How to Improve AI Search Visibility: GEO Methodology, Citation Tracking & Live AI Retrieval

Introduction: AI Search Demand in the US

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 0:00

Welcome, Chandan from Doodle Web. Thank you for joining us here on the NeuralAdX GEO and AI Search Strategy Podcast. Really pleased to see you. I understand you’re about eight hours behind us, so it’s about 7 a.m. your time?

Chandan — Doodle Web — 0:22

7:40, yeah.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 0:24

And you dragged yourself out of bed to have a podcast with me. What is wrong with you? You legend! I really appreciate it. You’re a true gentleman. I thought it would be interesting because, over in America, you guys had Google AI Mode a little over two months before its UK rollout. Google — U.S. AI Mode rollout Google — UK AI Mode rollout I think our listeners will be curious to get a brief understanding of the market there. You’ve effectively had an earlier start with AI Mode. How is the market maturing? What have you noticed with your clients and potential clients in terms of business interest? Has interest spiked, or is it still gradually catching on? What’s your view, Chandan?

Chandan — Doodle Web — 1:14

It’s a very hot topic in the US market right now.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 1:21

Glad to hear that.

Chandan — Doodle Web — 1:23

Everyone you speak to is interested in learning more about it. My current clients, previous clients and new prospects are all interested. Whenever we’re discussing new requirements and mention, “We’re also working on AI search for you,” they’re immediately interested. We build custom websites for businesses and work with a lot of technology companies. I’ve personally been doing this for more than 12 years. The demand is very high because the buyer experience is changing. People traditionally focused on SEO, but now, when you search Google, the first thing you often see is the AI-generated experience, followed by sponsored ads. Traditional organic SEO results are being pushed further down. That’s the shift we’re seeing, and businesses are increasingly asking, “How are you going to fix this for us?” That’s why we built our own internal tool called SiteSonar. Our process has four stages.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 2:58

Just to pause you there and give this a little structure, because we spoke about this just before the podcast and we’ve got limited time. I appreciate the insight into how the market is developing. My next question was actually about the methodology you implement for clients, which I think leads directly into what you’re saying about the SiteSonar software you use.

GEO Methodology: Measure, Diagnose, Fix & Re-Verify

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 3:05

Can you take us through some of that methodology?

Chandan — Doodle Web — 3:27

We measure, diagnose, fix and re-verify. There are a lot of tools in the market that simply tell you what to do. But the question is: how do you actually fix the problem and then verify whether the fix worked? We’ve built something around that entire process. You measure the current situation, diagnose the problem, make the changes and then re-verify the results.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 3:58

How does the measurement work? Let’s say you’ve got a client website page and the client obviously wants that page to rank or surface higher in AI search. What do you actually do, methodologically, using the software?

AI Citation Tracking: Prompts, Mentions & Competitor Positions

Chandan — Doodle Web — 4:15

The first thing we do is prompt tracking. We build a set of real buyer-style questions for that market. For example, for me as an agency, we might use a question such as, “Best web agency in Seattle.”

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 4:30

High-commercial-intent questions.

Chandan — Doodle Web — 4:37

Exactly. We run those questions against platforms such as ChatGPT, Perplexity and Google AI. For each query, we look at whether the company is cited, whether it’s mentioned, what position it appears in and which competitors are being cited instead.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 4:51

Do you track all of that through the SiteSonar system? So you enter the prompt, connect it with the client’s website and then monitor how well that website performs for that particular prompt. Essentially, it’s AI citation-tracking software. Is that right?

Chandan — Doodle Web — 5:15

That’s correct, yes.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 5:16

Fantastic. Let’s say you’re comparing the client against four or five competing companies and your client is coming fourth. Obviously, you want to move them further up the food chain. You’ve established their position through your measurement system. What do you then do to make changes that could influence how the AI engines respond to that client’s page for the prompt you’re measuring?

Why Competitors Get Cited: Reverse-Engineering Winning Pages

Chandan — Doodle Web — 5:55

That’s what we call the diagnosis stage. We ask: why aren’t you being cited? For every tracked query where the client isn’t cited, we fetch the pages the AI engines are actually citing for that question and analyse them. We examine things such as the headings, structure and content. The system produces information showing how many words or characters those cited pages contain, the number of data points, their structure, headings, comparison tables, pricing information and other relevant elements. That’s our diagnosis stage.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 6:44

So essentially, you’re reverse-engineering the company that’s come out on top. You’re identifying the differences between their webpage and your client’s webpage and then implementing improvements on the client’s page to give them the best possible opportunity of surfacing in a higher position. Academic research — What Gets Cited: Competitive GEO in AI Answer Engines KDD 2026 — SAGEO Arena

Chandan — Doodle Web — 7:07

Exactly. We then connect that to our resource hub. The resource hub is part of the SiteSonar software, but the content being created ultimately goes onto the client’s website. That’s how we close the gap. We look at two main types of gaps. One is technical. That might involve unblocking AI crawlers, generating or updating llms.txt and robots.txt files, fixing schema markup and automatically pinging IndexNow when content is published so it can be discovered. Then there’s the quality and content side. The cited page is analysed and we build a blueprint from the diagnosis. Content can then be created against that blueprint using things such as answer-first structure, verifiable facts, FAQs, internal linking and relevant brand information.

Technical GEO Fixes: Crawlers, Schema, llms.txt & Content Structure

Chandan — Doodle Web — 8:10

That content goes into an approval queue. You review it and approve it, and then it can be published on the website. After that, we re-verify the results. We automatically rerun the exact same prompts because the question set is fixed before and after the changes. We can then see whether a citation has appeared or whether the position has improved. That’s how we retest it.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 8:47

It’s really interesting hearing your methodology and application because GEO encompasses a wide range of skills and techniques, with technical overlap between GEO and SEO, even though they measure different output environments. Foundational GEO research — Aggarwal et al., KDD 2024 KDD 2026 — SAGEO Arena From what you’ve described, there’s a lot of similarity between your process and ours. We use our 11-Factor GEO Methodology, which began with the foundational GEO research and is now synthesised from a 13-paper academic evidence base covering retrieval, reranking, citation selection, source trust, content structure and repeated measurement. NeuralAdX — Academic Foundations of the 11-Factor GEO Methodology Foundational GEO paper — Aggarwal et al., KDD 2024 2026 critical survey of 45 GEO studies — Martinez When we obtain a new client, we also begin by running a set of high-commercial-intent prompts through AI engines. We observe the results and see where the client appears. Repeated, prompt-level measurement is important because generative-engine visibility is stochastic and can vary between engines and runs. 2026 GEO critical survey — Martinez If, for example, they’re appearing fourth and they want to reach first, we analyse the answer that’s currently surfacing in first place. We examine the citation sources, the information being used, the structure of the cited pages and other characteristics. Research on competitive GEO supports analysing retrieval, relevance, completeness, trust cues and the characteristics of sources that are selected for citation rather than treating ranking as a single fixed signal. What Gets Cited — 252,000 paired citation trials KDD 2026 — SAGEO Arena Then we reconstruct and improve the client’s page. The goal isn’t simply to replicate what the leading competitor is doing. We try to advance beyond it. We’re trying to future-proof the page to some degree by anticipating what could improve it further, because while we’re optimising to reach the current leader’s standard, they’re also going to continue optimising. So we try to be pre-emptive and add additional elements that might give the client an advantage. Current research supports stage-specific optimisation and repeated testing rather than assuming that one fixed set of GEO tactics will work permanently across engines. KDD 2026 — SAGEO Arena 2026 GEO critical survey — Martinez

NeuralAdX 11-Factor GEO: Live AI Retrieval & Benchmarking

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 9:45

That’s what we identify initially through our live AI retrieval assessment. We then connect the client to our AI citation benchmarking and AI answer visibility and share-of-voice benchmarking systems. NeuralAdX — AI Citation Benchmark NeuralAdX — AI Answer Visibility & Share of Voice Benchmark Like yourselves, we use predetermined prompts and continuously measure the feedback. If the client begins dropping relative to competitors, we can go back into the actual AI answer, analyse what’s changed and determine what needs to be improved. Hopefully, because we’ve already implemented changes, we’ll see movement in the right direction. We then reinforce that with live AI retrieval testing. We have examples on our website showing the tests we’ve carried out for NeuralAdX itself and the observed results. The evidence archive contains dated screen recordings, screenshots and platform-level results, while explicitly noting that these observations do not prove GEO was the sole cause of the outcomes. NeuralAdX — Live AI Retrieval Evidence Centre It’s fascinating because there’s a lot of correlation between what you do and what we do. And, to some extent, the US had an earlier start with Google AI Mode, which gave businesses and agencies there earlier exposure to the product. Google — U.S. AI Mode rollout Google — UK AI Mode rollout

Chandan — Doodle Web — 12:26

They’re definitely on top of it. Businesses really want to get cited. Another thing we do is educate clients using the data we’re gathering from live Google results and AI search.

AI Search Testing: Visibility Indexes & Changing AI Engines

Chandan — Doodle Web — 12:35

Google also provides data that can help you understand what people are searching for. You can use that alongside information about the business, ideal customer profile, competitors and brand to develop a better picture of the search environment.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 13:01

Do you guys do any additional testing specifically around AI search? For example, we run the NeuralAdX UK Business AI Visibility Index. It is a proprietary observational index that records dated retrieval snapshots across five AI platforms using a published 15-metric protocol. NeuralAdX — UK Business AI Visibility Index Protocol We use that to see which companies are surfacing in the top positions within particular sectors for high-commercial-intent queries. We record the results, assess them and observe the data. We’re essentially trying to continually learn more about how each AI engine works. We run those tests across platforms including Google AI Mode, ChatGPT, Microsoft Copilot, Claude and Google Gemini. NeuralAdX — UK Business AI Visibility Index Protocol We’re trying to understand their different behaviours because AI-search systems and their retrieval/output behaviour can change over time, and the same visibility assumptions should not be treated as permanent. 2026 GEO critical survey — Martinez Google — AI Mode continues to evolve You have to stay current and observe those changes through repeated live testing. 2026 GEO critical survey — Martinez Do you guys have anything similar underway, or is that something you’re planning?

Chandan — Doodle Web — 14:26

We already have something developing in that area.

Third-Party Citations, Directories & External Authority Signals

Chandan — Doodle Web — 14:30

We’re building a database around the differences we’re identifying. We’re looking at which businesses within the same industries are performing well and trying to understand why. It’s similar to what you’ve just described: creating our own system for understanding what is changing because the environment is constantly moving.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 14:58

Yes, exactly.

Chandan — Doodle Web — 14:59

For example, you might see changes in the prevalence of Reddit citations from one week to another. That’s why you can’t focus on only one factor. It isn’t simply a case of having a website, creating a microsite with resources and answers and expecting to immediately start surfacing everywhere. There are many other signals involved.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 15:20

There are a lot of third-party factors as well. For my own business, for example, there are important external sources and directories that can appear in the source ecosystems used by AI answer engines, so visibility cannot be treated as an owned-website-only problem. Generative Engine Optimization at Scale — 100K+ prompt responses across 100+ brands 2026 AI Recommendation Study — third-party directories and ranking sites in live grounded answers

Chandan — Doodle Web — 15:35

Exactly. We have a listening radar within the system as well. It can identify engines where you’re not being cited and directories where the business doesn’t currently appear. We can then say, “These are relevant directories you should be listed on. Register the business there, come back and we’ll check again.”

Closing: Continuous Testing as AI Search Changes

Chandan — Doodle Web — 15:55

So, to answer your question, yes. We’re building and testing every day to understand exactly what’s being cited and how these AI engines are changing.

Paul Rowe — Founder, Chief Generative Engine Optimisation Officer & CEO, NeuralAdX Ltd — 16:05

That’s great. We’ll pause it there and definitely come back to you again in the future. I deliberately wanted this to be a relatively short introduction focused on one particular topic. I think with podcasts covering subjects like this, once you start going much beyond 20 minutes, some listeners begin to tune out. But we’ve reached the core of what I wanted to discuss: the methodology and the ways we go about improving results for clients. Fundamentally, for any potential clients listening, that’s what they care about. I’m glad we concentrated mainly on that. Chandan, thank you very much for joining me. That’s a wrap.