Master AI visibility by understanding the Algorithmic Trinity

Jason says: “I'm going to explain how any marketer can optimize for AI-assistive engines.

People call it generative engine optimization; I call it AI-assistive engine optimization. Whichever name you use, the principle to SEO in the AI era is very simple, and it's called the Algorithmic Trinity.”

Why do you call them AI-assistive engines?

“They're assisting you to find solutions to your problems and answers to your questions.

Generative engine optimization describes how they function as opposed to what they do for you. You could also call them recommendation engines or conversational assistant engines, but they are assistants.

Google AI Mode, ChatGPT, Perplexity, and Copilot are all assistive engines. People don’t realise that you have explicit search like Google, and you have implicit research where you're researching a topic, and these assistive engines will bring you up in the conversation. For example, when you start talking about Knowledge Panels and it says, ‘Jason Barnard knows a lot about that.’

But there’s more! These AI assistive engines are being integrated into Gmail, Windows, Google Docs, Canva, etc., and they're going to assist people within the applications they're working in. That becomes ambient research, where your brand name can pop up in a conversation in somebody's workflow, and you have no way of knowing.”

What is the Algorithmic Trinity?

“When we talk about AI assistive engines, they all work on the same three technologies: knowledge graphs, large language model chatbots, and search engines. That's the Algorithmic Trinity.

Whichever flavour of assistive engine you choose, they're using those three technologies as the underlying way they deliver their answers. The difference between them is the extent to which they use each one in generating the answer.

In 2025, Perplexity is essentially a large language model reading search results and summarising them for you, with very little fact-checking using a knowledge graph. ChatGPT was essentially a large language model that then added search.

Google AI Mode is actually a mix of all three, because they use their knowledge graph to fact-check, the large language model Gemini to have the conversation, and the search results to get the most recent information.”

Does that mean that Google AI Mode delivers the best results?

“Google are way ahead of the field because they have the best knowledge graph. The knowledge graph is the encyclopaedic fact checker, so if you want to stop hallucinations, a knowledge graph is the solution.

Microsoft have a huge knowledge graph, Google have a huge knowledge graph, and I bet my bottom dollar that ChatGPT and Perplexity are building their own knowledge graphs. I've built a knowledge graph at Kalicube. If I can do it, they can do it. They've got a lot more money and a lot more super-talented Developers. (Editor’s note - in November 2025, after this interview was recorded, Andrea Volpini of Wordlift demonstrated that ChatGPT does, indeed, use a Knowledge Graph).

Then, the large language model gives the machine the capacity to have a conversation with the user. Fabrice Canel from Bing explained to me that Copilot is designed to bring the user down the acquisition funnel to what they call the ‘perfect click’. The conversational engine is saying, ‘Okay, what's your problem?’ It has a conversation with the user, digs down, and tries to figure out exactly what you're looking for.

It won't send the visitors to your website right away. What it will do is lead a conversation and figure out what the perfect click looks like for your particular problem. It will do that using, first of all, its own internal knowledge from its training data. It will then add search results to that to get up-to-date information that it didn't have in its training data, or to reassure itself that the training data is reasonably up-to-date and factually correct.

Then it will use a knowledge graph to fact-check anything it can, avoid hallucinations, and make sure that it is delivering the best solution to the problem.”

Is the perfect click the ideal user at the ideal time, looking for the ideal piece of content?

“Yes. For example, you (David) might be the ideal user for me (Jason) if you are asking ChatGPT, ‘Tell me about Knowledge Panels’.

ChatGPT will then say, ‘This is what a Knowledge Panel is.’ As it's saying that, it's going to explain it from the resources it already has. It doesn't necessarily know where it got the resource from, because it's part of the training data.

My aim, as a marketer, is for it to start associating my name with that topic. That's the topical authority that Koray GÜBÜR talks about, at the top-of-the-funnel. I need to be top-of-algorithmic-mind at that point, so that when the person says, ‘Who can help me with that?’ I'm one of the names the assistive engine will give the user. That's middle-of-funnel consideration.

Then the user says, ‘Okay, when push comes to shove, who should I choose?’, ChatGPT answers, ‘Jason Barnard, Kalicube,’ that's the perfect click. The user clicks on the link, goes straight to the service page, signs up, and buys.

They've been brought down the funnel and then convinced by the machine that Jason Barnard and Kalicube are the best solution for a Google Knowledge Panel.”

Why would ChatGPT build its own knowledge graph, and how can you ensure that it perceives your brand in the way that you want it to?

“The reason I believe they're building a knowledge graph is that they would be completely foolish not to, because their technology can't fact-check without one.

Looking at Google, with Search Generative Experience, Google was actually citing the knowledge graph as a source of information in Search Generative Experience back in the day, and that was their fact-checking.

The way to optimize for the Algorithmic Trinity is really simple, and that's the beauty of my approach. For an SEO, this is all going to be bread and butter because all three pillars of the Algorithmic Trinity use the same data source, which is the web. It doesn't actually matter to what extent ChatGPT is or isn't using a knowledge graph, because your web presence - your digital footprint – is going to feed the large language model, the knowledge graph, and the search engine in the same way.

The key to optimizing for AI Assistive Engines is to take control of your own digital footprint. You simply need to optimize your own digital footprint, which is your entity home website plus every single asset out there that talks about your brand.”

Why do you believe that you should duplicate your digital footprint on your own web property?

“I used to talk about entity home web pages, which would be the About page or the reference/canonical URL for that entity. John Mueller from Google talks about the point of reconciliation for an entity. The algorithms are actively looking for that.

It's the About page on your own website that you control, where you tell your version of the story, link out to all the corroborative sources, and create an infinite loop of self-corroboration. That is Kalicube’s incredibly powerful and simple strategy for getting into the knowledge graph and getting a Knowledge Panel. With that infinite loop of self-corroboration, focussed on the entity home/canonical URL/point of reconciliation, Bob's your uncle, you get a Knowledge Panel.

Anybody can have a great Knowledge Panel. It's just a question of whether Google understands you. The same is true with Bing. Does Bing understand you?

If you take that a step further, and think ‘My entire website is the source of information for my brand,’ it becomes the canonical resource for you as a person and/or your company. Then, you can do exactly the same thing with the entire website that you did with the entity home.

I can have a page on my site that lists the articles I have written, link out to them all, and show the algorithms that these are all indeed articles written by me. That reassures the algorithms that they have correctly identified me as the author of those articles on Search Engine Land, Semrush, Swydo, WordLift, or Authoritas, for that matter.

You can do the same with videos, webinars, and podcasts. What you are effectively doing is indicating all of the resources around the web that are indeed you, and linking out to them with relevant anchor text to explain how that fits into your digital identity.”

If they don't have an active knowledge graph of their own, how do AI engines decide what to include within the results, and how can you influence that?

“You can actually apply graph theory to all three: the knowledge graph, the LLM chatbot, and the search engines.

A knowledge graph is a practical example of graph theory. You have entities with attributes and relationships between the entities. The stronger, closer, and longer the relationship between the entities, the better. You build up confidence in understanding, within the knowledge graph, which is one of the ways you can expand your Knowledge Panel and get it to become both richer and trigger more reliably when somebody searches your name or your company name.

If you look at a large language model chatbot, you are a parameter in the LLM. The equivalent of an entity in an LLM is a parameter. You can strengthen that parameter over time by feeding more and more information that corroborates what you're saying. Then, the attributes of that parameter are other parameters, and then the relationships between the parameters are the vectorial space between them.

So, you can actually use the same optimization strategies for the LLM chatbot, the knowledge graph, and search engines.”

Can you measure the value of appearing in these AI search engines compared with Google search?

“There's a huge problem of opaqueness in these conversational environments, where they're bringing people down the acquisition funnel, because you don't know what the conversation looks like.

You can know what the kick-off question might be, but how the conversation then evolves is completely unique each time. So, you have no visibility and no way of tracking. What you can do is track terms or questions that you feel are likely to come up in the conversation.

Authoritas have an amazing tool for that that I highly recommend. We've been playing around with it, and it has an amazing capacity to identify who you're selling to, how you're selling to them, and what questions they're going to be asking. Then it tracks that.

However, you're just tracking a sample set of guesses rather than keywords we could accurately track before in Google. What we're doing at Kalicube is measuring the visits from ChatGPT and other AI assistive engines through Google Analytics. We have had a 31% month-on-month uplift just since last month.

What I've seen from clients is that the conversion rate from ChatGPT is between 10 and 20 times higher than it is from Google Search. You get fewer visits, but you get a much higher conversion rate. One client even said it was 100 times higher, but I find that hard to believe.”

Is this across different verticals and different categories of businesses?

“We're very much in the B2B space (we help entrepreneurs and their companies within the B2B space), so measurement is very difficult.

What I actually do now is, when somebody comes on a call is ask, ‘How did you find us?’ Pretty much everybody answers, ‘I researched the topic on Google search, then I used ChatGPT or Google AI Mode, and your name kept coming up.

Your red shirt kept coming up, and when I asked it, ‘Who do you actually recommend?’, ChatGPT said, ‘Jason Barnard, without a doubt.’”

Why do you think ChatGPT visits have a much higher conversion rate?

“The reason why the conversion rate is so much higher is because you're getting the perfect click. The ChatGPTs, Perplexities, and AI Modes of this world have guided the user down the funnel and presented them with what Fabrice Canel from Bing calls the ‘perfect click’. The conversion rate is significantly higher because the person is ready to buy. Also, the last thing they did before clicking on that link was get an explicit recommendation from the Assistive Engine, which they trust. They don't use the Assistive Engine if they don't trust it.

I'll give you a really quick example. I'm a bass player and I have a bass amp, but I wanted to play my guitar, and I didn't want to buy a new guitar amp.

• Jason: If I play my guitar through my bass amp, will I break the bass amp?

• ChatGPT: No, you won't break the bass amp, but it will sound terrible.

• Jason: Okay, how do I stop it sounding terrible and make my guitar sound really nice through my bass amp?

• ChatGPT: You need a compressor, you need an equaliser, and you need reverb.

• Jason: Okay, what do you recommend as a solution to that particular problem?’,

• ChatGPT: You need to buy three effects pedals for your guitar to use through the bass amp. Here are three really good suggestions…

• Jason: How much is that going to cost me?

• ChatGPT: $250.

• Jason: Can you find me something cheaper that's almost as good for somebody who isn't a professional musician?’

• ChatGPT: You can get these three pedals for $125, and it'll do the same job. If you're not a professional, you won't notice the difference.

• Jason: Where can I buy that?

• ChatGPT: If you're in America, buy it from Sweetwater. If you're in Europe, buy it from Thomann.

• Jason: Give me the link to Thomann.

It gave me the link. I bought the three pedals ChatGPT recommended. It was 20 minutes from starting with a problem to solving the problem with a purchase from Thomann in Europe.”

Do you see the majority of people moving to that kind of communication process in the future?

“People are already moving that way. I don't think it's going to be immediate because people tend to be slow to change their behaviour. Fabrice Canel, once again, was saying that supply creates its own demand. a well-known phenomenon in Economics (I've got a degree in economics, so it's something that really resonates with me).

When you provide something like a chat interface to have a conversation, it doesn't necessarily take away from the search. It creates a new use scenario that people will use in addition to the search they used before.

Little by little, people will transition across, and Google's going to move people across whether we like it or not. That idea of ‘supply creates its own demand’ is very powerful.

Chat interfaces are not yet replacing search. They're creating additional demand and an additional use cases. For example, the acquisition funnel, all in one session, guided by a trustworthy partner who will recommend the best solution after having understood my problem in detail, through a conversation.”

Jason, what's the key takeaway from the tip you shared today?

“The acquisition funnel is increasingly inside chat-based engines – ChatGPT, Perplexity, Google AI Mode – and you need to be top-of-algorithmic-mind throughout that funnel process in order to get the perfect click at the bottom of that funnel process.

To do that, you need to optimize your content for the Algorithmic Trinity: knowledge graphs, LLM chatbots, and Search Engines. To do that, you need to focus on your web-wide digital footprint (and not just your website!).

Your website is a representation of your brand that allows you to give your version of your story and point outwards to all of the information on the web that will help the algorithms answer the question and solve the problem, creating an infinite loop of self-corroboration.

If you diligently implement that strategy (that we call The Kalicube Process), then you will be top-of-algorithmic-mind throughout that acquisition funnel process, and you will get the perfect click, not your competitor.”

Jason Barnard is Founder and CEO at Kalicube. Find out more over at JasonBarnard.com.