- Now, ho- hopefully this, this topic aligns with what people are searching for and prompting.
Paris: but this is net new knowledge, so this is where we wanna be. how do you see this, this topic- So, so- You know, when we- Yeah, so my, my... Yeah, my, my question would be, you know, almost from a layman's point of view, I understand, you know, the, the informa- the level of information that is out there on the web harvested into all sorts of different areas, not necessarily even public.
Paris: what constitutes information gain? So if there is, you know, there's a topic, I wanna write about it, and just put it... Give us some, give us, maybe you can give us an example in terms of, you know, were you an e-commerce brand and you're selling a product that, you know, is sold by a few others, how do I then put myself into a position where there is information gain in what I am publishing?
Paris: 'Cause typically, you know, I've, I've just been asked by a client to build, a website, but he wants it... There's a competitor. He wants basically all of the products that they've got on there. And yes, the AI that I've trained for his business will go and take all those products, reword it all, place it onto his listings, add stuff to it. Now, the question is- Yeah ... what do we have to add and how different does it have to be?
Paris: And how... You know, I'm not asking you to tell us how your score works, but there must be elements in there that are reasonably straightforward Yeah, really the-- there's, there's no secret. Our-- We have an information gain scoring agent, and that, and that agent has a very simple prompt instruction, and that is to, to estimate the, the degree of novelty contained in a, in an embedded chunk within a vector database.
Paris: So if- Yeah ... just to take a step back, our, our architecture is based on a RAG system, retrieval-augmented generation. So you put in a, you put in a file with knowledge, it could be, anything, a, a document transcript. the, the AI is gonna break that into, into semantic chunks, and those usually could just be short passages, each, very short passages that have semantic similarity.
Paris: And then every one of those chunks gets an IG score, information gain score, and then the, the aggregation of that, entire, corpus becomes the information gain score for that, for that whole document. But then when, when you go to create content, you're not creating content from that single document because the, the vectorization process itself is gonna distribute all those chunks that-- or they're called embeddings, across the whole vector database so that the content model can draw, it can pull from the entire database.
Paris: So it might, it might actually pull in passages from, a dozen or more different documents because these passages all share some type of semantic relevance. And so the outcome or the output the information gain that's outputted from a piece of content is something actually different. It actually could have e- even a higher information gain score because it has, sort of cherry-picked, high information gain passages from, throughout the entire vector database, and that's how we've architected it.
Paris: And the, the actual prompt is very simple. It really just says, "Estimate the degree to which this chunk, this passage, represents net new knowledge that, the, LLM, LLMs have not yet trained on or have not yet discovered through any public scor- sources." And that's basically that, that's it, and then you trust- So when- the model is smart enough to give it the right score.
Paris: It's not- So when you, when you, when you say... So the interesting piece for me is, yes, okay, so you're saying, "Go and find out whether this is unique or it's got something extra in it that you don't know," in simplest terms. Yeah. if you took the entirety of the web and every single webpage thereon and every piece of information, there are, and bear with me on this, there's a whole chunk of it that's not true.
Paris: There's a whole chunk of it that is not structured correctly i- as far as I know, right? Maybe you can correct me on that. But if you took a good chunk of, and let's guess, let's guess that it's 50, 60, 70% of the internet is not structured in terms of schema, et cetera. And I'd imagine it may well be even higher. So if we were looking at, if we were looking at an LLM going out and learning, as it goes, and it's saying, "Oh, this piece of inform- this is actually...
Paris: I've got something similar to this," but, and, and it will be this clever, it'll be going, "Right, something, I've got something similar, but within this piece there's a bit extra. Is that true? And is, can I, you know, verify that? Ah, right. So that piece is, that's good, so I'll add that to my knowledge base." Now, it's not gone and done that on everything that's on the web, clearly, because a lot of it, you know, if it's self-published or non-verifiable- Yeah, it, it's not, it's not verifiable.
Paris: Yeah. So- So it really won't, yeah, it doesn't have confidence. So the, the confide- the, the verifiable piece is huge. but are you sort of saying that if I, say I publish an article now that is completely unique, right? So in the example where I've, I've created a, an hour and a half long video, which is all about answer-engine optimization, what I feel it does, how you can work with it, et cetera, et cetera.
Paris: That's a completely unique piece of p- content. I've also referenced my book, many other videos. Yeah. I've written a huge chunk of text that goes with it. It's not to say that any of that is unique. Now, it's unique to me, but are we saying that it needs to be picked up by, some other sources? It's far more important it gets cited somewhere or someone else references it.
Paris: And especially a trusted domain, you know, that's even more so. - I don't... Yeah, I'm just trying to understand how much of something needs to be unique to be- ... information gain. You know, could you take two articles that were essentially the same and completely spin one and add a few bits into it, and it's therefore unique and there's something extra in it? I don't think so.
Paris: Maybe not. No, no. But I think this comes also- I think, .. Go ahead, yeah. No, I think it also comes to the question that a lot of people are asking, for example, on LinkedIn, is, you know, can I use AI to write content in the knowledge that that content is then therefore not unique, right? In theory. But if I've trained my AI correctly and it, it can talk like me, in, in theory, then it can be unique to an extent.
Paris: because it's not just pulling pieces in and cobbling it together like a jigsaw and going, "There's your article, Simon." It's using my tone of voice. It's wording it differently. I th- are we saying that information gain has to just then therefore be, you know, some extra information that I'm... 'Cause I could, you could, you could make stuff up at that point, and it's then unique- ... and information gain.
Paris: Whether it's true or not is another matter. - Yeah. Well, yeah, I think, one thing I'll say immediately is that information gain doesn't, doesn't guarantee truth. I think it just guarantees novelty, but that could be a complete fabrication. If, if you decide you wanna take an article, and, and duplicate it and change up the facts and, and present it as the complete antithesis, and you've made up everything and it's, it's just a completely fabricated, it would have a high information gain score, I guess, in, in GEOforge.
Paris: But in order for the models to trust it and believe it as truth, it needs to be c- corroborated and validated by other sources, beyond, the, the original article. And that's where- Yeah ... so, I mean, yeah, that, that's getting into a trickier territory because, information gain alone, it is what the models are looking for, but it doesn't guarantee that if you feed information gain that's gonna, that's gonna give you AI visibility because the models still need to validate and they need that confidence and, and your, your story needs to be credible.
Paris: The other, the other point I wanted to make is that even if you take a, a ha- the... One of the hardest cases is e-commerce. I'm selling something where I'm, - maybe I'm just a drop shipper and I've got these products on my website. I use the manufacturer's product descriptions. I mean, I don't really have anything inter- anything other than to say, I mean, these are the facts about the product, the specs, and I, I can't really add any information gain on top of that.
Paris: What I would say is that, what you could do is to try to get customers who have bought the product, of course, review, to review the product, but to, to try to get stories of use cases out of that product which have not yet- Yeah ... been created. So maybe you bought a pair of, of running shoes, and they're supposed to be the most comfortable running shoes, and you use them for three months and, and you got blisters and you had a different experience.
Paris: that's, that is a unique perspective, and it is your, a true, a true perspective. and your, your, a wr- your, - Firsthand experience with the product, which might be very different than what, what's expected, is, is information gain. It's, it's a unique perspective. It's, if you even think back to the EEAT acronym, the first E, I always used to say the first E is the most important one because it's, it stands for experience.
Paris: a lot of people can claim expertise. A lot of people can claim authoritativeness. Trustworthiness is, I mean, in the SEO world, it's really your domain rating. But experience is what you as a person have done with this product or, in a lived experience, and that, I think, is what's different. And so with e-commerce, and I would even say the same for SaaS, like Stitch, the, the customer reviews is one of the most powerful and impactful sources of true information gain because it represents the, the real lived experiences with the, with the product or with the...
Paris: Whether, whether it's SaaS, SaaS product or physical product. But this is where we're getting into a more difficult, like, area of trust. Because, for example, you guys asked me to give, a review on G2, right? So- Yeah ... I've reviewed- Thanks for that, by the way ... I have reviewed. Yeah, yeah. I've reviewed GEOforge, right? Of course. And I gave it a great review. All right, thanks.
Paris: Oof. Top, top. Full, full everything, right? But the actual process of giving that review is a nightmare, right? In terms of- Really? Oh. Yeah. Like, no, so it went to the lengths of, it interviewed me. It wrote, the AI wrote the review based off of what I'd said. Fine. And at the very end of it, it then, and this is the trust piece, it said, "Can you now please share a screenshot of you logged into the platform to prove- ... that you are using it?" Ah, great.
Paris: I didn't know that they, they did that. So, so that's great, but you've really got to want to do the review at that point. Yeah. Right? Well, I, I appreciate your review even more now, Simon. I didn't know that... I knew that it was about, it was a process with an AI interview- Which I think is, is an easier process than before, which is a very, very lengthy survey.
Paris: but I didn't know that you had to take a screenshot, but that's, that's good because that means that G2 is a trustworthy source that AI can trust. So if you do write a good review, then AI is gonna believe it and they're gonna like us more as a result. And we're, we're pushing really hard- Well, I was, I was talking to someone at length yesterday about Reddit.
Paris: So obviously a lot of the Reddit, weight or domain authority or whatever you wanna call it in AI has slipped. - ... and the person said to me, and this is somebody who's selling, a tool to rank in AI, because I get, I must get offered 10 a week, right? And I do go and turn, I go and turn up on, occasionally on meetings if I think that it looks reasonably unique, I'll go and see what they're talking about.
Paris: And this was a black hat tool that they said, "Right, okay, we have got hundreds of Reddit accounts where we typically were just posting into threads, - ... and, and commenting and then adding in keywords and rich information and all that into these threads." And they were finding it coming up in AI.
Simon: And I said, "Yeah, but that doesn't work now, d- does it?" And he said, "Well, now we're having to use aged Reddit accounts that have multiple reviews, and we've gone from, like, 500 Reddit accounts down to, like, 70." But this is the level of black hattedness or whatever you wanna call it- Yeah, it's black hat SEO that some people are going to. Yeah.
Paris: This reminds me so much of, this was maybe 15 years ago, but I remember I went to a s- there, there was a conference called Search Engine Strategies, SES. It was in London. Do you, you happen to remember that or have you been to any of those? - I've been... So yeah, I'll be, hopefully I'll be meeting you at, Brighton, but yeah. Yeah, yeah.
Paris: I think it was in 2012, SES London, Search Engine Strategies London. And, and I remember going to see a guy's presentation and it was, he had basically created thousands of, fake Twitter accounts. And then what he was doing was, it was kind of like an enormous, - PBN network or something like that. But he was using Twitter accounts, and this... And he was showing this off because it worked, and it was working really well.
Paris: All these Twitter accounts were, of course, they were tweeting about a lot of stuff, but a portion of, of their tweets were about the product or the brand they wanted to promote. Sometimes they would link, sometimes they wouldn't. And then the other, the other Twitter accounts would, would like and, and boost those promotional tweets. And, and so they were all just cross-linking, cross-boosting, cross-chatter- chattering, and it was this massive network, and it worked like a charm.
Paris: And that what you're saying is just, it sounds exactly the same thing. It's just another Reddit PBN which is now getting dismantled. Well, one, one- Jonathan, please I know one of the biggest, the biggest SEO companies in the country, and I'm not gonna name them. But the, the way they rank all of their clients is they have thousands of websites And they literally just go and post.
Paris: You know, those websites have been there 10, 15 years, and they have authority to an extent. Yeah. and it, it comes to the point of hopefully, and I do genuinely hope that we enter an era where AI can sort this all out. Because I've done it.
Simon: I've... 20 years ago I was taking an article, putting it through a spintax, producing two, three, 400 versions of it, sending it out to forums, posting it on forums with... 'Cause th- back then you didn't even need a CAPTCHA to log into a forum, you just had a bot that would do it and post it. And that was enough to get three, 400 links in a day to a site.
Simon: And of course, you then set it up so that it would drip them out at 20, 30, 40, and suddenly you ranked at the top of Google, right? Yeah.
Paris: And the, the question of information gain, that's why I said to you what makes it unique. I don't know whether, I believe that AI can actually filter that out well enough yet. 'Cause there's a lot of debate around, you know, on LinkedIn. You w- you know, you're po- you've got loads of people posting AI slop or whatever, or people just- being lazy, going to go to ChatGPT or Gemini and they go, "I wanna post today about this, this, and this," and it'll mo- it'll write them a LinkedIn post and they post it. 'Cause you can tell half the time it's got no spacing in it, it's got em dashes, it's got...
Paris: Oh, and then you've got people saying, "Well, just put some spelling mistakes in it or some grammatical errors, and that'll get you round the AI." Of course it won't, right? No, it can be very, you know- No.
Simon: But the argument to me is, okay, text-based, in reality, as long as I'm using my tone of voice, my knowledge base, everything I've fed into Claude for me, so I have fed over 500 videos that I've been in on YouTube into my Claude so it can talk like me- right? Presumably that's a lot more than almost anybody else has done, right? That's a lot. So does that mean that...
Simon: So can, can I then tell Claude to go and write just like me? I think I can, right? But by the way, Simon- And you've done it to an extent ... a quick question. You fed the videos, but, Claude needs the transcripts. Did you, did you have- Yeah, yeah ... them transcribed? Yeah, yeah, yeah. And the, yeah, and the transcripts, right. Okay. So all of that.
Simon: So it's got- Wow, that's a lot ... all of everything. So, so at that point, in theory, it can write like me. I personally believe that a lot of the pushback is against imagery. So just going into Chat or Gemini or whatever and making an image and, or Grok and make an image and use it with my post, that's probably the biggest tell that it's not- non-AI, right?
Simon: I would think. Now, and- Yeah, yeah ... but in your, in your e-commerce example that you talked about earlier- Let's say you've got 1,000 products on your website, and none of which are unique because they're not, because they're all bought from... Everybody's selling the same thing, let's say, right? Yeah.
Paris: I'm a drop shipper. Now, the, the differential that you can give is, like you say, reviews, great. But also I've been telling people for years, "Why don't you just make a video of the product?" Right? So you've got, let's say you've got 100 people selling the same products, 100 people sell the same 1,000 products on 1,000 website, on 100 websites, right? If you can make it a video about it, and it's a genuine new piece of UGC or user generated content that's talking about that product and its benefits, et cetera, and questions people ask, questions that people ask about it- ... chances are 99 of those other competitors haven't bothered to do that, and it's, that would be information gain, I would think maybe.
Paris: And that's why- Yeah ... I think we probably see YouTube as one of the biggest, sources of content that's getting pushed into the LLMs, right? Yeah, YouTube has a massive presence now in AI Overviews. It seems to be... The last statistic that I read was 20% of all of the YouTube- 20% of all of AI Overviews citations are YouTube videos. I'll bet that, that number has gone up since I read, that was a couple of months ago.
Paris: Yeah. 'Cause I'm seeing them now, not only in the citation list, but I'm seeing them- the, the videos embedded in the answers themselves- ... in the main area. And I'm also seeing, this is a, you know, Google is taking this in tiny incremental steps, but they are clearly marching towards AI mode as a default. They're now starting to expand AI overvi- AI Overviews by default, so you don't have to click the expand thing because now most of the time, you get the, you get the window that's taking up most of the above-the-fold real estate.
Paris: But you can still scroll just a little bit to get to the links. When you expand that, it takes up, you know... That, that thing gets to be four, four or five times bigger vertically and just completely o- overwhelms and crowds out all the links. And when that happens, I mean, whatever the small click-through rate, you know, the declining click-through rate, or SEO click-through rate, that's gonna go to almost zero when AI Overviews is expanded by default.
Paris: And then there also have been reports in the wild that Google is rerouting, Google searches directly to AI Overviews without the user even, selecting that. Oh, wow. 'Cause right now the, the user can select it as, and they, they can change the default Google setting and say, "I want AI Overviews to be my default search engine." Yeah. But until they do that, they're getting classic Google.
Paris: Now Google is starting to actually route Google searches directly to AI Overviews answers without the user's, request or even permission. So Google is taking these incremental steps towards AI mode. you know, traffic, human traffic to websites is only gonna continue to just drop, drop, drop. But it, the AI visits to websites will keep going up and up and up. And what does that mean?
Paris: That your overall organic traffic's kind of meaningless when, when you think about it because you really need to see how much of that is bot traffic. Probably already most websites have more than half of their traffic, their organic traffic, as bot traffic, and they don't even know it. Human traffic is, is declining, and the bot traffic goes up. And, I'm sorry to be so blunt here, but websites are just becoming databases for AI crawlers- Yeah ... mostly.
Paris: And- And, and the f- like, and the- The people who, the people who embrace that first- ... the top of the funnel to convert. Yeah. The people that, that recognize that, they, I mean, they, they, they're basically going to y- yield to that reality. and then they optimize for that, and they optimize the formatting and the accessibility and the structure. I mean, that's, right now, that's how you win with the tech- on the technical on page side.
Paris: And then of course, with off page what, like we talked about last time, the citation building is critical. we are now moving towards a, the, the new feature, the next new big feature for GEOforge is gonna be placing, Content Forge... Certain Content Forge pieces are gonna be, are going to be earmarked for, for pitching replacements. Right. And, in fact, the, the citation opportunity area in GEOforge within SiteForge, that's actually gonna be the originator of the opport- of the, of the pitch targets.
Paris: So SiteForge opportunities is gonna have some way to determine which sites are most receptive or should be most receptive to accepting and publishing third-party content, and those sites are gonna be identified. And then that, SiteForge agent is gonna talk to the Content Forge agent and say, "Hey, question.marketing has been publishing... In the last six months, they published a dozen pieces that appear to be from guest authors.
Paris: this guy, Simon Young, is a great guy to pitch, and right now this is a topic that he seems interested in, so let's, let's generate that piece of content. And instead of push- pushing it down the owned media path, we're going to divert it to the earned media path, and we're gonna actually target this topic, this piece of content for Simon at question marketing." And, that's what's coming next in SiteForge, and I think it's gonna be a big, big improvement in the citation win rates.
Paris: Well, t- t- talk- 'Cause the reality is, - ... talking to, talking to citations, again, I was pitched this week a tool to get citations, so which on the insi- on the outside looked great. Once you got on the inside, it was basically buy citations. So a list- Oh ... of websites- It's like, yeah ... on your topic. A database file you can just order. But not, not cheap.
Paris: Hundreds and hundreds of dollars per placement, which is ridiculous. You know, sh- surely genuinely earned citation through sharing something, worthwhile to an audience and then them publishing it is the real value. There are a bunch of these directories and, and I even have spoken to one of these agencies that is based here in Bulgaria, and it's all based on domain authority and domain rating. And that, that currency's going away because AI doesn't care what your domain authority is.
Paris: They care about your specific relevance. So all these link buying, marketplaces, basically, I think what you describe sounds like a link buying marketplace, the, the val- the value of these links and all those prices is based on, DR and DA, and those are, those are gonna fall apart. I think that that's gonna implode pretty soon because people aren't gonna necessarily need to go out and chase links.
Paris: eh, I mean, imagine a world coming very soon when Google is, Google is AI mode. Why, why do you need to rank? Why do you want links so that you can rank? Rank for what? It's gone. What you need is you need, you need mentions in the right type of s- sites. And if you just go around chasing domain rating, it's a, it's a fool's errand, I think.
Paris: So all, all those prices, I don't know, $500 a link because I'm a DA, I'm a DA 70 plus doesn't matter if it's, if it's not relevant. the AI's not gonna care about, about that. And that's a great thing for the industry. That's really moving us forward, I think. Hmm. Have you seen... Without putting a, an ad in, I'm not put- this is... Have you seen a product called Trustmary?
Paris: Do you know them? No, I don't think so. So they're like, they're based in Finland, and they're like Trustpilot, but they- ... supposedly have a, an easier review route than many of the other, you know, platforms.
Simon: But they were- ... saying their domain rating is in 80, 85, and a lot of their clients, I was talking to a guy called Johannes who runs it, and he was saying that this could be a route into AI in terms of trusted reviews, which are then- a, like asking someone a question to genuinely get feedback on a product rather than- ... you know, can you give us a review?
Simon: It's can we have- ... some proper feedback? - This sounds like- And then- I'm sorry, go ahead. Go on This, this sounds like the G2 process that was, it was a, it was a little bit excruciating for you, but because of that, that's the reason why AI can trust G2.
Paris: And if Trustmary has a... I mean, this is a great... This is, this is like review sites 2.0, AI native review sites. Now that you have AI, that you really have to, you really have to verify using AI, you have to verify that this person is who they say they are, that they really have purchased this product, they really have used the product. In the case of G2, send a screenshot, or here maybe, maybe you take a picture or something or, prove that you've been to the restaurant.
Paris: Like, that's the next layer that, that Yelp isn't, isn't on, Trust- Trustpilot's not on yet. No. And, I think that's gonna make these types of, th- this new AI first, review, review sites gonna be, these are gonna be a, a hit, I think But, or maybe you're gonna get to a stage where influencers are conducting their own genuine reviews, you know, to then feed that out as content.
Paris: Because, for example, somebody comes to me and talks to me about an AI tool, I'll gladly give them a review, but it, it will be a genuine review. And then at that point, it- does it then sit with how authoritative someone like me is to then talk on the subject? You see what I mean? And then if you get- Yeah if you extrapolate that out into other industries- ... you end up with a layer of people that are trusted and those that aren't.
Paris: Yeah. You know, there was an interesting thing that I r- one of my readouts from the, the GEO show this morning was about something that Profound... no, it's actually Air Ops. Let me, let me go there and try to find it, because it's b- I think it's pertinent to this discussion. Give me a quick second. I'm gonna go over to my briefing here, and I'm gonna...
Paris: Yeah, this is basically what I do with the, the podcast, is I go through, 10 stories that get served up in, in a brief to me every day. And, one of them is... No, actually, this was from Claude, so let me go over there and find it. Hmm. Okay, I don't think I'm gonna find it, but basically it says that the GEO is expanding, th- the entire discipline of AI search is expanding beyond just, let's say, on-site technical structure, publishing content, and getting citations.
Paris: It's going now beyond that. It's going to, s- social and UGC corroboration and also influencer marketing. But it, it actually pulled in, as another large pillar, it pulled in social and UGC, which is basically how does social media as a, you know, how does the voice of social media, the voice of the crowds speak about your product? And then how, how are influencers speaking about your product as well? 'Cause the- these influencers also have trust because they're validated by their own audiences.
Paris: And I think that part of the GEO/answer-engine optimization game is going to be beyond publishing this high information gain content, beyond distributing that content to, to high third-party trusted sites. The next step is gonna be influencing social media conversations and influencer marketing. So one of the things I'm already looking at is identifying... I had Claude go out and find me, 20 of the top influencers speaking about GEO and answer-engine optimization on LinkedIn and on YouTube, and I haven't done any outreach yet, but I've got the list, and I'm gonna start contacting them.
Paris: And, and I think that this might be a, actually a better go-to-market motion than, - Than advertising or maybe than, my own distribution efforts. Because, if, if someone like, Marques Brownlee, I don't know if he's the right guy, but you know the, the guy that, he's famous mainly for his Apple reviews, Apple product reviews. If he does, if he does a one-minute video on GEOforge, it's game over.
Paris: I mean, we are... We're golden. You know? The... I don't know how much he would charge. Maybe he'll charge me $100,000 for that, but I would get more f- from that than I would at plowing $100,000 into, into Google ads or into LinkedIn ads, I'm pretty sure.
Simon: So- Oh, I've, I have paid, I have paid some influencers in the crypto space 50 and 100K per video. Jesus. Yeah, that's... Well, with crypto, I mean, you can't advertise, so you have to go, you have to go to influencers, and they, they know they can command a... They've... I mean, they, they've got you over a barrel, so to speak. Well, have a, have a look at some of the genuine anal- analytics in the back of some of those channels and you might sort of scratch your head and say, "Hmm, is all of that real as well?" But, there- Yeah ... you know, there's a whole industry on the back end of that for manipulating followers and views and everything, so- Yeah that's pr- that...
Simon: Okay, we're talking about the real deep, dark, horrible hole of crypto, right? But, yeah. We're, we're, .. It's gonna be really interesting over the next year or so to see how it all pans out, but I think seeing- ... collective, collective groups of influencers in particular spheres, that'd be- a really interesting topic to talk about in one of our future lives. Yeah. Well, let's do influencers next time maybe.
Simon: We'll think about it. I, I think we both have s- hard stops coming up, Simon, so this has been another... I, I'm gonna stop, making the LinkedIn Live event 30 minutes because we just can't stop. maybe I'll just do it 45 minutes, but it's been fun as always, Simon. Great, great to see you. congrats on a productive week.
Paris: I'll be thinking of you when I, when I crack open my first beer in pr- probably an hour and a half or so from now. And, I'm looking forward to seeing you in Brighton in about a month, and until then we're gonna speak again on Tuesday as always, same time. and, and we, we will go probably longer than 30 minutes. but, it's always fun.
Paris: So have a, have a great weekend, Simon. You too. Have a good one. See you later. All right. Cheers.
