Latest / Semantic SEO Podcast / Microsemantics for AI SEO: Winning LLM Visibility with Semantic Triples
Transcript
- James Dooley: Why micro-semantics is important for AI SEO and LLM visibility? So, with regards to micro-semantics, that's been probably the biggest learning curve for me with the whole semantic SEO training, development, um the idea of triples or the semantic triples, um just the sentence structure, trying to be as concise as possible, removing the fluff. Kyle: Do you think that's cuz we're from the north? Um we're not really taught how to put a sentence together. James Dooley: Yeah, [laughter] yeah, I think it is, yeah. I think linguistics-wise, is we say certain things that it's just like, "All right." Kyle: [laughter] James Dooley: It doesn't kind of work very well from a linguistic point of view. I think a lot of time Kyle: We've been learning English. James Dooley: Corey, I'm I'm going, "Yeah, it's all good." And all everyone's name when I introduce, [laughter] Carl Hudson is not too bad. But when you've got certain names, I'd I'd pronounce them Kyle: I usually say that I'm going to butcher it. James Dooley: [laughter] Kyle: You tell me my golf that one. James Dooley: But with regards to micro-semantics, specifically, like I say, for SEO and for LLMs, so within you've got Claude, you've got ChatGPT, you've got Gemini. Everyone now is obsessed with all the different acronyms, whether it's LLM, SEO, AEO, GEO, whatever they want to call it. Why is micro-semantics very important? And I believe it's something that's actually not talked about enough. That sentence structure, making certain that each word is connecting, an entity to an entity, an entity to an attribute. Why is micro-semantics important for LLM visibility? Kyle: It's basically an LLM only or machine learning model will only it only works in black and white. It doesn't really understand the gray area. And that's the whole point of the sentence structure side when it comes like subject, predicates, objects, or RDF triples, cuz obviously there is a few other other types of them. Um that's where it eliminates that grayness. So, it's literally, you know, James is 26 year old. James Dooley: Well, that's not actually very good, is it? [laughter] Kyle: It's false. James Dooley: No, yeah, yeah. No, but I'm saying James you don't want to be putting James. It should be James Dooley. So, the actual full entity name. Where Give an example of where someone might be doing that wrong. So, James is actually wrong, but what other example Kyle: Paul, would it be where they might use it where it's an I or stuff like that? Explain why saying I or he or whatever is incorrect and how the way the LLMs are passing information, they don't know what that means. James Dooley: Yeah, so essentially if you imagine if you own Let's say it's your own website and it's the about page. Nine times out of 10, a lot of people go, um we we are a plastering company. We have been around since 1994. Who the hell's we? A LLM, yes, it's landed on you. It's landed on your web page. So, it understands that this domain is about something. But then it says it's about this we character. So, then it doesn't have fully understand and then you mention plastering. So, it's saying, right, well, there's a plastering entity here. But what's that got to do with 1999? Because you're saying we. So, it's all about connecting those dots. So, if it was James James James's plastering services, James's plastering service was established in 1999. James's plastering service is a two-man team. James's plastering service is owned by James Dooley. You know, that's how you would be doing it. The weird thing is it it looks a little bit spammy. Kyle: Yeah. James Dooley: Like from your old mentality of like SEO, you think you're just keyword stuffing. But the reality is you're not as long as you're defining attributes in in in like defining the attributes and entities more. Kyle: Yeah, it's like you what you need to do is you need to remove the entity out as being you can't keyword stuff the entity name. So, the entity name needs to be in there as much as possible. I saw people say, "Oh, but Google or Gemini or ChatGPT or Claude, they understand that the about page on that website is about that entity." Well, they can do. But, the issue is now the cost of information retrieval is going up cuz now they've got to try and work out who is we. And if they're doing any sort of grounding and chunking, and they want to try and take out a piece of information, if you go and take that paragraph or sentence out and put it into a Google snippet, or you put it into an LLM answer, "We do this, this, and this." Doesn't make any sense anymore. So, therefore, the semantic triple and defining the entity within the RDF triple becomes very important. James Dooley: It's also going to probably if you did that that's probably a bang-on example. If you did that, "We do X, Y, and Z." And you put that in probably in a quote, you're probably going to bring up another 4 billion websites that also say, "We do X, Y, and Z." Kyle: Yeah. James Dooley: So, again, there's it's ambiguity. Kyle: Yeah. James Dooley: Who is we? And you've just said you have 4 million websites. Kyle: That's that's a great word, the ambiguity, and the the uncertainty from the LLMs that they're not certain. How important is disambiguation and trying to make certain that the confidence and the clarity of your content is there because if they're not confident and there's not clarity, they're just not going to cite you. How important is the disambiguation and making certain How many times have we gone in and strengthened a KG ID score and a knowledge graph just by re- replacing I and we with the entity name, it's scary how and then how many people doing this on third-party websites saying, "We did this, this, and this. We won this award." And it's just like you're not doing the nodes and the edges. Don't like within the way the semantic graph works, you have to define the entity and connect it to the objects, the attributes, and everything else with nodes and edges. So, the easiest way of explaining it if anyone doesn't understand what I mean by nodes and edges is if you go and type in like knowledge graph visualization, you'll see a lot of dots and a lot of lines. And the lines are the edges, the dots are the nodes, are the entities that connecting things together. And it's not done in the triplets, either semantic triples, there isn't two connections in the line being done. And at that point, you're not feeding the knowledge graph, which then means you're not feeding the knowledge vault, which then means you're not feeding the training data, which then means the LLM visibility is nowhere near as good whether it's Claude, whether it's ChatGPT, or whether it's Gemini. So, it's crazy when people start looking at word level and sentence level James Dooley: But it can also just be like one word knocks it off. Like we've seen it like 50 points. Like one word will it can destroy the entire page that one word cuz it's created that ambiguity when it when it wasn't needed. And you thought, "Oh, I'll just quickly go and update this page." Kyle: Yeah. James Dooley: You know what I mean? And you think you're doing well cuz you you're getting some Google freshness. Kyle: So, I've got one last question for you because we we chat and we joke about this quite a lot. And with regards to when someone's looking to do backlinks, are they looking to do an off-page topical map, and they're looking to do third-party corroboration? Branded mentions now is brilliant for LLM visibility, right? And a lot of time now people saying you don't even need the backlinks that much anymore. There is still the still page rank algorithm, there's still you still want to be a link juice to increase the crawl budget. But how important is the micro semantics on backlinks, on press releases, on guest posts and get that micro semantics right on the off page just as much as the on page. Yeah, that micro semantics right, which is almost forming the connections that know the links are not needed cuz of the branded connections. James Dooley: I'd say for people who know us, this is going to be quite a big statement. So I'd say probably more important than building the actual anchor text link itself. Now, this is specifically within the realms of LLM optimization. Kyle: Yeah. James Dooley: Obviously when it comes to trying to rank higher on Google, I'd still argue yes, it's very important and will help, but you still need that link. Kyle: Yeah. James Dooley: Um but yeah, when it comes to the LLM optimization, utilizing the micro semantics and semantic SEO, it is like fundamental. Um if you're structuring that page wrong, you're basically just adding you're you're building the um ambiguity if you're building the Kyle: Yeah, yeah, yeah, yeah, yeah. At times, I've seen people building links and it's weakened their knowledge graph score because now it starts to raise. We had one where one of my members of staff wrote an article about Fat Rank and it started doing connections and it then started to think that I wasn't the founder. It was connected it to somebody else because the way the sentence structure was put together, I was mentioned in the sentence, but it looked like one of the members of staff was the founder. And I was like, what's going on here? Like, it's not I am the founder. The sentence structure has been done isn't done in the correct manner. That should have been broken down into two sentences. You're throwing them both in together and the way that they've passed that with natural language processing it's had a detrimental effect and raised this ambiguity. James Dooley: found that on um niche edits. Obviously, you know, I'm not really a huge advocate of niche edits anyway. I think why do niche edits when you have a guest post? Um but we did actually find that in niche edits we built a link and it changed because it was a niche edit in a in like a related post, so you think it's good? It actually changed who who it thought we the founder of our business was. Kyle: Yeah, yeah, yeah. James Dooley: Which is like crazy. It was like, "What? Like, why is it saying that?" And then you'd find the link and say, "That's the one." And then you delete that out and it came back. Kyle: Yeah, yeah. James Dooley: Got rid of the disam- that that ambiguity in it. Kyle: Yeah, it's crazy. James Dooley: And that's like I'd say it's super important and obviously considering I enjoy building links, it's it's a crazy world now that, you know, it's almost like a mention is as powerful um as a link. Kyle: For LLM visibility, yeah. I still love the backlink when it comes down to Google rankings and page rank distribution, but yeah, like for me, micro semantics is almost like the anchor text of backlinks. It is key in making certain those branded mentions are done in a micro semantic way to make certain that it's feeding the LLMs, whether that's Claude, ChatGPT, or Gemini. There's many others as well. Could be Grok and other places, but getting those micro semantics for LLM visibility is key. Kyle, it's been an absolute pleasure. James Dooley: Cheers, mate.