Semantic Content Networks for AI Search: SEO, GEO, Entity Clarity & Query Fan-Out

Introduction: Steven Dunlop and Semantic Content Networks

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Hello and welcome to the NeuralAdX GEO and AI Search Strategy Podcast.

Today, my guest is Steven Dunlop, Senior Account Manager at Epresence.ie, an experienced marketing professional who moved fully into digital marketing back in 2010.

Since then, he has built extensive experience across SEO and PPC, working with businesses in Ireland, the UK, the US and South Africa.

More recently, Steven has focused heavily on AI search, GEO and AEO, and has already increased Epresence’s AI visibility by 50% in just one month.

He is also a qualified marketing manager with a postgraduate diploma in digital marketing.

Today, Steven is joining me to help explain semantic content networks — what they are, why they matter and how they can help search engines and AI platforms better understand the relationships between your content.

This particular topic is something I’ve looked into many times, but there is a certain amount of intellectual bandwidth required to fully conceptualise and visualise exactly what it means.

I’m very happy today that you’re going to elaborate on it further, not just for potential listeners, but genuinely for myself, because I can still see that there’s more knowledge I need to draw from it.

So, with the actual definition of those words, how do you interpret semantic content networks so that you can fully understand them and then implement them on the websites you work with on a daily basis?

Steven Dunlop — Senior Account Manager, Epresence.ie:

Thanks for the introduction and the invite, Paul. It’s a pleasure to be here.

What Semantic Content Networks Mean

Steven Dunlop — Senior Account Manager, Epresence.ie:

The way I understand it is that, as we move from traditional SEO into artificial intelligence, we’re moving away from simply thinking about keywords and pages and more towards meaning.

That meaning goes beyond individual paragraphs and pages. It looks at the whole website in terms of what it represents.

For example, if you’re a digital marketing agency, your higher-level entities might be your primary services — SEO, PPC, web design and so on.

AI looks at those entities to help determine who you are and what you do.

By bringing those entities together, it can recognise that you are a digital marketing agency, even without you explicitly having to say that in every place.

This happens at different levels.

You have your top-level pages, such as SEO, PPC and web design.

Under those, you can have secondary pages. For SEO, for example, that might include technical SEO, on-page SEO and off-page SEO.

Then you can take it a step further and create node pages underneath those.

Under technical SEO, for example, you might have pages covering Core Web Vitals, schema, site speed and sitemaps.

By structuring all of those things together, you’re immediately telling an AI platform much more about who you are and what you do.

The House Analogy for Website Entities and Internal Links

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That makes sense.

The way my mind tends to perceive it is that you have the website itself, and then individually you have each separate page.

Each page can represent a particular entity or subject.

For that entity to be properly understood, you need to provide the unique information that enables an AI engine to interpret what the page is about.

That’s primarily done through the specific content on that page, but also through things like entity clarity and entity disambiguation.

Those are the areas where my brain sometimes needs a more visual explanation.

So I’ve been trying to think about it through an analogy.

Imagine a semantic content network as a house.

The house is the website.

The rooms are the individual pages.

Then the things inside each room help define what that room actually is.

If you walk into a kitchen and see a fridge, a cooker and other kitchen equipment, those objects collectively help you recognise that you’re standing in a kitchen.

In the same way, the content on a webpage helps define the entity or subject of that page.

Then you have the doors between the rooms.

Those doors are like internal links.

If you have a kitchen next to a living room, it makes sense that those rooms are connected because there is a logical relationship between them.

It’s similar with a website.

When you link one page to another, ideally you want that link to connect to the most semantically relevant and useful page.

So the idea is to create a website structure where everything integrates in a way that is understandable to humans and machines.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

If you walked into a room and saw a fridge, a stove and a table, as a human being you would automatically recognise it as a kitchen.

AI works in a similar way.

The house is the website.

The room represents a main topic or page.

The furniture in that room gives further context.

And then what you might have inside the cupboards can represent the deeper node-level content.

So I think the house analogy works very well.

If I dropped you into a building blindfolded and then removed the blindfold, you could walk around, observe what was there and start recognising what type of building you were in.

You could walk into another environment and immediately recognise that it was a coffee shop based on what you saw around you.

AI is effectively doing something similar through the information architecture and content.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Exactly.

That’s what I find useful about the analogy because complex terminology can sometimes create unnecessary cognitive load.

My mind needs to create a visual image of what those terms actually mean.

Once I can create a coherent visual model, I can understand the concept much more deeply.

How AI Understands Website Structure

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

So, essentially, what you’re trying to do with semantic content networks is illuminate as clearly as possible what each page is, what it represents and how it relates to the rest of the website.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Yes.

At the higher level, the main entities allow the AI platform to recognise what the overall website is.

For Epresence, for example, the main entity pages help establish that we are a digital marketing agency.

Then the supporting pages provide deeper context around those areas.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That then leads nicely into something else we discussed earlier.

How Epresence Increased AI Visibility by 50%

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Can you explain what you did for Epresence?

You’ve achieved what I think is a very strong result by increasing your AI visibility by around 50% in the last month.

Whether that measurement is based on brand visibility, brand mentions, citations or a combination, a 50% increase is still significant.

So can you take us through what you did?

Because clearly something you implemented is working.

Building a Semantic Content Network in Practice

Steven Dunlop — Senior Account Manager, Epresence.ie:

There’s no magical secret.

It’s largely a lot of hard work.

A lot of websites historically had fairly flat structures.

You had your primary services on the main menu, and the URL structure might also be relatively flat.

Topic Hierarchies, Node Pages and Query Fan-Out

Steven Dunlop — Senior Account Manager, Epresence.ie:

With AI, we need to think more deeply about how information is organised.

So I started by identifying the main service areas and giving those high priority in the navigation and URL structure.

For example, services like SEO and PPC would sit at the first level beneath the domain.

Then I took the relevant subtopics and built those as secondary pages underneath the main service areas.

So rather than having lots of blog posts randomly distributed across the site, you start organising them logically underneath the relevant secondary pages as node pages.

You’re effectively building a network from SEO, through the different areas of SEO, and then into increasingly specific topics.

Using the house analogy, you’re moving from the house, into the room and then into the individual things inside that room.

There are a couple of reasons for doing this.

First, it creates a clear network that helps an AI platform understand which topics and pages you consider important.

Second, AI platforms do not necessarily perform only one search when responding to a prompt.

They can use query fan-out, where one user query results in many related searches or retrieval operations.

For example, somebody might ask for the best running shoes.

The system may also explore things like:

best running shoes for men; best running shoes for women; best running shoes for wet weather; best running shoes for different surfaces.

So if your site contains detailed, semantically relevant pages across those related areas, you increase the chances of being relevant not only to the original query but also to some of the fan-out queries.

That helps the AI platform understand who you are, what you know and the depth of your expertise.

It also forces us to think more carefully about the products and services we provide.

We need to be genuinely knowledgeable about what we publish.

And as you’ve said, we also need content that is trustworthy, citable and supported by useful research.

We can’t simply put keywords onto pages and expect that to be enough.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

So what you’ve realised is that this becomes a much more considered, long-term strategy.

It’s significantly different from the old mindset of simply publishing content and trying to optimise around keywords.

You have to step back and ask:

What are we actually dealing with here?

How does this system work?

Then you potentially have to restructure parts of the website and continue learning while you implement it.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Absolutely.

AI Search Optimisation vs Traditional SEO

Steven Dunlop — Senior Account Manager, Epresence.ie:

And it goes right down to page level.

You have to think about things like schema, FAQs, how the page is written and how the content is laid out.

You might need concise sections, strong headings and direct answers.

It’s a different strategy from traditional SEO, although there is still a huge overlap.

We still need SEO.

AI optimisation is another layer built on top of it.

And in some ways, it forces the industry to become more professional because low-quality approaches become harder to sustain.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

I agree.

Something I regularly think about is the idea that you should create content as if you were presenting it to a judge.

That’s a useful mental framework.

If you were presenting evidence in front of a judge, they wouldn’t simply accept everything you said without supporting evidence, proof, verification and credibility.

I think the same principle applies to content.

Alongside the infrastructure you’re describing — navigation, topic relationships, node pages and internal linking — you also need evidence.

That’s one of the key ingredients in generative engine optimisation.

How have you found that when dealing with clients?

Educating Clients About GEO, AEO and AI Search

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

A lot of businesses are used to traditional SEO.

Now you’re potentially having to tell them that things have changed and that additional work is required.

How do you navigate that?

Steven Dunlop — Senior Account Manager, Epresence.ie:

There are definitely people who have taken an interest in it and started to understand the opportunity.

But there are still many people who haven’t really seen it as a serious channel yet.

Internally, we also need to think about how we market the service compared with traditional SEO because it is almost becoming a separate service.

I see it as a layer above traditional SEO.

You still need the foundations of SEO, and then AI search optimisation and AEO are built on top of that.

As an industry, we need to work out how we explain that.

Clients sometimes assume that if you’re already doing SEO, everything connected with AI search should automatically be included.

But it can involve a lot of additional work.

So there is an educational challenge.

Measurement is also still developing.

With Google Search Console, people became accustomed to seeing impressions, clicks and average position.

AI search is more complicated because the measurement tools are still catching up.

It’s still a steep learning curve for everyone.

That’s why people like you and I, who have started working on it relatively early, can at least speak from direct experience.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That’s the fascinating part.

We’re working in an area where there isn’t decades of established knowledge.

Why GEO Requires Testing, Research and Patience

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

You have to investigate things yourself.

You have to conduct testing.

You have to read academic studies.

You have to extract what you think is useful, implement it and then wait to see what happens.

It isn’t necessarily an overnight process.

There can be a huge amount of work going on behind the scenes that nobody sees.

After six months, you might discover that five particular things appear to work.

But people don’t necessarily see how much research and experimentation went into reaching that point.

It requires patience.

You have to be willing to say: this is a long game, but I’m going to keep applying what the evidence currently suggests and continue adjusting as I learn.

And I think one of the reasons relatively few people are doing that deeply is because of time.

If you’re already running existing SEO campaigns for clients, where do you suddenly find the time to investigate GEO to that level?

Ultimately, you have to decide whether you believe this is going to become a significant market and whether you’re prepared to invest heavily in understanding it.

Measuring AI Visibility and the Limits of Current Tools

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

There’s a lot of best-practice advice available, but we don’t yet have decades of evidence where somebody can say, “If I press button X, I will definitely get result Y.”

We’re still building that evidence base.

But I can see results from much of what I’ve been doing.

The difficulty is measuring the exact effect.

Google Search Console gives you relatively straightforward metrics for traditional search.

AI visibility is harder because the tools are still developing.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Have you tried some of the AI visibility tools?

I use Otterly.ai, for example, to measure things such as brand mentions and domain citations.

Are you implementing similar software with clients?

Steven Dunlop — Senior Account Manager, Epresence.ie:

We use several different tools.

Google Search Console is now showing some AI-related activity, although the reporting still has limitations.

We also use SE Ranking, which offers AI tracking capabilities.

We have another tool that provides AI-focused site assessments and reports.

And internally at Epresence, we’ve built some of our own tools using different AI platforms and prompts that we’ve tested over time.

That has been useful as well.

But it’s still an experimental area.

The tools themselves are changing almost week by week.

And prompt tracking is inherently more complicated than keyword tracking.

A traditional keyword might only be two or three words.

Prompts can be much longer, more nuanced and far more variable.

Prompt Variance and What Drove the 50% Visibility Increase

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Exactly.

Even changing one word in a prompt can change which brands appear.

That variance is significant.

But ultimately, you work with the best measurement methods available at the time.

Going back to your 50% improvement, would you say that increase is primarily connected to the semantic content network work you’ve been doing, or were there other implementations contributing to the result?

Steven Dunlop — Senior Account Manager, Epresence.ie:

Building the network naturally forces you to create more content.

And importantly, it forces you to create more nuanced, semantically relevant content.

As part of that process, we’ve also been looking at other factors such as schema and E-E-A-T.

Once you start building the semantic network properly, you inevitably start improving other areas of the website at the same time.

And when you get down to individual pages, you have to make sure those pages are optimised for AI retrieval as well.

How to Structure Pages for AI Retrieval

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

When you optimise a page for AI retrieval, do you have a specific process?

For example, imagine you’re building a key service page and you’re thinking about a prompt such as “best GEO agency in Ireland”.

Do you have a step-by-step framework for constructing that page around the type of information the AI engine might need?

Steven Dunlop — Senior Account Manager, Epresence.ie:

For a prompt like that, I wouldn’t necessarily try to simply declare that we are the best ourselves.

AI is intelligent enough to understand self-promotion.

There are third-party websites that publish those kinds of comparisons and lists.

So for some recommendation-style prompts, external validation can be very important.

In terms of page creation, our process is largely research-driven.

I often ask AI platforms themselves what factors they would look for when answering a particular type of query.

Then you can identify the relevant criteria and create supporting pages around those areas.

It isn’t necessarily a rigid formula.

It’s an iterative process.

You create something, assess it, improve it and keep refining it.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That’s similar to the framework I use.

Third-Party Validation and Recommendation Queries

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

I call it TCCEA.

I start with the prompt I’m optimising around.

Then I look at:

Target — what exactly is the user asking for?

Context — what is the context behind the query?

Criteria — what criteria would the AI engine need to evaluate an answer?

Evidence — what evidence can support those criteria?

And then:

Answer — how do I bring those components together into the page or article?

That becomes a workflow.

It’s something I developed because I wanted a repeatable way of thinking through page creation rather than simply copying somebody else’s advice.

And so far, it has worked reasonably well for some of the benchmark and methodology pages we’ve created.

But like everything else, if the results change, I’ll adapt the framework.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

Nothing is ever completely finished.

You have to keep adjusting.

Research-Led AI Content Optimisation

Steven Dunlop — Senior Account Manager, Epresence.ie:

And the old days of simply stuffing keywords onto a page are definitely gone.

Even small things matter.

Spelling mistakes, vague wording and poorly written content can reduce credibility.

You have to be precise.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That reminds me of YouTube transcripts.

People publish a video and rely entirely on the automatic transcript.

Why Accuracy, Spelling and Transcript Quality Matter

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

But automatic transcripts frequently contain errors and colloquial speech.

If the AI engine is trying to interpret that content and the transcript is messy, incoherent or inaccurate, you may be making the content harder to understand.

So even if the video itself is excellent, you still need to clean up the transcript.

You need to make it coherent, technically accurate and universally understandable without making it unnecessarily complicated.

It’s another example of additional work that people often don’t realise matters.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

You have to look at the page and actually read it.

Ask yourself whether it comes across clearly.

If you phrase things vaguely or with too much uncertainty, AI may not treat the content as strongly.

If you are presenting a factual claim, you need to be clear about what the claim is and where the supporting evidence comes from.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That ties directly into authority.

AI retrieval often works at passage level.

Passage-Level Retrieval and Direct Answers

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

So if you have an H2 and then immediately provide a direct answer underneath it, you’re reducing the amount of work required for the system.

If your content precisely matches the user’s intent, is semantically relevant and easy to understand, then the AI system may be able to retrieve that passage and use it as part of its answer.

That’s the ideal scenario.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

Trust, E-E-A-T and AI Source Selection

Steven Dunlop — Senior Account Manager, Epresence.ie:

AI isn’t lazy, obviously, but it does operate under constraints.

From the moment somebody enters a prompt, the system may have to search, retrieve information, evaluate it, synthesise an answer and return it quickly.

So your website needs to be as easy as possible for these systems to process.

It needs to load quickly.

The answers need to be clear.

The structure needs to make sense.

And another interesting thing is that an AI system may not always return the answer that is technically the most accurate.

It may favour the answer or source it trusts most.

That’s why E-E-A-T and trust signals can be so important.

One area I’ve been working heavily on is personal profiles.

We often take team profiles for granted, but they’re becoming increasingly important.

Author Bios, Team Profiles and Expert Entities

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Are you talking about author bios?

Steven Dunlop — Senior Account Manager, Epresence.ie:

Author bios, yes, but also the wider team profiles.

We’ve connected team profiles and author bios.

We’re starting to use specialist authors for relevant pages.

My own specialist areas include SEO, GEO and AEO, so my profile reflects those areas, my experience and my qualifications.

That helps establish expertise.

And that’s one of the reasons conversations like this are useful as well.

The more I talk publicly about a subject, the more external references there are connecting my name with that expertise.

AI search increasingly operates across many different online environments.

Wherever it repeatedly sees your name connected with a particular area of expertise, that can contribute to the wider picture of who you are.

Building Authority Through Third-Party Mentions

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

And that becomes even stronger when it happens across third-party platforms.

Something like this podcast creates another independent context where your expertise is visible.

You’re appearing somewhere else, discussing a specialist subject and demonstrating knowledge.

The more frequently that happens across different sources, the less ambiguity there is about your expertise and professional identity.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

The days of SEO specialists sitting quietly in a dark room at the back of the office are disappearing.

We also have to get out there and demonstrate that expertise publicly.

We can’t simply hide behind the website anymore.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Which can be uncomfortable.

But in some ways that’s also valuable.

It forces you to grow.

I’m not naturally the most outgoing person either.

I’d probably describe myself as an ambivert.

But because of the way AI visibility works, I can’t necessarily build the type of business I want to build without appearing on platforms and having conversations like this.

Sometimes you simply have to go into situations that feel uncomfortable and trust the work you’ve already done.

Once you get into the conversation, all the research, experience and knowledge you’ve built starts to come through naturally.

And when you step back and look at the algorithms themselves, the complexity is incredible.

Steven Dunlop — Senior Account Manager, Epresence.ie:

I think we’ve reached the point where no single person necessarily understands every part of what is happening.

The systems are simply too complicated.

The Complexity and Rapid Evolution of AI Search

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That’s how I imagine it as well.

There are probably teams working on different areas — trust, entity understanding, semantic meaning and so on.

The more you learn, the more you realise how much you still don’t know.

Somebody from outside the industry might hear us talking and think we know an enormous amount.

But when you work inside it, you’re constantly aware of the depth that remains.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Absolutely.

I know what I know, but I also know how much I still don’t know.

And the rate of development means there is always something new to learn.

When you eventually combine AI with things like quantum computing, the potential becomes even more dramatic.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That acceleration is something I think about a lot.

People imagine technological improvement as a smooth progression.

But recursive improvement could potentially produce periods where progress accelerates dramatically.

Something that historically might have taken years could potentially happen much faster.

That creates huge uncertainty.

All you can really do is remain disciplined, continue researching and apply the best evidence available.

Trying to keep up with every single development in AI would overwhelm anyone.

There is something new almost every day.

And especially in our industry, new research is constantly emerging.

I’ve spent an enormous amount of time developing our own GEO methodology.

Then new academic studies appear that challenge parts of it.

At that point, you have to be prepared to change the methodology.

You have to be honest and say: this new evidence suggests something different, so the framework needs updating.

That can be frustrating because you may have invested hundreds of hours developing something.

But that’s the reality of working in a rapidly developing field.

I would rather adjust the methodology based on evidence than pretend the original framework was perfect.

Why AI May Reduce the Impact of Black-Hat SEO

Steven Dunlop — Senior Account Manager, Epresence.ie:

I think that approach will ultimately be rewarded.

Before AI search, Google spent years trying to deal with black-hat SEO techniques.

There were constant updates trying to combat manipulation.

But you could still sometimes perform a Google search and find a site ranking at number one even though it appeared to go against a lot of recognised best-practice guidance.

I think AI has the potential to shift things more towards rewarding genuine expertise, effort and useful content.

AI systems can potentially identify manipulative approaches much more quickly.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That’s certainly the direction I see as well.

Truth, evidence and trustworthiness become increasingly important.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Trustworthiness is a huge word here.

How Epresence Builds Trust With AI Engines

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

So, bringing that back to Epresence, what have you done specifically to demonstrate to AI engines that Epresence is an entity that can be trusted?

What kinds of trust signals have you been building?

Steven Dunlop — Senior Account Manager, Epresence.ie:

It’s multifaceted.

There isn’t one single thing.

We’ve upgraded author profiles and team profiles.

We’ve added qualifications and areas of expertise.

We cite facts and link to research.

One of the biggest things is doing our own research.

AI systems appear to value useful original research.

We’ve had particularly good results from producing in-depth, factual, Ireland-specific research.

A lot of industry data is based on the US.

For smaller countries such as Ireland, there can be fewer high-quality local datasets.

So producing our own research gives us something genuinely useful and distinctive.

We’re also connecting individual team members with their specialist areas.

My profile is linked closely with SEO, AEO and GEO.

Other colleagues are connected to their areas of expertise.

It’s about building trust through experience, expertise, authority and trustworthiness.

You follow recognised best practices and demonstrate to Google and AI platforms that you genuinely understand what you’re talking about.

E-E-A-T, Independent Mentions and Search Everywhere Optimisation

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

And that brings us back to E-E-A-T.

Some people assume that because GEO is developing, certain traditional SEO concepts have become less important.

But in many cases the opposite may be true.

You’re not abandoning those principles.

You may actually need to execute them more diligently.

And then third-party independent mentions become another important trust signal.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

SEO itself is evolving towards what people increasingly call Search Everywhere Optimisation.

It isn’t just traditional search engines anymore.

AI platforms can draw information from across the wider web.

YouTube can matter.

Reddit can matter.

Other industry websites can matter.

And unlike traditional SEO, the value of a third-party mention doesn’t necessarily depend entirely on whether there is a backlink.

If your company or name is being mentioned in a relevant authoritative context, that can still contribute to how systems understand you.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That’s where I think it becomes useful to distinguish traditional SEO authority from what I describe as GEO authority.

SEO Authority vs GEO Authority

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Traditional SEO authority has historically been heavily associated with backlinks and domain strength.

But GEO authority appears broader.

It can include things like:

the quality and uniqueness of your content; original research; statistics; factual evidence; live testing; video verification; consistent entity information; third-party mentions; and demonstrable expertise.

I use the term GEO authority because I needed a way of describing that broader collection of trust and evidence signals.

It isn’t an established industry metric.

But conceptually, I think it helps explain why a relatively young company can sometimes outperform much older competitors in AI visibility.

For example, NeuralAdX has been performing strongly against several established brands in the prompt sets we benchmark.

If you looked only at traditional backlink authority, you might not expect that.

So something else is clearly contributing to AI visibility.

Why Unique Research and First-Party Evidence Matter

Steven Dunlop — Senior Account Manager, Epresence.ie:

You used the word unique, and I think that is important.

AI is unlikely to be particularly interested in citing content that it could easily generate itself.

The more unique the content is, and the more specific it is to your own experience and research, the more valuable it can become.

Original research, proprietary evidence and first-party data can be extremely useful.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Exactly.

Because you’re effectively adding something new to the wider knowledge environment.

If you run a specific experiment that nobody else has conducted in exactly the same way, then you’ve created genuinely new information.

For example, if Epresence runs a benchmark in Ireland with a particular set of competitors, prompts, dates and measurements, nobody else has that exact dataset.

That can make it valuable.

Steven Dunlop — Senior Account Manager, Epresence.ie:

And that is especially important in markets where much of the existing research is US-based.

If you’re operating in the UK or Ireland, producing your own country-specific research can fill a real information gap.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

And potentially that research can then become useful internationally as well if it is relevant to a particular query.

That’s one of the strongest GEO strategies I’ve seen: producing first-party evidence, testing, statistics, tables and genuinely useful research.

AI systems have a lot more to work with when the content contains specific, verifiable information.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

And I’ve genuinely enjoyed learning and practising this.

GEO as an Ongoing Practice

Steven Dunlop — Senior Account Manager, Epresence.ie:

The word practising is important because I don’t think any of us will ever reach a point where we’ve completely finished learning it.

The systems are going to keep adapting.

So we have to keep learning with them.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Do you ever step back and look at how the systems appear to be designed and think: that actually makes complete sense?

Take third-party validation.

Of course a first-party website is going to say positive things about itself.

But if 20 independent websites consistently associate Steven Dunlop with strong expertise in a particular field, that is a more powerful external signal.

When you start breaking down some of these mechanisms, the logic behind them becomes quite elegant.

Different AI Platforms Use Different Sources

Steven Dunlop — Senior Account Manager, Epresence.ie:

What makes it even more interesting is that different AI platforms also use different sources of information.

So you have to remember that you’re optimising for systems that may behave differently.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

And their algorithms and preferences differ as well.

One thing I’ve noticed, for example, is that Google AI Mode seems relatively comfortable citing first-party content directly.

Google AI Mode vs ChatGPT Citation Behaviour

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

It will frequently cite NeuralAdX pages for certain queries.

ChatGPT can behave differently and may sometimes surface us through third-party references instead.

So you have to learn the nuances of each system.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Exactly.

And even within one AI platform, the behaviour can change depending on the prompt.

Some prompts can be answered from information the model already knows.

Other prompts require live web retrieval.

If you ask something like, “What is photosynthesis?”, that is relatively stable knowledge.

When AI Uses Live Web Retrieval

Steven Dunlop — Senior Account Manager, Epresence.ie:

But if you ask what time a particular film is showing at a local cinema today, the system has to access current information.

So the prompt itself influences whether retrieval is required.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Exactly.

Retrieval-Augmented Generation and Grounded Content

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

That brings us into retrieval-augmented generation and grounded information.

If the system doesn’t have enough reliable internal knowledge to answer a query confidently, it may need to retrieve current external information and ground the answer using sources.

There are enormous numbers of layers to all of this.

Steven, I think you and I could probably continue talking about it for hours.

But I’m mindful that you’ve got other things to do, so I’m going to bring things to a close.

Thank you very much for joining me.

I really appreciate the information you’ve shared.

Closing Thoughts and Advice for Getting Out of Your Comfort Zone

Steven Dunlop — Senior Account Manager, Epresence.ie:

Thanks for getting me out of my comfort zone.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Me as well.

It’s an ongoing process.

I think this is only around the fourth podcast conversation I’ve done.

The first one naturally felt more uncomfortable.

But you adapt.

You start trusting yourself more.

And eventually you realise that you’re more capable than you initially thought.

A lot of it comes down to trusting the work you’ve already done and allowing yourself to have the conversation.

So thank you again, Steven.

I really appreciate you joining me.

I’ll speak to you soon.

Take care.

Steven Dunlop — Senior Account Manager, Epresence.ie:

Thanks, Paul.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd:

Cheers. Bye, everyone.