In this LinkedIn Live talk, Mark Williams‑Cook addresses common misconceptions about schema markup, especially the claim that adding schema will directly boost AI citations or improve Google rankings. Citing an Ahrefs study and a comment from Andrew Kokstaki, he clarifies that Google explicitly states schema does not affect core ranking. He then walks through Google’s internal ranking architecture: the core ranking system (internally called “Mustang”) that evaluates site quality, popularity, and links; the query‑time re‑ranking layer (the “Twiddler” framework) that uses contextual signals such as intent, freshness, and predicted CTR to reorder results after a user’s query. He explains how the Knowledge Graph operates separately from these ranking systems and how schema helps Google recognize and disambiguate entities, making it easier for a site to be incorporated into the Knowledge Graph—particularly valuable for new or challenger brands. While schema isn’t a direct ranking factor, it’s a low‑cost, low‑maintenance way to improve entity clarity, increase the chance of Knowledge Graph inclusion, and indirectly influence SERP features. The talk concludes with practical advice on implementing schema and emphasizes that the primary benefit is entity definition rather than ranking.
Chapters
Hello everybody. I wanted to talk a little bit about schema and what it means to actually rank. So I posted this video yesterday or early this morning, which was about the Ahrefs study showing that schema didn't boost citations on AI platforms. And I got a really interesting comment today from Andrew Kokstaki, who said, but if we're saying adding schema markup won't get you AI citations and attributing that to snake oil sellers. But we're also saying clarifying entities and their relationships, especially for challenger brands, is good and useful and helps machines understand, which may lead to increased likelihood of improved ranking and also citations. Are we saying snake oil is good? Obviously not, but perhaps accidentally good, indirectly stumbling upon the right answer. Well, I think that's a really good, um, really good question. And. Probably one of the challenges of making a 45-second video around a really complicated topic. So I thought it was worth talking about. So I guess the first thing is I'm, I'm saying that if you're out there saying I'm selling geo, and to improve your visibility in AI, all you need to do is add schema, then I I say that's not true at the moment, based on the fact that there's absolutely no evidence that shows that's correct. If you want to say it improves ranking in Google, Google specifically says schema does not improve ranking. And again, we've got people like SearchPilot, like I mentioned in the video, that show that's true. However, as Gianluca was pointing out, it is still worth doing schema. So I think that's where the interesting discussion is, which is, well, if it's not improving ranking, it's not improving visibility in these cases, what is it for? So I thought I'd run through with everyone, we can have some fun in MS Paint. So I think this is a really interesting conversation about what we call ranking and what we don't, right? So if we have Google on this side, and this is Google's big algorithm that is doing everything to do with rankings, right? What that means to Google and what that means to us can be completely different. So rankings to Google, or ranking is their system, which I believe they call Mustang. I should probably change this color for accessibility purposes to make it easy to read. So I think this is called Mustang internally. And this is things that make your site rank better. So we're talking about things like Site quality, popularity, so things like links, things that can be directly attributed to your site and how well it should rank. At the other end of the, I guess, scale pipeline, whatever you want to call it, you have our very happy users, who are generally This is high pressure here, um, who are going to be doing searches. So that's a very happy user. As you can tell, it looks just like me. And they are doing a web search, right? So we don't know what query is yet. And this is kind of the point. So they're doing a search query right now. This doesn't work in that people do queries and then this just goes straight across to Google and then Google's like, okay, let me check what is the best ranking for your search query. There is a whole stage in the middle, which is pretty big and complicated in itself, which is to do with re-ranking. So we'll give that green. And what re-ranking is, and the interesting thing about re-ranking is it happens at— a lot of it, at least as I understand, happens at query time. So there's this moment after someone has done a search. So they've completed their search and there's information that's passed in that search query that's useful to determine what kind of result might be good for them that Google don't have in their rankings. So re-rankings can only be done after, in a lot of cases, the query's been done. So they get information like, is it a commercial query? Is it a local intent query? Does the query deserve freshness? What's the predicted CTR of this query? And based on this, there is a whole thing called the Twiddler framework, which sits at this front end of Google. And that Twiddler framework is essentially a bunch of independent little programs themselves that will re-rank the results. So there might be one for local. And so there's our little local one there, there might be one for if something needs to be— Pressure. So once Google's decided what the query is, these rankings that are passed are actually reordered. And that's called re-ranking. Now the confusing thing is from an outsider's perspective, all of this affects the SERP that we see. So we get our final SERP. This is our top 10 that we get here eventually, or whatever now AI adds. And then probably some kind of rankings afterwards. But because all of this part affects what we see here, there's a lot of pressure just to think about all of this whole process as ranking, right? Which isn't, isn't the same way at least Google speaks about it, which is important. Because this is where a lot of arguments have occurred where someone says, oh, click-through rate is a ranking factor. And Google was like, we definitely don't use that to rank websites. And they're speaking the truth because they're not applying your direct CTR to your, your site. They're looking at it in a broader, more aggregated sense as to what's expected for the search query. And then they're re-ranking results they had already ranked. So where does schema come into all of this? Well, there are lots of individual microsystems that come into the final SERP that we see. And we've got, we, you know, lots of people talk about Google's knowledge graph. And that kind of, again, sits, you know, over somewhere else on its own. And that's something Google builds over time. And when you do a search query, and this is interesting, when you do a search query, you can, before the search is done, Google's already deciding if it's picking something out of its knowledge graph for you, or whether it's like a document search. You can kind of see this when you start looking at Google Suggest. So you can see entities from the knowledge graph popping up there. And actually, I'm hopefully going to be doing quite an SEO talk on this. You can look at some of the network traffic and what's going on in Chrome DevTools and see the IDs that Google assigns these different search queries, because they have all different IDs. Here are some of the ones that I found already. So like ID 3, and it's type code for that query is trending. So it might deserve freshness. If they're a specific knowledge graph entity, if it's just a normal document search, all of these things are Things Google's calculating at the time the user is doing the query, right? So this knowledge graph is nothing to do with rankings in that sense. It's when this search query happens before the SERP is generated, Google's trying to figure out, okay, do I know about this thing? Is it a specific company? Is it a specific person? Do I need to lean onto my knowledge graph for an answer? Or do I need to kind of look at document rankings or a combination of both? This is, I think, the point Gianluca was making in that if you are already established in this knowledge graph, then adding kind of schema to your site or not isn't going to make any difference because you're already in there. Also, adding schema to your site is not a thing here where it's like, okay, well, they've got schema, so we need to rank their site better. It's not even a re-ranking thing where it's like, oh, it's got schema. It's a— everyone that is using schema to describe you is what's removing this ambiguity and strengthening your position so you can become an entity within this knowledge graph. And then the microsystems that all come into play at the time of the search query to figure out who should be surfaced and what sites connected to what thing and which things are connected to each other. Is all knowledge graph stuff. So it's not, I would say, really anything to do with ranking as such. And again, it's why Gianluca was saying that especially for new websites, for certain types of queries, you want to be in that knowledge graph. So doing everything you can do to try and get in there is going to be really helpful. And again, schema is one of those things where it's not particularly expensive to generate for most companies, unless you're a particularly sluggish enterprise. Once you've kind of done it, you've done most of the work, it tends not to need so much maintenance. So it is worth it in the long run. So yeah, that's what I wanted to say about that. Hopefully that's cleared up some questions and something you found interesting.
